Gaggia Classic - temperature study (long)

[Sparky]

Here is my report on the Gaggia Classic temperature stability during a
shot. The machine is slightly modified, with PID temperature control
and an aluminium block glued to the boiler to act as an additional
thermal mass. The PID control uses a K-type thermocouple mounted to the
top of the boiler under the metal bracket that holds the thermal fuse
in place.

Ambient temperature was 20 deg Centigrade during the course of the data
logging, which indicates the water temperature in the reservoir. All
data was logged using a single PID set point (99.5 deg C). All pours
were stopped after 60 ml was extracted, using a double basket. The
error with the volume was judged using a small glass that had
previously been calibrated. The main error was due to the crema, which
meant that the volume was usually below 60 ml (once the Guinness effect
had ended). The timing between shots was no less that 11 minutes and
usually 15 minutes or more to allow thermal recovery between shots.

The data was logged using a K-type thermocouple with an exposed bead
junction. The bead is no more than 1 mm in diameter. The thermal
response of this thermocouple was extremely rapid due to its small
thermal mass. The fast response is evident in the data, so there is no
need to quantify it further. The data was logged on a QM1538 digital
multimeter with RS232 output. The QM1538 outputs data at 0.4 second
intervals. A data logging routine was written in Matlab, which logs the
data and corrects for the nonlinear thermocouple response. The
thermocouple was calibrated against a digital thermometer, which was in
turn calibrated in an ice/water mixture and in boiling water (sea
level). A non-linear curve fitting routine was then used to provide a
correction function. The TC reading is estimated to be accurate to
within +/- 0.2 C following this procedure.

The data was logged for two different cases; directly pouring the shot
and using a shot pre-infusion period (to allow the boiler to recover
from the large thermal load that occurs due to the rapid flow of water
into the puck during this period). In all cases, the basket was filled
to the point of touching the shower screen. A standard Gaggia double
basket was used. The filter basket was removed from the porta-filter
for filling with grounds, while the porta-filter was left locked onto
the group at all times, to maximize thermal inertia.

The thermocouple was situated on top of the grounds and was introduced
over the rim of the basket. When the porta-filter was locked in place,
the thermocouple wire formed a good seal with the lip of the basket and
the rubber group gasket. In most cases, no leaking was observed,
otherwise only a few small drops (2-3) of water slowly escaped via the
seal (representing a negligible increase in flow rate).

The results are presented graphically and can be found here:
No pre-infusion

Loading Image...

Here we directly see the shocking temperature stability during a shot.
While a range of different flows are indicated, the fastest flow
results in a drop of approximately 7.5 C during the shot. All the other
shots resulted in 9-10 C temperature variations.

With pre-infusion

Loading Image...

We see that pre-infusion helps slightly to reduce the temperature
variation during the shot. Here it ranges from 5 C to 9 C. It may be
possible to reduce the variation down to 3 C if ristretto pours are
used. However, the pre-infusion step adds variability to the PID set
point temperature, resulting in less accuracy in the starting
temperature. It may be possible to minimize this effect with good
repeatable grind and tamp as well as accurate pre-infusion timing.

A group head temperature measurement was also performed by placing the
thermocouple up next to the group casket. It shows the effect of
brewing on the group temperature outside the porta-filter basket.

Loading Image...

Here was see that the rapid fall in temperature of the group head. The
effects of pre-infusion and brewing are clearly visible in the data.

The results presented here would be even more evident in an unmodified
Classic, indicating that temperature stability is clearly not possible
with this machine. The likely culprit for the instability is the
boiler-on-group design with the cold water injection from the bottom
between the boiler and the group. Clearly, the large thermal load from
cold water injection is transferred directly to the group as well as to
the hot water drawn from the top of the boiler.

As for PID modifying a Classic, it does have a noticeable effect on
shot-to-shot repeatability. Also the thermocouple mounting position may
also improve things along with aggressive PID parameters. It would seem
the best place to mount the thermocouple should at the bottom of the
boiler as close to the boiler-group junction as possible, to allow the
PID controller to best respond to the load.

Ultimately the poor thermal design of the Classic can best be remedied
by the use of pre-heated input water using either a heat exchanger or
second pre-boiler. This would remove the large thermal load during the
shot pour. The group could also be actively heated and PID controlled,
but this may only provide a diminishing return.

As for the taste, I had found the bitterness can easily be removed with
the PID controller, but that there was always some sourness that could
not be removed with the temperature set point. These results now
explain what I had been experiencing. If the bitterness is removed, the
shots invariably drop into the low temperature regime producing a sour
note.

As for my credentials: I am an experimental physicist with over 20
year's experience. I believe I followed a reasonably rigorous
experimental procedure for this study. BTW, not a single quality bean
was harmed in the pursuit of this data.

I hope people find this study useful when future discussions return to
the Classic vs XXXX debate.

Regards,

Mark.

[jim schulman]

Post by Sparky
I hope people find this study useful when future discussions return to
the Classic vs XXXX debate.

Have you tried resetting your PID for on/off control at about 1F
differential (p=0 for the Fujis)? The Gaggia makes up for its
deficiencies with a big honking heater that can heat the water at
espresso flow rates on the fly. The results you are getting are far
worse than were observed for the stock machine by the Norwegian

Post by Sparky
Se også vår test av
38 espressomaskiner.
Kanskje finner du din favoritt?

Then "resultater" to get the tabulations

[Moka Java]

Post by jim schulman

Post by Sparky
I hope people find this study useful when future discussions return to
the Classic vs XXXX debate.

Have you tried resetting your PID for on/off control at about 1F
differential (p=0 for the Fujis)? The Gaggia makes up for its
deficiencies with a big honking heater that can heat the water at
espresso flow rates on the fly. The results you are getting are far
worse than were observed for the stock machine by the Norwegian

Post by Sparky
Se også vår test av
38 espressomaskiner.
Kanskje finner du din favoritt?

Then "resultater" to get the tabulations

You might also consider that the aluminum block you added acts as a heat
sink and draws heat away from the boiler rather than add thermal mass.
Your recovery times seem very long.

Richard "yeah, right, like I might know what I'm talking about" F

[Ming W.]

Especially so as it has external rather than internal heating elements.

Post by Moka Java

Post by jim schulman

Post by Sparky
I hope people find this study useful when future discussions return to
the Classic vs XXXX debate.

Have you tried resetting your PID for on/off control at about 1F
differential (p=0 for the Fujis)? The Gaggia makes up for its
deficiencies with a big honking heater that can heat the water at
espresso flow rates on the fly. The results you are getting are far
worse than were observed for the stock machine by the Norwegian
Specialty Coffee Association: http://www.kaffe.no/ and click on the link

Post by Sparky
Se også vår test av
38 espressomaskiner.
Kanskje finner du din favoritt?

Then "resultater" to get the tabulations

You might also consider that the aluminum block you added acts as a heat
sink and draws heat away from the boiler rather than add thermal mass.
Your recovery times seem very long.
Richard "yeah, right, like I might know what I'm talking about" F

[Sparky]

Hi Jim,

I just checked the Norwegian site again and they are using 7 gm of
coffee (ie a single basket) that applies only half the thermal load of
a double so their 87-92.5 C temperature range seems to be consistent
with what I'm finding, if it means there is a drop of 5.5 C during the
shot. Otherwise you can't read much more into their data.

The PID controller can actively heat the boiler during the shot using
aggressive PID parameters. I suspect that the thermocouple needs to be
mounted near the bottom of the boiler to better sense the load and
compensate for it. I'll be trying that in the future.

Cheers,

Mark.

Post by jim schulman

Post by Sparky
I hope people find this study useful when future discussions return to
the Classic vs XXXX debate.

Have you tried resetting your PID for on/off control at about 1F
differential (p=0 for the Fujis)? The Gaggia makes up for its
deficiencies with a big honking heater that can heat the water at
espresso flow rates on the fly. The results you are getting are far
worse than were observed for the stock machine by the Norwegian

Post by Sparky
Se også vår test av
38 espressomaskiner.
Kanskje finner du din favoritt?

Then "resultater" to get the tabulations
--
jim schulman

[jim schulman]

Post by Sparky
I just checked the Norwegian site again and they are using 7 gm of
coffee (ie a single basket) that applies only half the thermal load of
a double so their 87-92.5 C temperature range seems to be consistent
with what I'm finding, if it means there is a drop of 5.5 C during the
shot. Otherwise you can't read much more into their data.

The only things that matter are the rate of flow (can the heat keep
up); and in the opposite direction, is the group losing heat so fast
that the inflowing water cannot warm it.

Since you saw the same temeprature drop in your choked 80 second
shots, the single flow rate is not helping your machine.

I doubt the heat loss from the group is a major factor; although
heatloss by the water to the group could be if the head is
inadewquately warmed prior to the shot.

The Gaggia is essentially a thermoblock machine done right -- but it
requires the fastest possible control to work.

Here's a simple test -- set your PID without alterations to 400C. This
will force the heat flat out on, even with sluggish tc placement and
parameters. Now measure the shot temperature. If the profile is still
falling, then the machine is inadequate. If, as I suspect it rises,
then your control system is the problem.

The US 700 watt Silvia cannot keep up, a flat out on boiler still
produces falling temperatures. My 1200 watt Solis produces rising
temperatures when I force the heat on. So I suspect the 1450 watt
Gaggia will too.

[Sparky]

It's not just flow rate, it's also total thermal load (ie integrated
flow rate). A single has nearly half the total load, so the low point
will be not as low as a double. My data shows that by simply looking
half way along the brew curve to an approximate 30 ml point. Also the
pre-infusion/wetting volume is less, so that should further improve
things.

I don't know about "thermoblock done right". Introducing the cold water
directly above the group which has excellent thermal contact with the
boiler is exactly counter to what a good thermoblock should do.

With better thermocouple placement and the PID controller set to on/off
control, the machine's tendencies towards wild thermal instability may
just be reigned in.

Can't do any more experiments now, as my wife is home today and she'd
object to me taking over the kitchen :)

[Sparky]

I'll return the Classic to thermostat control and log some pours when
the heating element is on to see how it can cope with the load. That'll
determine if it's capable of dealing with the system under load.

[Sparky]

Here is a follow up on my first report. The conditions were essentially
the same, but this time I investigated the ability of the Gaggia
Classic heating element to actively control the brew temperature during
the pour. This was done in a "zeroth order approximation" by making
sure the element stayed on during the pour. I used Jim Schulman's
suggestion and set a high temperature set point which I switched to
just prior to brewing. I chose a rather long pour in order to represent
the lightest load condition. Faster pours will invariably fare worse.
The PID controller reached it's higher set point in the first 5
seconds of the pour, at which point the element was heating the boiler
at full capacity.

Here is the result:
Loading Image...

I chose to use just this data set, as it is a best case scenario. All
my other pours (7 in total) were mainly too fast (because of the old
- extremely stale beans). It is clear that the high thermal load at
the beginning of the pour resulted in a rapid drop of the brew
temperature in the basket. The element was running at full power during
most of this region, and yet it clearly can't compensate for the
initial high thermal load. As the pour was quite slow and long, the
heating boiler eventually starts to modify and increase the brew
temperature in the filter basket, but it is driven to the point where
the steam thermostat shuts it off. Simply by looking at the slow effect
of an over-heated boiler on the brew temperature during an extremely
slow pour, it is clear that any active control of the brew temperature
fluctuations is impossible. It takes 30 seconds just to stop the drop
and start the temperature recovery. Any faster (usable) flows will only
increase this recovery delay beyond the 20-30 second brewing window.

My conclusion is that the feed water and the brew water are
inextricably thermally coupled in the Gaggia Classic. This isn't too
surprising in hindsight, as the feed water enters the boiler through
the bottom plate, which is made of high thermal conductivity brass and
forms the top of the group head. Thus the feed water actively cools the
group and brew water during the brewing process. The design, as it
stands, cannot provide a stable temperature brewing platform.

Finally I should add, that science aside, I enjoy the coffee from my
Classic. I think the combination of good beans (both fresh and tasty
blend), good grinder and good technique goes an extremely long way to
making a nice beverage, machine limitations aside.

[Sparky]

It's been pointed out that my machine may in fact be defective. That is
true. For instance I haven't removed the boiler to see if the brew feed
pipe is actually there. If this was missing, then brew and feed water
would actively mix....

I'd be happy to test any other Classics. If there's anyone in Brisbane
who would like their machine tested, let me know. It's entirely
harmless to the machine and it would aid the cause of scientific
repeatability.

[jim schulman]

Post by Sparky
My conclusion is that the feed water and the brew water are
inextricably thermally coupled in the Gaggia Classic.

You may be right, but I don't understand how it follows from the data.
If the inflowing water were affecting the temperature it would show up
later in the pour -- initially you're getting the temperature of the
group and the water in the boiler, only later does the inflow water
temperature's degraded the performance as it leaches heat out of the
system.

Your scenario can only be correct if the Gaggia is totally
misdesigned, so that the puck gets a shot of inflow water directly,
and the heated water only gets there later. Iirc, there is a tube that
takes the intake water to the top of the boiler (as in every other
home machine), so it has mix with the boiler's water in order to get
to the group at the bottom. If by some patented Italian piece of
quality control, you boiler is missing this, it could explain what you
are seeing.

To me this looks more like a very old nose -- the group or a pocket of
water just above the group that is at a very low temperature as the
shot starts.

[Sparky]

Hmmm. I've had a long hard look at the data and can define 4 regions.
The range of data makes it easy to inspect some of these regions.

Region 1 is before the brew switch is pressed and it's just logging the
temperature in the basket.

Region 2 is the pressing of the brew switch and the in rush of water to
the puck. This can be quite a considerable volume (est. 30 ml for a
double basket).

Region 3 is the short time after the in-rush of water before the pour
starts. Here the flow is extremely slow, as water is absorbed and
forced through the puck

Region 4 is the pour region, when you see crema laden coffee streaming
from the porta-filter.

Region 1 sets the temperature of the basket before water in injected.
In most cases the grounds haven't come up to group temperature. In
region 2, you can see the temperature leap up to brew temperature and
in many cases it either maintains this temperature or climbs further
for a few seconds more before starting to fall. This usually takes not
more than 3-4 seconds. So the initial large volume is NOT mixed with
the in-flow water according to this data. I'll call region 3 the
no-flow region. This is starkly illustrated in the pre-infusion data
set, where there is strictly no flow occurring. Here the temperature
falls in an exponential manner, indicative of passive cooling towards
thermal equilibrium. This clearly shows that a large heat sink is
present in or near the group. As no water is flowing, it can't be a
result of mixing. Region 4 is where the flow increases and the coffee
starts to pour. It has a humped shape in general, with thermal loss
increasing throughout the shot. This correlates with the flow
increasing steadily towards the end of the pour. In this region the
thermal loss is now being driven by the cold water in-flow.

The no-flow period in region 3 is also clearly visible in the group
head temperature data, where you see an exponential relaxation prior to
the brewing region. This really suggests that the heat sink is actually
in the group and not just heat lost to the puck (which I wouldn't
expect to show up so quickly in the group temperature). Then in the
brewing region you can see the group temperature being driven down.

To summarise; the large initial temperature rise and the few seconds of
steady temperature during the brief period of large in-flow of feed
water, really indicates that there is no substantial brew water/feed
water mixing. So until I can varify my findings with another Classic,
I'm optimistic that the data suggests that my Classic is at least
intact and not missing the brew-feed pipe or otherwise misbehaving.

So, yes, on the face of it, the Classic is a poor design for thermal
stability. I guess that's what happens when you mount your boiler on
top of the group and than inject cold feed water into the boiler/group
junction. You now have an actively cooled group!

[Sparky]

Post by jim schulman
To me this looks more like a very old nose -- the group or a pocket of
water just above the group that is at a very low temperature as the
shot starts.
--
jim schulman

Sorry, I just re-read this. Yes indeed, it does look like that.
Afterall, that's where the cold feed water is injected. (Sorry for
labouring the point)

If I can get to another Classics, I'll test them as well, but from what
I can determine from the data, my machine is in perfect working order.

Alan Frew ranks the ranks the Silvia above the Classic, stating that it
extracts just that little bit extra. He has tentatively attributed this
to better shower head water dispersion. I think that this little bit of
science has shown that better temperature stability may in fact be the
true culprit.

Now to fix my machine....

[D. Ross]

| Alan Frew ranks the ranks the Silvia above the Classic, stating that it
| extracts just that little bit extra. He has tentatively attributed this
| to better shower head water dispersion.

Note that this is based on the fact that he also ranks the Silvia just above
the *old* Silvia, with the former dispersion screen.

- David R.
--
Less information than you ever thought possible:
http://www.demitasse.net

[Coffee for Connoisseurs]

Every Silvia ever sold in Australia has the same water distributor & dual
layer dispersion screen, and has since day 1, sometime in 1998. This is
Rancilio's standard setup, see http://www.coffeeparts.com/ran/ran4.html.


--
Alan

***@coffeeco.com.au
www.coffeeco.com.au

[D. Ross]

"Coffee for Connoisseurs" <***@coffeeco.com.au> wrote:

| Every Silvia ever sold in Australia has the same water distributor & dual
| layer dispersion screen, and has since day 1, sometime in 1998. This is
| Rancilio's standard setup, see http://www.coffeeparts.com/ran/ran4.html.

Don't know what I was thinking of then. I was under the impression that
there was *some* change in the Silvia that improved its output 5-6 years
ago, but to be honest I was by then in the habit of tuning out of
discussions of Silvia technical minutiae.

- David R.
--
Less information than you ever thought possible:
http://www.demitasse.net

[Sparky]

Okay, it I get what you're suggesting. The on/off mode will throw the
heaters into full on during the shot if the temperature drops too low.
Good thinking, and I'll look into it.

Post by jim schulman
Have you tried resetting your PID for on/off control at about 1F
differential (p=0 for the Fujis)? The Gaggia makes up for its
deficiencies with a big honking heater that can heat the water at
espresso flow rates on the fly.
--
jim schulman

[Dan]

Post by jim schulman

Post by Sparky
Se også vår test av
38 espressomaskiner.
Kanskje finner du din favoritt?

Then "resultater" to get the tabulations
--
jim schulman

I like the comment in the milk steaming section:

"Bra skum"

chuckle..

[Jack Denver]

Thanks for posting this.

I'm not sure I understand what your pre-infusion technique was. Please
explain.

Schomer would certainly say that the temperature drop during the shot was
too great for sweetness to be achieved. Illy seems to indicate that some
drop is actually desirable as the cooler water prevents overextraction late
in the shot.

I think you've identified the source - in such a small boiler, there's no
way to prevent the cold feed water from intermixing with the boiler
contents, nor can the heating element keep up "on the fly". Although vibe
pumps are usuall rated for only very cool water, my Oscar preheats the
standing water in the tank up to around 130F just from waste boiler heat
whether I like it or not. It's been doing this for years w/o pump damage
(YMMV) , so a 1st step might be to start w/ lukewarm water in the reservoir.
Since the Gaggia heating coils sit on the surface of the boiler, it might be
possible to scavenge considerable preheat by wrapping a coil of copper
tubing around the surface of the boiler (as many turns as will fit). I
don't know if there is clearance inside the case for this.

Post by Sparky
Here is my report on the Gaggia Classic temperature stability during a
shot. The machine is slightly modified, with PID temperature control
and an aluminium block glued to the boiler to act as an additional
thermal mass. The PID control uses a K-type thermocouple mounted to the
top of the boiler under the metal bracket that holds the thermal fuse
in place.
Ambient temperature was 20 deg Centigrade during the course of the data
logging, which indicates the water temperature in the reservoir. All
data was logged using a single PID set point (99.5 deg C). All pours
were stopped after 60 ml was extracted, using a double basket. The
error with the volume was judged using a small glass that had
previously been calibrated. The main error was due to the crema, which
meant that the volume was usually below 60 ml (once the Guinness effect
had ended). The timing between shots was no less that 11 minutes and
usually 15 minutes or more to allow thermal recovery between shots.
The data was logged using a K-type thermocouple with an exposed bead
junction. The bead is no more than 1 mm in diameter. The thermal
response of this thermocouple was extremely rapid due to its small
thermal mass. The fast response is evident in the data, so there is no
need to quantify it further. The data was logged on a QM1538 digital
multimeter with RS232 output. The QM1538 outputs data at 0.4 second
intervals. A data logging routine was written in Matlab, which logs the
data and corrects for the nonlinear thermocouple response. The
thermocouple was calibrated against a digital thermometer, which was in
turn calibrated in an ice/water mixture and in boiling water (sea
level). A non-linear curve fitting routine was then used to provide a
correction function. The TC reading is estimated to be accurate to
within +/- 0.2 C following this procedure.
The data was logged for two different cases; directly pouring the shot
and using a shot pre-infusion period (to allow the boiler to recover
from the large thermal load that occurs due to the rapid flow of water
into the puck during this period). In all cases, the basket was filled
to the point of touching the shower screen. A standard Gaggia double
basket was used. The filter basket was removed from the porta-filter
for filling with grounds, while the porta-filter was left locked onto
the group at all times, to maximize thermal inertia.
The thermocouple was situated on top of the grounds and was introduced
over the rim of the basket. When the porta-filter was locked in place,
the thermocouple wire formed a good seal with the lip of the basket and
the rubber group gasket. In most cases, no leaking was observed,
otherwise only a few small drops (2-3) of water slowly escaped via the
seal (representing a negligible increase in flow rate).
No pre-infusion
http://i4.photobucket.com/albums/y129/mjfernee/Classic/Coffeeno-pre-infuse.jpg
Here we directly see the shocking temperature stability during a shot.
While a range of different flows are indicated, the fastest flow
results in a drop of approximately 7.5 C during the shot. All the other
shots resulted in 9-10 C temperature variations.
With pre-infusion
http://i4.photobucket.com/albums/y129/mjfernee/Classic/Coffeepre-infuse.jpg
We see that pre-infusion helps slightly to reduce the temperature
variation during the shot. Here it ranges from 5 C to 9 C. It may be
possible to reduce the variation down to 3 C if ristretto pours are
used. However, the pre-infusion step adds variability to the PID set
point temperature, resulting in less accuracy in the starting
temperature. It may be possible to minimize this effect with good
repeatable grind and tamp as well as accurate pre-infusion timing.
A group head temperature measurement was also performed by placing the
thermocouple up next to the group casket. It shows the effect of
brewing on the group temperature outside the porta-filter basket.
http://i4.photobucket.com/albums/y129/mjfernee/Classic/Coffee_brewing_in_group.jpg
Here was see that the rapid fall in temperature of the group head. The
effects of pre-infusion and brewing are clearly visible in the data.
The results presented here would be even more evident in an unmodified
Classic, indicating that temperature stability is clearly not possible
with this machine. The likely culprit for the instability is the
boiler-on-group design with the cold water injection from the bottom
between the boiler and the group. Clearly, the large thermal load from
cold water injection is transferred directly to the group as well as to
the hot water drawn from the top of the boiler.
As for PID modifying a Classic, it does have a noticeable effect on
shot-to-shot repeatability. Also the thermocouple mounting position may
also improve things along with aggressive PID parameters. It would seem
the best place to mount the thermocouple should at the bottom of the
boiler as close to the boiler-group junction as possible, to allow the
PID controller to best respond to the load.
Ultimately the poor thermal design of the Classic can best be remedied
by the use of pre-heated input water using either a heat exchanger or
second pre-boiler. This would remove the large thermal load during the
shot pour. The group could also be actively heated and PID controlled,
but this may only provide a diminishing return.
As for the taste, I had found the bitterness can easily be removed with
the PID controller, but that there was always some sourness that could
not be removed with the temperature set point. These results now
explain what I had been experiencing. If the bitterness is removed, the
shots invariably drop into the low temperature regime producing a sour
note.
As for my credentials: I am an experimental physicist with over 20
year's experience. I believe I followed a reasonably rigorous
experimental procedure for this study. BTW, not a single quality bean
was harmed in the pursuit of this data.
I hope people find this study useful when future discussions return to
the Classic vs XXXX debate.
Regards,
Mark.

[Mike Garner]

Post by Sparky
Here is my report on the Gaggia Classic temperature stability during a
shot. The machine is slightly modified, with PID temperature control
and an aluminium block glued to the boiler to act as an additional
thermal mass. The PID control uses a K-type thermocouple mounted to the
top of the boiler under the metal bracket that holds the thermal fuse
in place.

This proves exactly what? This is far from a stock Gaggia machine. I
agree with Jim's assessment that you're probably defeating the ability
of the Gaggia's boiler to heat on the fly, as well as having added a
heatsink with your aluminium block.

The results I've personally measured are in line with the Norewgian
study referenced by Jim as well. Actually, they are tighter than that
on my current machine.

As I said, this is what YOU are seeing on YOUR heavily modified machine.

Mike

[Randy G.]

Post by Mike Garner

Post by Sparky
Here is my report on the Gaggia Classic temperature stability during a
shot. The machine is slightly modified, with PID temperature control
and an aluminium block glued to the boiler to act as an additional
thermal mass. The PID control uses a K-type thermocouple mounted to the
top of the boiler under the metal bracket that holds the thermal fuse
in place.

This proves exactly what? This is far from a stock Gaggia machine. I
agree with Jim's assessment that you're probably defeating the ability
of the Gaggia's boiler to heat on the fly, as well as having added a
heatsink with your aluminium block.
The results I've personally measured are in line with the Norewgian
study referenced by Jim as well. Actually, they are tighter than that
on my current machine.
As I said, this is what YOU are seeing on YOUR heavily modified machine.

I agree that the value of adding the large heatsink is questionable.
it would have made a lot more sense to lag (insulate) the boiler. It
certainly would be interesting to remove the aluminum chunk and see
what effect it has.

Randy "q'na" G.
http://www.EspressoMyEspresso.com

[tckroll]

Post by Sparky
Here is my report on the Gaggia Classic temperature stability during a
shot. The machine is slightly modified, with PID temperature control
and an aluminium block glued to the boiler to act as an additional
thermal mass. The PID control uses a K-type thermocouple mounted to the
top of the boiler under the metal bracket that holds the thermal fuse
in place.

Thanks for putting this together. If the end result of further study
suggests that this machine may be less than ideal - so be it. However, I
think it premature to voice such strong conclusions.

Ignoring, for the moment that you and Mark, who comments below, have two
different machines, just what might account for such major differences
in machine behavior?

One thought would be to remove the boiler heat sink. A couple of years
back, via email, I talked with someone who had investigated this unit.
He partially decoupled the boiler / group head from the steel case. I
forget the details, though he used some sort of high temp plastic
washers and bolts to replace the group head to case connectors.

Post by Sparky
As for PID modifying a Classic, it does have a noticeable effect on
shot-to-shot repeatability. Also the thermocouple mounting position may
also improve things along with aggressive PID parameters. It would seem
the best place to mount the thermocouple should at the bottom of the
boiler as close to the boiler-group junction as possible, to allow the
PID controller to best respond to the load.

In case you want to try this, I have a exposed junction thermocouple
sandwiched between the boiler and group head of an old gaggia. Have
experienced no leaking.

As others have mentioned below, I would be interested in the effect seen
with the PID set to on / off control.

thanks again for the effort. please post whatever else you find.

regards

terry

[tckroll]

Post by Sparky
No pre-infusion
http://i4.photobucket.com/albums/y129/mjfernee/Classic/Coffeeno-pre-infuse.jpg

Sparky

I am not sure I reading the graphs correctly?

Are you using data from pours in the 60 - 80 second range? Are these
pours drinkable?
Would it not be more insightful to use data from pours with more similar
- standard - pour times.


thanks again

regards,

terry

[Cordovero]

I think the graph is saying that even in the first 20 seconds, he registers
a drop of 7 degrees Celsius, though I agree that I'm confused by the
function of the later times/arrows.

C

Post by tckroll

Post by Sparky
No pre-infusion
http://i4.photobucket.com/albums/y129/mjfernee/Classic/Coffeeno-pre-infuse.jpg

Sparky
I am not sure I reading the graphs correctly?
Are you using data from pours in the 60 - 80 second range? Are these pours
drinkable?
Would it not be more insightful to use data from pours with more similar -
standard - pour times.
thanks again
regards,
terry

[Sparky]

OK, my results certainly shocked me. I was aware of the Norwegian
results and was quietly optimistic that I would confirm what they have
published. I do notice that no experimental method was mentioned in
their study, so exactly how they logged their data is a mystery.

As for the "heat sink" I've attached :) A heat sink needs a high
surface to volume ratio. The Al block that is attached is a polished
block measuring 60x40x30 mm and glued on the 60x40 side. Hardly a heat
sink, but an additional thermal reservoir to passivly compensate for
load fluctuations. You can have thermal mass in metal as well as water,
and the metal can respond to temperature changes about 100 times more
rapidly. Personally, I noticed no change in the espresso's before and
after the block was added. Furthermore, the temperature drop in the
data is driven, not an exponential drop and so can't be the result of a
passive device. Periods of exponential fall are evident in the data at
periods when the pump is not on.

The PID controller is actively heating the boiler during the shot. As
I've mentioned, the TC is attached to the top of the boiler and during
a shot, the largest drop it registers is about 0.5 -1 C during the
initial high water flow period, then it stabilizes to about 0.1 C from
set point during the shot. From my study, mounting the TC close to the
group/boiler junction would allow the best active control during the
shot.

The flow rates that were used cover all useful rates for brewing. Even
though some shots were far longer than "recommended" they serve to
illustrate how the different thermal load rates affect the temperature.
The total integrated load is 60 ml of cold 20 C water (plus the initial
void filling/puck wetting volume) and results in essentially the same
final temperature in all graphs, just the rate of reaching this
temperature is varied. You may also notice that during the shot the
temperature drop curves downwards. This is a result of the flow rate
increasing towards the end of the shot and clearly shows that it is the
water that is driving the temperature down. The sharp drop in the
beginning is the result of the large flow rates (ie high thermal
loading) at the beginning of the shot.

By pre-infuse, I simply mean that the pump was activated for 3-4
seconds to fill the puck/group void with water and wet the puck, then
the system is left for a time to allow the PID controller to come back
to set point. Then the shot was poured.

I know that people will want to believe what they want. But I have this
machine and am interested in improving its design. The first step is to
study how it performs. It does make a brilliant cappuccino, and an
acceptable espresso. However, I had always noted that PID could remove
the bitterness, but not all the sourness. Recently I have been visiting
coffee shops known for their espresso and calibrating my taste buds :)
I noticed that their shots were neither bitter nor sour. I can't
achieve that.

If you take my results into consideration when thinking about the
boiler/group design of the Classic, you have to question the effect of
introducing cold water into the boiler/group junction. This should
actively cool both the boiler AND group, due to the high thermal
conductivity of the metal. I for one wouldn't think of designing a
thermally stable coffee machine that way.

As I said, I expect these results to be even more pronounced for a
"standard" Classic that isn't actively heated during the pour of the
shot.

If I were to do any follow up studys, I'd look at varying the TC
position on top of the puck, rather than have it near the side of the
basket. But for me, the data speaks for itself, and the machine is what
it is.

[Ming W.]

Is yours actively heated at full power during the pull or is it flashed by
the PID? The PID's temperature stability should only be a setpoint for
temperature at the begining of the pull isn't going to deal with
feedwater/boilerwater mixing very well.

Since the heating coils are on the outside, you *specifically* don't want
there to be too much thermal mass hugging the boiler stealing away the heat
if the coils are compensating for the cold feedwater, as the mass would
effectively steal the heat away from the boiler itself as metal does react
to heat quicker than water will - as you said yourself.

Post by Sparky
OK, my results certainly shocked me. I was aware of the Norwegian
results and was quietly optimistic that I would confirm what they have
published. I do notice that no experimental method was mentioned in
their study, so exactly how they logged their data is a mystery.
As for the "heat sink" I've attached :) A heat sink needs a high
surface to volume ratio. The Al block that is attached is a polished
block measuring 60x40x30 mm and glued on the 60x40 side. Hardly a heat
sink, but an additional thermal reservoir to passivly compensate for
load fluctuations. You can have thermal mass in metal as well as water,
and the metal can respond to temperature changes about 100 times more
rapidly. Personally, I noticed no change in the espresso's before and
after the block was added. Furthermore, the temperature drop in the
data is driven, not an exponential drop and so can't be the result of a
passive device. Periods of exponential fall are evident in the data at
periods when the pump is not on.
The PID controller is actively heating the boiler during the shot. As
I've mentioned, the TC is attached to the top of the boiler and during
a shot, the largest drop it registers is about 0.5 -1 C during the
initial high water flow period, then it stabilizes to about 0.1 C from
set point during the shot. From my study, mounting the TC close to the
group/boiler junction would allow the best active control during the
shot.
The flow rates that were used cover all useful rates for brewing. Even
though some shots were far longer than "recommended" they serve to
illustrate how the different thermal load rates affect the temperature.
The total integrated load is 60 ml of cold 20 C water (plus the initial
void filling/puck wetting volume) and results in essentially the same
final temperature in all graphs, just the rate of reaching this
temperature is varied. You may also notice that during the shot the
temperature drop curves downwards. This is a result of the flow rate
increasing towards the end of the shot and clearly shows that it is the
water that is driving the temperature down. The sharp drop in the
beginning is the result of the large flow rates (ie high thermal
loading) at the beginning of the shot.
By pre-infuse, I simply mean that the pump was activated for 3-4
seconds to fill the puck/group void with water and wet the puck, then
the system is left for a time to allow the PID controller to come back
to set point. Then the shot was poured.
I know that people will want to believe what they want. But I have this
machine and am interested in improving its design. The first step is to
study how it performs. It does make a brilliant cappuccino, and an
acceptable espresso. However, I had always noted that PID could remove
the bitterness, but not all the sourness. Recently I have been visiting
coffee shops known for their espresso and calibrating my taste buds :)
I noticed that their shots were neither bitter nor sour. I can't
achieve that.
If you take my results into consideration when thinking about the
boiler/group design of the Classic, you have to question the effect of
introducing cold water into the boiler/group junction. This should
actively cool both the boiler AND group, due to the high thermal
conductivity of the metal. I for one wouldn't think of designing a
thermally stable coffee machine that way.
As I said, I expect these results to be even more pronounced for a
"standard" Classic that isn't actively heated during the pour of the
shot.
If I were to do any follow up studys, I'd look at varying the TC
position on top of the puck, rather than have it near the side of the
basket. But for me, the data speaks for itself, and the machine is what
it is.

[Sparky]

My machine is only trying to maintain the PID set point. The PID
controller can be made to at least partially deal with the high thermal
load if the thermocouple is placed close to the cold water inlet. Of
course, if it is set up like that (with aggressive PID values) then it
is less stable and will take longer to achieve the set point at warm
up.

The block is a passive thermal reservoir that does indeed increase the
thermal inertia of the system. I agree that for active control it would
be best to minimise the thermal inertia of the system. However, the
total increase in boiler mass is probably on the order of 10%, so not
too bad (also not too effective either).

I may try disabling the PID control and switching to full heating
during the shot to see what can be achieved. I haven't given up on this
machine yet :)

[Moka Java]

Post by Sparky
My machine is only trying to maintain the PID set point. The PID
controller can be made to at least partially deal with the high thermal
load if the thermocouple is placed close to the cold water inlet. Of
course, if it is set up like that (with aggressive PID values) then it
is less stable and will take longer to achieve the set point at warm
up.
The block is a passive thermal reservoir that does indeed increase the
thermal inertia of the system. I agree that for active control it would
be best to minimise the thermal inertia of the system. However, the
total increase in boiler mass is probably on the order of 10%, so not
too bad (also not too effective either).
I may try disabling the PID control and switching to full heating
during the shot to see what can be achieved. I haven't given up on this
machine yet :)

Since aluminum is so much more conductive than water and the heating
element is on the outside of the boiler perhaps the only place to put
the TC is in the boiler to measure the water directly. Don't know how
you might do that with a Gaggia boiler but folks have put a thermowell
in brass Silvia boilers with some success. Putting the TC at the shower
screen will not help you, it's too much after the fact.

Richard "still don't know what I'm talking about" F

[tom b.]

Post by Sparky
I may try disabling the PID control and switching to full heating
during the shot

Has it been mentioned to turn the steam switch on during the shot?

Thanks for sharing your data.. >>:)

[Dave B]

Very intriguing! Info and research on something other than a Silvia!

excellent

[Danny]

-snip big test-

Whilst not wishing to argue with your results, I was very surprised,
since I have taken measurements from a Gaggia Baby I own and hadn't
seen anything like the fluctuations you describe.

So much so that I had to enter the dark pit that is my front room,
fight between the piano accordions and vintage radios, tossing aside
the Leak stereo 70 and other bits of tat in my search for one of the
two Gaggia Baby's I knew were in there somewhere. Between a piano and
a shelf load of vintage toasters and Sunbeam food mixers I found a
white Baby (does this make a difference, I thought to myself).

I created sufficient space near a wall outlet in the kitchen and
filled the water reservoir. Switched on and primed the pump. Nice
mix of grey cloudy water and some pure aluminium recovered from the
drip tray (just like panning for gold) and the machine was at temperature.

I dug out my fluke-yellow multimeter (alas, not a Fluke) and stuffed
the bead TC up one of the holes of the double spout whilst grinding
for a double. Loaded the pf and I was off. I waited for the temp
light to go out and then hit the brew switch.

During the whole pour the temp stayed between 89-87 degrees. I
discarded this one and loaded and locked again. Same result, but a
better pour, since the metal was nice and hot now, so 88-90.

Decided to risk the pf gasket and threaded the bead into the next
carefully crafted and tamped pf, resting it on the coffee. Locked it
in the machine and went for another pull. 93-91 degrees this time,
and the heater came on during the pour. I stopped looking at the
heater light. No water from the gasket seal where the TC was
squashed, so I pulled some more shots. My, is this machine unusually
stable. I pulled 10 shots, all within 2 degrees of each other. To be
absolutely fair, there was never at any time a difference of more than
4 degrees between shots, and the temperature never rose or fell by
this amount or greater.

I appreciate that my data is empirical, that I did nothing but write
some numbers down on an old Weetabix box, and that my TC and meter
aren't exactly calibrated for this type of work. Nevertheless, even
if the temps were out, the mean high/low would still be fairly OK, so
I'm happy that a Gaggia Baby (at least this one) is able to produce
consistent shots that don't vary by more than 2 degrees.

Anyone living near Portsmouth want a Gaggia Baby? Needs a steam arm
extension and the drip tray cover. Guaranteed consistent pours.
Probably could do with descaling...

[Sparky]

Thanks for going to the extra trouble to do a few measurements. If we
are to reconcile your findings with mine, then we'd have to compare
extperimental methods. For instance the TC placement can have a
dramatic effect on the observed temperature. Also, is your machine
designed the same way as mine? My machine was built in 2004, so is
probably the current model.

I'm planning to log other Classics and other machines, thus keeping the
experimental mathod the same and offering a direct cross-machine
comparison.

[Danny]

Post by Sparky
Thanks for going to the extra trouble to do a few measurements. If we
are to reconcile your findings with mine, then we'd have to compare
extperimental methods. For instance the TC placement can have a
dramatic effect on the observed temperature. Also, is your machine
designed the same way as mine? My machine was built in 2004, so is
probably the current model.
I'm planning to log other Classics and other machines, thus keeping the
experimental mathod the same and offering a direct cross-machine
comparison.

I don't know all the differences between a Classic and a Baby. I have
two Baby Gaggia's, one of '90's and one of '80's vintage. I used the
later model - white with a removeable water reservoir.

I placed the TC on top of the coffee cake, a couple of mm above the LH
lug as you look at the basket.

[Sparky]

Post by Danny
I don't know all the differences between a Classic and a Baby. I have
two Baby Gaggia's, one of '90's and one of '80's vintage. I used the
later model - white with a removeable water reservoir.
I placed the TC on top of the coffee cake, a couple of mm above the LH
lug as you look at the basket.
--
Regards,
Danny

Hmmmm, I positioned the TC on top of the puck but slightly below the RH
lug. In my machine the water inlet to the shower distribution block is
on the far LH side. So the path the water has to travel is greater in
my test, than in yours. Something to consider perhaps?

Does your machine introduce the cold feed water into the bottom of the
boiler/top of the group, like what is done in my machine? One more
question. Does your machine have exposed external elements, or is the
element embedded in the boiler wall? I've seen an old Gaggia boiler
like this.

[Sparky]

Very briefly, I have just logged two successive temperature profiles on
a Bezzera BZ99 (aka. Livia 90) HX machine using my equipment and same
experimental method (ie same probe placement in the basket and
operating conditions). No cooling flush was used and no barista
technique was used to get the best operating conditions. The brew
switch was just pressed and the data logged.

The results are here:
Loading Image...

For a comparison with know HX temperature profiles, have a look at the
home-barista article on HX machines here:
http://www.home-barista.com/hx-love.html

As you can see, I'm logging very similar results with my equipment.

Now just for the sake of comparison, have a look at how my Classic
compares:

Loading Image...

As this wasn't a true side-by-side comparison, I won't say any more
about it.

More cross-comparisons are on the way, thanks to some very helpful
people. I hope you'll stay tuned.

[jim schulman]

Post by Sparky
Very briefly, I have just logged two successive temperature profiles on
a Bezzera BZ99 (aka. Livia 90)

Nice, the first shot shows the influence of the cold nose early in the
shot, and the overheated HX water later.

But nobody is questioning your temperature logging technique. We're
trying to figure out what's wrong with your machine, since almost
everyone else is getting better results with theirs.

You can't say the Danish test parallels yours. They are measuring at
the outlet, and giving the maximum deviation over several shots. they
explicitly say the Classic is in the same range as the Giotto,
Zaffiro, and the other category winners. In your defence, they noted a
lot more variation on the Gaggia Coffee, which should have identical
characteristics. I'm guesing a tstat with a wider deadband.

You still haven't published a curve with removed heatsink and PID
converted to tight on/off control. Such a reading may iadd
information. For instance, your hypothesized cold water influx effect
could also be the heatsink sinking heat.

To be honest, what puzzles me, and perhaps others, is your not
disassembling the machine and tracking the fault; if not for our
enlightenment, then yours.

If I had results from any machine as miserable as the ones you got, my
first actions would be to try to fix it. If I got a result as
miserable from a machine with a 25 year sterling reputation for shot
quality; I'd do a very meticulous diagnosis of everything I could have
possibly done wrong. The actions required to fix the shot would
automatically locate the fault, whether inherent in the design, in the
PID execution, or screwed up assembly of your particular machine.

[Sparky]

Post by jim schulman
But nobody is questioning your temperature logging technique. We're
trying to figure out what's wrong with your machine, since almost
everyone else is getting better results with theirs.

It's just to remove any doubt (in my mind at least) that the logging
was not carried out correctly.

Post by jim schulman
You can't say the Danish test parallels yours. They are measuring at
the outlet, and giving the maximum deviation over several shots. they
explicitly say the Classic is in the same range as the Giotto,
Zaffiro, and the other category winners. In your defence, they noted a
lot more variation on the Gaggia Coffee, which should have identical
characteristics. I'm guesing a tstat with a wider deadband.

I haven't translated the Norwegian result, so it is hard to know
exactly what they found, how they conducted the tests and what the
numbers really mean.

Post by jim schulman
You still haven't published a curve with removed heatsink and PID
converted to tight on/off control. Such a reading may iadd
information. For instance, your hypothesized cold water influx effect
could also be the heatsink sinking heat.

We could discuss this, but the best way to resolve this is to log a
standard Classic. This is in the pipe.

Post by jim schulman
To be honest, what puzzles me, and perhaps others, is your not
disassembling the machine and tracking the fault; if not for our
enlightenment, then yours.

Wife and child :)

Post by jim schulman
If I had results from any machine as miserable as the ones you got, my
first actions would be to try to fix it. If I got a result as
miserable from a machine with a 25 year sterling reputation for shot
quality; I'd do a very meticulous diagnosis of everything I could have
possibly done wrong. The actions required to fix the shot would
automatically locate the fault, whether inherent in the design, in the
PID execution, or screwed up assembly of your particular machine.
--
jim schulman

You're assuming the coffee my machine makes is bad. Before PID and
"heat sink" the shots were often bitter. It was hard to control the
quality with time sufing. After, I could easily remove the bitterness
with PID adjustments.

I also note that older Gaggia's had very different boilers, with almost
3 times the mass of metal than the current design (I had the
opportunity to juggle both of these boilers today). These older boilers
had their heating element embedded into the boiler body and so is not
visible. The internal boiler volume is cylindrical in these older
units, incontrast to the newer shape which defies description (by me at
least, maybe wedge shaped). I was told that these boilers were about 5
years old.

[D. Ross]

| If I had results from any machine as miserable as the ones you got, my
| first actions would be to try to fix it. If I got a result as
| miserable from a machine with a 25 year sterling reputation for shot
| quality;

Well, that reputation is based on the taste of the shot, not the shape of
the temperature curve; all this home espresso metrology is relatively new,
and historically not applied often to Gaggias.

- David R.
--
Less information than you ever thought possible:
http://www.demitasse.net

[k***@hotmail.com]

The very old Baby's had the cold water inlet through the side of the
boiler rather than through the group itself. This suggests that the new
Gaggias could be improved by boring out the hole and putting a length
of teflon hose in between the cold water junction and the outlet in the
boiler (with an expander at the inlet end to keep it from being forced
into the boiler). In fact the hose could protrude into the boiler a bit
to direct the inflow slightly away from the group head which should
further aid stability. Preheating the inlet water would also help and
improve recovery time.

One thing that makes me a bit skeptical of your Classic is that my
Gaggias produce richer, fuller, sweeter shots using a double basket for
a single shot - very, very close to a commercial HX shot. By your
arguements of thermal loading of the basket, the double basket shot
should be even cooler and hence thinner and more sour, shouldn't it?

Greg

[Sparky]

Post by k***@hotmail.com
One thing that makes me a bit skeptical of your Classic is that my
Gaggias produce richer, fuller, sweeter shots using a double basket for
a single shot - very, very close to a commercial HX shot. By your
arguements of thermal loading of the basket, the double basket shot
should be even cooler and hence thinner and more sour, shouldn't it?
Greg

I'm going to log some shots with a single basket, both in the basket
and in the spout (to replicate the conditions of the Norwegian study).
I think single baskets have their own problems. They're harder to
prepare without microchanneling and, I for one, have always gotten far
less crema from the singles. Otherwise,as you suggest, the thermal
loading should be less. We'll see.

I've been getting some excellent shots out of my machine lately.
Joining the science with art can be fraught with difficulty. Schomer
assures us that temperature stability is the key. Alan Frew has found
Silvia (which is believed to be thermally stable) shots to be slightly
better than Classic shots. HX people suggest that variaible temperature
profiles might be better to adapt to particular blends... and so it
goes. Still the correlation between temperature profile and shot taste
is vague at best. Temperature profiles might be great for comparing how
machines operate and their thermal design, but what is really important
for taste?

As for how a Gaggia Classic operates, I've modelled the heat flow in a
boiler-on-group system typified by the Gaggias and have found that the
cold feed water dramatically cools the group, with significant cooling
in the first second and dramatic cooling after 10 seconds. It's a
simple 1-D model, but it should apply pretty well to the Gaggia case
(and even neglects extra cooling of cold water flow through the
group-boiler junction). What it shows is that the group provides most
of the heat to warm the cold feed water. I'll leave the details out for
now until I've investigated a few more Gaggia machines.

Nevertheless, for those pursuing the thermal stability mantra, or those
who eschew it, these investigations are worth pursuing. The
investigation continues.

[Sparky]

Okay, here's some more results to ponder.

I just ran some more data logging on my Classic, but this time with
single filter baskets and with the thermocouple both in the filter
basket and in the spout, monitoring the temperature of the liquid
exiting the porta-filter. I also ran some in-the-spout shots with a
double basket to see how that fared. As I understand it, the Norwegian
study logged temperature of the liquid exiting the portafilter using 7
gm of coffee in a single basket and 30 ml pours, so I wanted to see
what they saw.

Here's the in the spout results for the single basket:

Loading Image...

and here's the same experiment done with a double basket and 60 ml
pours:

Loading Image...

What's going on? The data looks very temperature stable when you log it
like this. Both single and double pours show variations in temperature
of around 2 C, some much less. The profiles ramp up in the beginning as
the hot water has to heat up the puck, but then it does seen to plateau
and remain quite stable.

No have a look at the single basket when you monitor the temperature in
the basket:

Loading Image...

Now that looks familiar. Well to me at least. I see the same initial
"Gaggia spike" that I've observed with the doubles and the region of
the pour has a falling temperature profile. However, the total fall in
temperature is not too large in this case. If you disregard the initial
spike, the temperature only falls about 2-3 deg C during the pour.
That's not too bad, especially when compared to the results found for
the 60 ml double pours.

Finally in an effort to further understand what is going on here, I
decided to set up a model of the Gaggia boiler/group area and see what
happens when you put cold 30 deg C water directly on to the top of the
group/bottom of the boiler. I have made a plot of the temperature fall
of a point in the group 10 mm below the boiler floor. You can see it
here:

Loading Image...

It clearly shows the effect of the large initial thermal load on the
group temperature. The key point to note is how fast the temperature
falls. By one second the temperature has dropped nearly 20 degrees. Of
course this is only a model, but it is good enough to give some insight
into what cold water injected into the group is capable of.

Thus ends this installment.

[Aaron Cooper]

Post by Sparky
Okay, here's some more results to ponder.
I just ran some more data logging on my Classic, but this time with
single filter baskets and with the thermocouple both in the filter
basket and in the spout, monitoring the temperature of the liquid
exiting the porta-filter. I also ran some in-the-spout shots with a
double basket to see how that fared. As I understand it, the Norwegian
study logged temperature of the liquid exiting the portafilter using 7
gm of coffee in a single basket and 30 ml pours, so I wanted to see
what they saw.

http://i4.photobucket.com/albums/y129/mjfernee/Classic/Gag_single_in_spout.jpg

Post by Sparky
and here's the same experiment done with a double basket and 60 ml

http://i4.photobucket.com/albums/y129/mjfernee/Classic/Gag_double_in_spout.jpg

Post by Sparky
What's going on? The data looks very temperature stable when you log it
like this. Both single and double pours show variations in temperature
of around 2 C, some much less. The profiles ramp up in the beginning as
the hot water has to heat up the puck, but then it does seen to plateau
and remain quite stable.
No have a look at the single basket when you monitor the temperature in

http://i4.photobucket.com/albums/y129/mjfernee/Classic/Gag_single_in_basket.jp

Post by Sparky
g
Now that looks familiar. Well to me at least. I see the same initial
"Gaggia spike" that I've observed with the doubles and the region of
the pour has a falling temperature profile. However, the total fall in
temperature is not too large in this case. If you disregard the initial
spike, the temperature only falls about 2-3 deg C during the pour.
That's not too bad, especially when compared to the results found for
the 60 ml double pours.
Finally in an effort to further understand what is going on here, I
decided to set up a model of the Gaggia boiler/group area and see what
happens when you put cold 30 deg C water directly on to the top of the
group/bottom of the boiler. I have made a plot of the temperature fall
of a point in the group 10 mm below the boiler floor. You can see it
http://i4.photobucket.com/albums/y129/mjfernee/Classic/group_temp_fall.jpg
It clearly shows the effect of the large initial thermal load on the
group temperature. The key point to note is how fast the temperature
falls. By one second the temperature has dropped nearly 20 degrees. Of
course this is only a model, but it is good enough to give some insight
into what cold water injected into the group is capable of.
Thus ends this installment.

This is very interesting stuff. I recently took the boiler of my Gaggia
Coffee apart and was pondering at the design. I don't really understand how
you made this model-- could you explain a bit further?

I've been enjoying reading your followups and I now regret being so skeptical
at first. If the Norwegian study was performed as you have said, it's not
difficult to understand how they determined the machine to be 'stabile.'

Aaron

[Sparky]

Post by Aaron Cooper
This is very interesting stuff. I recently took the boiler of my Gaggia
Coffee apart and was pondering at the design. I don't really understand how
you made this model-- could you explain a bit further?
I've been enjoying reading your followups and I now regret being so skeptical
at first. If the Norwegian study was performed as you have said, it's not
difficult to understand how they determined the machine to be 'stabile.'
Aaron

Thanks Aaron. Skepticism is healthy and has spurred me to really try
and get to the bottom of this. The only information I have regarding
the methods used in the Norwegian study was that posted by Jim
Schulman. So I ran some shots accordingly. It's interesting that even
double pours appear to be temperature stable under those conditions.

The model is only pieced together from snippets of design information.
As I haven't actually pulled my machine apart (as it's still under
warranty, I can only go by parts diagrams, external inspection of where
the pipes run in my machine, some pictures posted on Alan Frews site
and a replacement Gaggia boiler that I've seen. The crucial component
is the group. The feed water is introduced into the boiler through the
top of the group. As the water is cold and the boiler water is hot, it
should stratify rather than mix (if it did mix, the small boiler volume
would make it terribly unstable). In my model I use a 20 mm thick brass
plate (as the group is made of brass). I'm assuming red brass or marine
brass, but yellow brass will not change the timings more than a
fraction of a second. I place a 10 mm layer of 30 deg C water directly
on top of the 93 deg C brass plate. A 60 mm deep layer of brew
temperature water sits on top of the cold water layer and a 10 mm thick
layer of brew temparature water sits below the group plate. I then use
the heat equation to determine the heat flow in this 1-D model. I use
actual engineering values for the diffusivity constants for both water
and marine brass to get the simulation running in real time.

Last night I ran a full 3-D model, well a cylindrically symmetric 2-D
model with heated walls (more like my PID situation) to test for edge
effects. I found edge effects only extended about 10 mm from the edge,
then the 1-D model was satisfied. This all assumes that there is no
other insulating layer in the group/boiler flange, just straight
conduction through metal until the water distribution block. I'm using
approximately a 30 ml injection, which is about how much water is
rapidly injected into the system when you hit the brew switch with a
double basket(due to wetting of the puck and filling the puck-shower
head void). The model is all about heat conduction and how fast it
occurs following cold water injection into the bottom of the boiler.

I have a really interesting test in the pipe: Stock Gaggia Classic
compared to a stock Silvia compared to a commercial machine. I'm pretty
excited about that one.

[Sparky]

Here's the 3D heat transfer model. It's a cross-section of the modelled
boiler-group arrangement showing the stratified brew water, cold feed
water, brass group layer and a hot water layer below the group. It's a
more visual representation of what's going on, showing the actual
boiler dimensions assumed in the model.

Loading Image...

A colour bar is included so you can read the temperatures (in
centigrade). The heat transfer was calculated 3 seconds after the feed
water is injected. Edge heating effects are visible but do not prevent
rapid cooling of the group by the cold feed water layer. Note, all the
grey edges are maintained at a constant temperature of 95 C, which
simulates constant heating (more appropriate for a PID controlled
machine). The non-heated boiler would fare even worse, but not
significantly so with respect to the group cooling.

[Aaron Cooper]

Post by Sparky
Here's the 3D heat transfer model. It's a cross-section of the modelled
boiler-group arrangement showing the stratified brew water, cold feed
water, brass group layer and a hot water layer below the group. It's a
more visual representation of what's going on, showing the actual
boiler dimensions assumed in the model.

http://i4.photobucket.com/albums/y129/mjfernee/Classic/Heat_transfer_3s_2D_gag

Post by Sparky
gia.jpg
A colour bar is included so you can read the temperatures (in
centigrade). The heat transfer was calculated 3 seconds after the feed
water is injected. Edge heating effects are visible but do not prevent
rapid cooling of the group by the cold feed water layer. Note, all the
grey edges are maintained at a constant temperature of 95 C, which
simulates constant heating (more appropriate for a PID controlled
machine). The non-heated boiler would fare even worse, but not
significantly so with respect to the group cooling.

What does this model look like after 10-25 seconds? Isn't that the real
concern? Also, I think you should up the temperature of the cold water
injection to 50 C or so, as the water in the reservoir of my machine gets
toasty pretty fast. Not sure about yours, though.

Since your machine is still under warranty, feel free to email me if you need
detailed pictures or measurements of the boiler innards. I'd be happy to take
it apart again, and I do believe that our machines have the same boiler
design.

Aaron
kehanren /at/ gmail /dot/ com

[Sparky]

Post by Aaron Cooper
What does this model look like after 10-25 seconds? Isn't that the real
concern? Also, I think you should up the temperature of the cold water
injection to 50 C or so, as the water in the reservoir of my machine gets
toasty pretty fast. Not sure about yours, though.

The model reaches a constant thermal gradient throughout the group in
those 3 seconds. That's nearly a steady-state condition for that model,
so going further in time would yield pretty much the same result
(actually the interfaces aren't at steady-state and thats where the
heat transfer takes place). I used 3 seconds in order to show how fast
the system responds to the rapid cold water injection in the beginning
of a shot. It's pretty hard to simulate the entire brewing process as
that involves fluid flow, which is pretty complicated. What the model
tells me is that putting cold water into the bottom of the boiler has
an immediate effect on the group temperature, and that most of the heat
used to heat the cold water in-flow in fact comes from the group (at
least initially).

The ambient temperature here is about 20 C at the moment. The water in
the reservour is pretty much isolated from the hot boiler compartment,
so unless I left the machine on all day, it would have little effect on
the reservoir water. I've used 30 C water, which is already 10 C hotter
than the reservoir water.

As for the innards. I know the boiler, but I am unsure if the bottom of
the boiler/top of the group (where the water distribution block bolts
on) is one lump of solid metal, or if there is some intervening space
that breaks the connection.

I'm planning another temperature log of my machine as soon as I have
the time. This time I'll introduce the TC bead into the left hand side
of the basket, as close to where the water exits the group as possible,
rather than diametrically opposite. There may be a large thermal
gradient across the puck dependent on the path the water has travelled
across the bottom of the group.

[hudson]

Post by Sparky

Post by Aaron Cooper
What does this model look like after 10-25 seconds? Isn't that the real
concern? Also, I think you should up the temperature of the cold water
injection to 50 C or so, as the water in the reservoir of my machine gets
toasty pretty fast. Not sure about yours, though.

The model reaches a constant thermal gradient throughout the group in
those 3 seconds. That's nearly a steady-state condition for that model,
so going further in time would yield pretty much the same result
(actually the interfaces aren't at steady-state and thats where the
heat transfer takes place). I used 3 seconds in order to show how fast
the system responds to the rapid cold water injection in the beginning
of a shot. It's pretty hard to simulate the entire brewing process as
that involves fluid flow, which is pretty complicated. What the model
tells me is that putting cold water into the bottom of the boiler has
an immediate effect on the group temperature, and that most of the heat
used to heat the cold water in-flow in fact comes from the group (at
least initially).
The ambient temperature here is about 20 C at the moment. The water in
the reservour is pretty much isolated from the hot boiler compartment,
so unless I left the machine on all day, it would have little effect on
the reservoir water. I've used 30 C water, which is already 10 C hotter
than the reservoir water.
As for the innards. I know the boiler, but I am unsure if the bottom of
the boiler/top of the group (where the water distribution block bolts
on) is one lump of solid metal, or if there is some intervening space
that breaks the connection.
I'm planning another temperature log of my machine as soon as I have
the time. This time I'll introduce the TC bead into the left hand side
of the basket, as close to where the water exits the group as possible,
rather than diametrically opposite. There may be a large thermal
gradient across the puck dependent on the path the water has travelled
across the bottom of the group.

I am sorry to say this again, and I do appreciate the effort that has gone
into the science, but do not really understand the point of most of it, the
Gaggia Classic is a very simple machine. It consists of a very simple boiler
and a very simple steam system. It is an old but good design. It has
probably not changed much because it works. What is most important is what
ends up in the cup.

Best Regards

Stuart Hudson

[Aaron Cooper]

Post by hudson

Post by Sparky

Post by Aaron Cooper
What does this model look like after 10-25 seconds? Isn't that the real
concern? Also, I think you should up the temperature of the cold water
injection to 50 C or so, as the water in the reservoir of my machine gets
toasty pretty fast. Not sure about yours, though.

The model reaches a constant thermal gradient throughout the group in
those 3 seconds. That's nearly a steady-state condition for that model,
so going further in time would yield pretty much the same result
(actually the interfaces aren't at steady-state and thats where the
heat transfer takes place). I used 3 seconds in order to show how fast
the system responds to the rapid cold water injection in the beginning
of a shot. It's pretty hard to simulate the entire brewing process as
that involves fluid flow, which is pretty complicated. What the model
tells me is that putting cold water into the bottom of the boiler has
an immediate effect on the group temperature, and that most of the heat
used to heat the cold water in-flow in fact comes from the group (at
least initially).
The ambient temperature here is about 20 C at the moment. The water in
the reservour is pretty much isolated from the hot boiler compartment,
so unless I left the machine on all day, it would have little effect on
the reservoir water. I've used 30 C water, which is already 10 C hotter
than the reservoir water.
As for the innards. I know the boiler, but I am unsure if the bottom of
the boiler/top of the group (where the water distribution block bolts
on) is one lump of solid metal, or if there is some intervening space
that breaks the connection.
I'm planning another temperature log of my machine as soon as I have
the time. This time I'll introduce the TC bead into the left hand side
of the basket, as close to where the water exits the group as possible,
rather than diametrically opposite. There may be a large thermal
gradient across the puck dependent on the path the water has travelled
across the bottom of the group.

I am sorry to say this again, and I do appreciate the effort that has gone
into the science, but do not really understand the point of most of it, the
Gaggia Classic is a very simple machine. It consists of a very simple boiler
and a very simple steam system. It is an old but good design. It has
probably not changed much because it works. What is most important is what
ends up in the cup.
Best Regards
Stuart Hudson

It sounds as if you are not interested in any temperature study of the Gaggia
Classic. As this is the title of this thread, I recommend that you refrain
from reading it rather than making such counterproductive comments and
thereby clouding it up.

Aaron

[Sparky]

Here's an update on my progress. I logged 4 shots last night with the
TC in the just above the right hand lug, just to test for possible
temperature variations across the puck. In general the results looked
pretty similar to what I'd already obtained. So there doesn't seem to
be a significant temperature gradient across the puck.

To test if the feed water was driving the temperature down, I used hot
tap water in the reservoir (55 C). Compared to the 20 C reservoir water
I was using, it represented an approximately 50% reduction in
temperature difference. The temperature profile accurately reflected
this with an approximate 3 C drop during a 60 ml shot.

Now for the taste comparison. I decided not to use beans that I had
been drinking, not garbage beans. Therefore I did a taste comparison.
Sure enough, the cold feed water shots had a sour note. In contrast,
using the 55 C feed water, the shot temperature went from 91 - 88 C
(not down to 84-85 C) and the taste was now devoid of the sourness and
taste was much more complex with flavours becoming much more
pronounced. A very surprising result.

BTW: the pump didn't like working at 55 C at all. The pumping sound
changed completely, so I wouldn't recomment running with heated
reservoir water.

Now the one question that remains in my mind is: Is my machine faulty?