Temperature Stability . . . . from a Heat Exchanger

Need help with equipment usage or want to share your latest discovery?
Ken Fox
Posts: 2447
Joined: 18 years ago

#1: Post by Ken Fox »

I am posting this both here and on alt.coffee because there are people I respect who participate in one venue or the other and I'd like to get opinions from all of them.

Having wasted about three whole days datalogging shots with the Scace Thermofilter on my Cimbali Juniors, I had nothing to show for it but a bunch of horrible looking graphs that were beginning to convince me that one could never get temperature stability out of a HEX machine and that the PID Jim Schulman and I put in (with some assistance from Barry Jarrett) a couple of years ago had accomplished nothing. I was getting ready to call my banker to see about a loan on a new LM GS3:-)

I've had a number of emails going back and forth with such local luminaries as Jim Schulman, Barry J., even Andy S., and I was getting ready to jump off the roof or hang myself in the closet. Jim kept saying I didn't have the flush down low enough, I should try 110ml, Barry said, hey, temperature stability, HEX, no can do, which was more or less what I inferred from my emails with Andy as well. I seemed to have my choice of good first shots but overheated seconds and thirds and so forth, or hot first shots followed by good later shots.

And then----- I got LUCKY. I started to do some datalogging while playing around with the boiler temperature and the flush volumes. And here is what I found: by reducing the boiler temperature to well below anything I've used before, 231F, which equates to about 0.6 bar on my front panel gauge, and markedly reducing the cooling flush down to 45ml (an ounce and a half, about a third of what I've used before), I struck PAYDIRT.

It would be nice to compare these results to what I can get on the same machine under pstat control, down at the same temperature range and with the same flush; I don't have time to do that now since I'm headed off on a trip soon, but I'll try to get to it in January. I don't think that pstat control can match these results, as the gentleness of the curves and the consistency of them would not easily be obtained with the ~6 degree hysteresis present in my boiler with the pstat calling the shots.

I think these results, which I have now duplicated in 3 separate series this afternoon, show that this works, and validates the approach of reducing HEX boiler temperature, reducing flush size, and replacing the Pressurestat with electronic temperature control. If anyone else out there has a PID'd HEX machine, it sure would be nice to see this duplicated on another type of machine.

Just a note about methodology before I post the links to the graphs: The idea was to start off from an idle period, which is the hardest for any sort of espresso machine as this tends to produce either overheated or cool groups depending on the machine. I pulled one thermofilter shot about 15-30 seconds after the small cooling flush, then waited 1min 30 seconds to pull the next one, and so forth. Although on my earlier testing I'd removed the Scace PF in between shots, I've been convinced by my own testing that this accomplishes nothing because the Thermofilter just measures the water as it flows through and any impact of the PF temperature (it is a chopped PF and the probe is far away from metal or other heat sinks) is eliminated after 1 second into the shot. Therefore, after the small cooling flush I returned the Scace device to the group and left it there throughout the 10 or 12 minutes it took to do each series. One other reason for doing this is that the basket they are using on the Thermofilter doesn't seal well with Cimbali group gaskets and once I have a good seal in there I don't want to risk a poor seal on later shots which would make it hard to trust the measured results.

Have a lookie at these graphs of shots, pulled as above, on a lowly Heat Exchanger machine:

Image

Image

Image





ken

User avatar
another_jim
Team HB
Posts: 13871
Joined: 19 years ago

#2: Post by another_jim »

(cross-posted on alt.coffee)

I had a chance to look at these before and was stunned and mute. With
a bit more thought, I have some thoughts, although they may be a bit
incoherent:

-- I'm guessing you've gotten close to converting your machine to a
single boiler. With the Cimbali one piece HX and group, and the group
acting as a heat sink; I'm guessing the HX stabilizes around 210F -
220F, and that with the cold inflow, you're basically getting shot
temperatures inside the HX all the time.

-- One question will be how well the steam works at this boiler
pressure. Home lever machines, like the Elektra, Pavoni, and
Micro-Cimbali operate at around this pressure. These are all very good
steamers, better than low end home HX machines. So, with the proper
steam tip, you may get good performance, albeit more slow.

-- The other question is what happens if you start nudging the boiler
pressure up and the flush amount along with it (adjusted to keep the
average temperatures of all the shots in each series roughly equal). Does the
stability get more ragged slowly? Or will it stay at this stability
level for a while, then go back to normal HX behavior.

-- In Italy, most bars keep steam pressures low, usually around 0.8
bar, and have HXs suited to that temperature. Dual boiler machines are
not even a blip there. Perhaps you are getting closer to the way these
machines were designed to perform.

-- Finally, there's always been a "Nah, no way" reaction to running
the steam boiler of an HX with a TC placed inside the HX. Your data
shows it may not be so farfetched after all.

Advertisement
lennoncs
Posts: 234
Joined: 19 years ago

#3: Post by lennoncs »

Congratulations Ken!

You have found that "DPlot Religion"


nice looking graphs, I am going to ruminate on them a bit before I comment but it looks promising.

Cheers,
Sean

Ken Fox (original poster)
Posts: 2447
Joined: 18 years ago

#4: Post by Ken Fox (original poster) »

another_jim wrote:(cross-posted on alt.coffee)

I had a chance to look at these before and was stunned and mute. With
a bit more thought, I have some thoughts, although they may be a bit
incoherent:

-- I'm guessing you've gotten close to converting your machine to a
single boiler. With the Cimbali one piece HX and group, and the group
acting as a heat sink; I'm guessing the HX stabilizes around 210F -
220F, and that with the cold inflow, you're basically getting shot
temperatures inside the HX all the time.

-- One question will be how well the steam works at this boiler
pressure. Home lever machines, like the Elektra, Pavoni, and
Micro-Cimbali operate at around this pressure. These are all very good
steamers, better than low end home HX machines. So, with the proper
steam tip, you may get good performance, albeit more slow.

-- The other question is what happens if you start nudging the boiler
pressure up and the flush amount along with it (adjusted to keep the
average temperatures of all the shots in each series roughly equal). Does the
stability get more ragged slowly? Or will it stay at this stability
level for a while, then go back to normal HX behavior.

-- In Italy, most bars keep steam pressures low, usually around 0.8
bar, and have HXs suited to that temperature. Dual boiler machines are
not even a blip there. Perhaps you are getting closer to the way these
machines were designed to perform.

-- Finally, there's always been a "Nah, no way" reaction to running
the steam boiler of an HX with a TC placed inside the HX. Your data
shows it may not be so farfetched after all.
Hi Jim,

I'm eating dinner and in between courses running Scace/Fluke runs. I pushed the PID down to a boiler temp of 229 and ran two more sets, with (I think) EIGHT shots (or was it nine?) in the latter of the two sets, filling up my fluke. I just turned the temp down two degrees more, to 227; I think that varying the flush amount will just make this too complicated. It seems as if reducing the boiler temperature does not reduce the brew temp, degree for degree, but less than that. As to raising the temp, I'm already in the 202's with the graphs I already posted. Does anyone advocate brewing much above 203.5? I could go higher but I think it is pretty obvious that I could go up a degree without a problem and beyond that may be a second round of bean roasting rather than providing useful data:-) If this setup can do, say, 198-203 or thereabouts, I think that is most of the useful range.

As to steaming performance, you can hide all sorts of sins in milk. I think the response time is quick and one could just bump it up 2 or 3 or 5 degrees if one wanted to make cappas, leaving it at a lower level the rest of the time.

I'd give up a tiny amount of temperature stability not to have to fool around with 55ml vs. 40ml flushes and just standardizing on 45 for straight shots. Plus, it is pretty hard to measure 5 ml differences so I have programmed in the flush on one of the programmable buttons and just leaving it there.

Thanks for your input!

ken

Ken Fox (original poster)
Posts: 2447
Joined: 18 years ago

#5: Post by Ken Fox (original poster) »

lennoncs wrote:Congratulations Ken!

You have found that "DPlot Religion"


nice looking graphs, I am going to ruminate on them a bit before I comment but it looks promising.

Cheers,
Sean
Thanks for your help, Sean! I think I"m onto something with this, and since I seldom am, it is a bit of a rush. If it wasn't so repeatable I'd have my doubts but when you see the same number for 15 seconds straight on the datalogger, it is hard to ignore.

Take care, I'll have more to post later.

ken

Ken Fox (original poster)
Posts: 2447
Joined: 18 years ago

#6: Post by Ken Fox (original poster) »

Some more shot temperature graphs at different boiler temperatures:

At 227F, the shot reproducibility is only fair and there is quite a spread between shots although not following any real trend:

Image

and

Image

228 was not (yet) tested and might well get the range of reproducible shot temps down to 200F

229F has nice reproducible shot temperature characteristics:

Image

and

Image

231F was shown in the original post at top.

232F has nice tight reproducible curves. I only did one series, but with 10 consecutive shots. One of the later shots lost a couple seconds of initial data when I forgot to push the datalogger button on time:-) this does not effect the important part of that curve so I included it in the plot:

Image

It is obvious from viewing all the graphs I've posted that this technique, e.g. a low boiler temperature in a PID'd HEX boiler, with a very small initial flush, works well over a certain temperature range and less well when you get below it. This range appears to be from about 200F to 204F intended shot temperature. It would not surprise me if the new double boilers also have a certain range in which they can maintain stable and consistent shot temperatures and fail to maintain them at other intended shot temperatures. Has anyone subjected their dual boiler machine to this kind of testing?

ken

edna713
Posts: 20
Joined: 18 years ago

#7: Post by edna713 »

After slogging through the lengthy prose from Fox, who was testing shot temps between dinner courses, and is NOW calling his Cimbali jr. a lowly HX machine, I have a question:

WHY IS IT SO HARD TO GET CONSISTENT TEMPS FROM HX DESIGN MACHINES?? why all the fiddling and folderol?

Is an (expensive) HX espresso machine REALLY an upgrade??

Dave

Advertisement
Ken Fox (original poster)
Posts: 2447
Joined: 18 years ago

#8: Post by Ken Fox (original poster) »

I've managed to waste a bunch more time datalogging my PID'd Cimbali D1, and found a range where I can get very tight temperature distributions (aka "flat shot profiles") with the current setup of low boiler temperatures and a small initial flush. Some of these temperatures (especially at the far ends) don't produce as tight a temperature distribution or reproducibility from shot to shot. Even in these cases, however, it appears that the first couple of shots are pretty consistent, just the reproducibility suffers after two or three shots. This could be a function of PID programming, flush volume (which I've kept constant for ease of use), or other factors I've not considered. This is not a perfect system by any means and although it offers a good deal of control in the desired shot temperature range of 200 to 204, it falters above and below that and doesn't offer intra-degree control. It may well be that even the new dual boiler machines have a range where they produce consistent temperature shots and a range where they are less reliable. This may be a design limitation in 2005 of trying to accomplish this degree of temperature control. It would be great if some users of GS3s, Synessos, LMs, Brewtus's, Technos, and others, who have access to a Scace Thermofilter and Datalogger, if they could run series like these and post their results. I must warn you, however, that this sort of stuff eats a LOT of time!

The data collection (shot) frequency was set at one shot per 2 minutes to approximate the work rate of a reasonably skilled home barista who takes the time to dose and distribute and tamp well enough to avoid channeling and other common pitfalls we know know we all commit in the era of the bottomless PF.

Here are the curves in ascending order from boiler temperatures of 227F to 233F:

227F:

Image

Image

228F:

Image

229F:

Image

Image

230F:

Image

231F

Image

Image

Image

232F:

Image

233F:

Image

At their best, I think the demonstrated stability is probably about as good as the results I've seen posted from any other machine, but again, as set up this works best over only a certain range of shot temperatures. I made a cappuccino this morning at a boiler temp of 231F, and although the frothing was slow, copius microfoam was produced. I'd say the froth quality was better than I have been getting at higher boiler temps but it does obviously take a few seconds longer to get the job done. If I was making mostly milk drinks, where shot quality is less important, I'd bump the temperature up a few degrees and maybe go back to a larger flush. For straight shot making, however, low boiler temps and a smallish flush seem to produce the best results, at least on paper.

ken

User avatar
AndyS
Posts: 1053
Joined: 19 years ago

#9: Post by AndyS »

edna713 wrote:After slogging through the lengthy prose from Fox, who was testing shot temps between dinner courses, and is NOW calling his Cimbali jr. a lowly HX machine, I have a question:

WHY IS IT SO HARD TO GET CONSISTENT TEMPS FROM HX DESIGN MACHINES?? why all the fiddling and folderol?
It's hard to get consistent temps from ANY espresso machine. Even a DAVE-PIDed Silvia!

If you get yourself a Scace device and do some testing, you will find this out for yourself.
-AndyS
VST refractometer/filter basket beta tester, no financial interest in the company

Ken Fox (original poster)
Posts: 2447
Joined: 18 years ago

#10: Post by Ken Fox (original poster) replying to AndyS »

Hi Andy,

You are among the most rigorous and conscientious "espresso-scientists" out there, and I really respect your judgement and your observations.

Could you tell us what if anything you have found on the idea of there being a "sweet spot," a certain band of temps that a machine might be able to pull flat curve shots at, and that the ability to do this declines as one gets to the fringes of this "band?" It seems only logical to me that you can only tune such a system (as an espresso machine) so well, and in the design the engineers would have to decide where they wanted the machine to be the most stable.

Any observations you have on this issue would be very interesting, at least to me.

ken

Post Reply