Errors in temperature and pressure measurements

[John]

Anytime that anything is measured, there are errors, to some degree, in that measurement.

Now that some people have become very interested in various measurements regarding espresso extraction, I thought it was time to present what these measurement errors can amount to. This will enable experimenters to understand why their numbers don't match.

Let's look at temperature first, as it is the most popular. [For reference a degree F = 1.8x a degree C.]

The first measurement limitation is the measuring instrument. Below are three typical cases. In all cases, unless otherwise noted, I am providing the total possible error band - not plus or minus.

Keithley 2700 Multimeter and Data Acquisition System -
the most accurate device for reasonable money ~ $1500
accuracy with type E or T thermocouple = .36 degree F

CN77000 PID controller ~ $230
typical of high accuracy controllers
accuracy with thermocouple .7 degree F

DVM typical digital volt meter less than $200
accuracy with thermocouples stated as ± 1% which equals 3.96 degrees F @ 198 F

Next to consider are thermocouples themselves.

Conventional thermocouples
Standard limits E = 3˚F T = 1.8˚F J or K = 4˚F

Better thermocouples use "Special limits of error" wire
Accuracy E type 1.8 degree F
Accuracy T type .9 degree F

The combination of the instrument and thermocouple errors are as follows.

Keithley 2700 with E type (special limits) = within 2.16 degrees F
with T type (special limits) = within 1.26 degrees F

High quality PID controller
with E type (special limits) = within 2.5 degrees F
with T type (special limits) = within 1.6 degrees F

DVM typical with E type standard = within 6.96 degrees F
with T type standard = within 5.76 degrees F
with J or K type standard = within 7.96 degrees F


I particularly mention the DVM with J or K type thermocouples because these are typical inexpensive systems such as most people will start out with.


Calibration
It is possible to get thermocouples by themselves, and even together with the measuring instrument, calibrated. The limits of this calibration are typically .7 degree F total.



Now we come to pressure measuring. This is yet to be a big issue but when people finally get the chance to adjust the extraction pressure in fine degrees it will become a big deal.

A typical gauge that might be fitted to a high quality espresso machine has an accuracy of ± 2%. At 9.2 bar this equates to an error of about ± .2 bar, or a range of .4 bar.

We know routinely that we can taste the difference of .05 bar.

In a machine such as the M3, which uses a calibrated pressure transducer (a thing that turns pressure into electricity) and an analog to digital converter to tell the computer what the pressure is in digital bits, there are errors here as well. There doesn't seem much reason to go into all those errors yet.

[malachi]

As noted previously, a good argument for very large data sets.

[lennoncs]

Poor measurement practices are another whole can of worms.

sean

[DavidMLewis]

Hi John,

Nice to see you posting here. I had one question about your notes, which by the way, I greatly appreciate. The question is: what is the repeatability of the various methods, as opposed to the absolute accuracy? While the latter is clearly important if you're developing a machine for production, if you're an individual user closing the loop with your taste buds, which covers most of us, it seems like the former is more crucial.

Best,
David

[John]

Hi John,

Nice to see you posting here. I had one question about your notes, which by the way, I greatly appreciate. The question is: what is the repeatability of the various methods, as opposed to the absolute accuracy? While the latter is clearly important if you're developing a machine for production, if you're an individual user closing the loop with your taste buds, which covers most of us, it seems like the former is more crucial.

Best,
David

Usually repeatability is not worth worrying about. There are many tiny errors that are below what I have sketched out here, but we have yet to find them important. On the other hand, if you are using an inexpensive DVM it would be worth contacting the manufacturer about repeatability - particularly temperature coefficient (changes in gain due to warmup internally and also ambient - room temp.). Anyone has to cut corners on inexpensive equipment, and the input amplifiers on the DVM could have thermal sensitivity.

Accuracy is an issue when someone says that this blend needs 203.5 and 8.5 bar. How was it measured? With what overall accuracy? For instance, we have finally realised that we have about a 1 degree difference in temperature reading between our M3 machine and that at Kaladi in Denver. Makes a big difference what each of us says about a particular nuance change in a roast.

Then there are issues about what is being measured with a particular technique. This will also foil the individual.

[HB]

John, good to see you on HB, I feel like I already know you a bit from your interview with Abe.

David brings up a good point about repeatability. I'm pretty slack about calibration and tend to work in terms of x+1.5F or x-2.0F from a given setpoint rather than worrying about accuracy compared to "true" standard because of the problems you noted. That prevents me from comparing precise temperatures with others, but then again, I'm skeptical that's a realistic expectation unless the two comparison machines are identical in design.

(Add to that the influence of Abe, Chris, Steve, Peter and others who have persuaded me to adopt a more holistic approach to espresso than my natural tendency to break out instrumentation at the drop of a pin. Go figure).

We know routinely that we can taste the difference of .05 bar.

You mentioned a similar point in your interview. I'm curious whether this is a perceptible difference that your everyday taster could discern and appreciate, or does this require a more refined palate?

Conventional wisdom has it that even 10x the difference you cite isn't worth discussing, but all the talk of the relationship between clarity and pressure delivery has encouraged me to rethink it. In fact, two of the early posts on this board are on this subject (Received wisdom about brew pressure and The Next Breakthrough in Espresso Technology) and it's come up several times in different forms (e.g., The Elusive Clarity in the Cup and The Air Pump Driven Espresso Machine). Of course I wouldn't be the first to suggest that uber-precise temperature control is passe and pressure technology is the new frontier.

[AndyS]

Of course I wouldn't be the first to suggest that uber-precise temperature control is passe and pressure technology is the new frontier.

Just like I wasn't the first to suggest that espresso tastes better when pulled in Celcius and bars rather than Fahrenheit and PSI.

[swines]

Absolute accuracy is not necessary if the device can be proven to repeat a measurement. For example, if you found that your thermocouple + meter showed +2 degrees higher temperature (when compared to a calibrated thermometer) everytime you took a measurement you'd feel confident that your measurement was accurate each time you used the instrument. You would know that the temperature was actually 2 degrees less than what the instrument was showing and would apply this offset to the measurement.

Conversely, if you tested the setup against a calibrated thermometer and found that the temperature reading varied each time you took a reading, then the variability renders the readings useless if absolute accuracy is needed.

I had a US Army research program where we ran into this exact problem using an array of thermocouples used to measure soil temperature across a defined area and at various depths over the area. We ended up using a NIST traceable process thermometer to find each thermocouple's temperature offset over the temperature range within which we needed to make the measurements. Once the offsets were know and verified as being repeatable for each thermocouple, we could generate the data sets knowing the offset required for each measurement.

[John]

We know routinely that we can taste the difference of .05 bar.

You mentioned a similar point in your interview. I'm curious whether this is a perceptible difference that your everyday taster could discern and appreciate, or does this require a more refined palate?

Let me preface the following with... By way of being in dialogs in Home Barista I must talk about my own experiences. The possible problem is that these experiences are with equipment that we have for sale. And therefore anything I say may be thought to be advertising. I hope that I don't transgress here...

The set up was: the M3 system at the Denver Nascore show last month. We encouraged every serious visitor to have two shots. We varied the pressure .05 bar between them. Since temperature is a much more time consuming adjustment, I thought the rather instant availability of pressure adjustment might cause a stir. Every person, several editors of magazines, local roasters, even a young woman (there with her boyfriend) with no real interest in coffee, all immediately aware of the difference. Our friends who visit us here taste it. So I don't think it takes an extraordinary palate.

It is our feeling that these are not minor flavor changes. We have learned, just this summer, about just how sensitive this process is. We were getting lots of variations in the cup that we thought was the natural result of the variability in the beans from one shot to the next or one day to the next. What it was was the small variations in brew parameters that we didn't imagine would affect the flavor, and that we therefore paid no attention to from shot to shot or day to day. Now we have serious repeatability.

We are not imagining things - nor exaggerating - when we say that the flavor change of .05 bar on the M3, 100 rpm grind speed on the M3 grinder, or 1-2 degree change on the Sivetz roast temperature on one of three beans in a blend - is a very serious flavor change. Any single one of these can make or break the flavor in a cup. And any one of these is repeatable.


I am quite hopeful about Home Barista. Glad to be here. Earlier this year I was very concerned that the entire web presence for espresso was in unfortunate circumstances - that is now changed very much for the better.

[HB]

The possible problem is that these experiences are with equipment that we have for sale. And therefore anything I say may be thought to be advertising. I hope that I don't transgress here...

Not to worry, that hasn't been a problem to-date because the participating sponsors / professionals understand the difference between sharing information and promoting their products. While there are no formally published rules for the board yet, Team HB has collectively created a mission statement and guidelines in the spirit of moderators as your hosts and discussion facilitators, not enforcers. Since moderatoring in the common use of the word hasn't been necessary (*knock on wood*), I've seen no need to post rules.

It is our feeling that these are not minor flavor changes. We have learned, just this summer, about just how sensitive this process is. We were getting lots of variations in the cup that we thought was the natural result of the variability in the beans from one shot to the next or one day to the next. What it was was the small variations in brew parameters that we didn't imagine would affect the flavor, and that we therefore paid no attention to from shot to shot or day to day. Now we have serious repeatability.

We are not imagining things - nor exaggerating - when we say that the flavor change of .05 bar on the M3, 100 rpm grind speed on the M3 grinder, or 1-2 degree change on the Sivetz roast temperature on one of three beans in a blend - is a very serious flavor change. Any single one of these can make or break the flavor in a cup. And any one of these is repeatable.

I consider myself an average espresso lover and thus assume such minute subtleties would either be beyond my abilities to detect or acceptable variance. To put it in basic terms by way of example... I like red grapes more than white grapes, but I'll happily eat either for lunch. But if I'm really in the mood for grapes and we only have bananas in the house, well, I'm disappointed. I am struggling to grasp if the differences between such minute pressure changes would mean nicely ripe bananas, hard green bananas, or simply red grapes / white grapes. I don't mind a little serendipity in my espresso as long as it's within an acceptable range. In contrast, based on the writings of some professionals, I imagine them flying into a rage if the temperature is off by 0.4F from the previous extraction. That's not me.

More seriously and directly to your point: Is the tight control you describe within .05 bar of a fixed pressure, or do you intentionally vary pressure through the extraction to produce a "pressure profile"? Jim Schulman postulated on the positive effects in Received wisdom about brew pressure (excerpted below) and I'm wondering if his findings are consistent with your own.

There's two disputable assertions on lever groups - that they have greater clarity and less crema, and that the reason is lack of vibration.

The first is an oft repeated observation -- I get it reliably and invariably when I set the La Peppina and Tea side by side, Mark mentions it in the Elektra review. Almost all owners of lever machines, home or commercial attest to it. Obviously, it could be wrong, but it's about as solid as any other fact in this game.

The reason for it being lack of vibration is mostly unresearched. I was the one who raised the pressure profile as an altrnative possibility; and, so far, am the only person I know of who's tried it. Most people just don't have big variacs lying around into which they can plug their machines or pumps. of course, even modified the E61's profile has the 10 second ramp up, so that's still a point of difference to the typical lever having a no pressure presoak for around the same length of time.

...

Thinking about this. I guess there's a lot of received wisdom about pressure, most of it virtually untested, and perhaps wrong. The knowledge about temperature variations is hugely more reliable, since adjusting temps is a part of lots of people's daily shot making practice. Adjusting pressure is usually done for maintenance or gross defects, not taste adjustments. So there's several orders of magnitude less data on it.

[rfc]

Are you sure you do not mean a half a bar? As in about 7.5 psi? I've never seen a machine with gradations more frequent than tenths of a bar. How is this being measured? If what is being claimed is a dramatic taste difference by .75 psi, then I have something yet to experience. For now, I'm not convinced.

[malachi]

Just because machines don't have gauges to display that degree of detail doesn't mean it isn't relevant.

My experiments show the same results as John's work does.

[rfc]

I was confusing the brew pressure with the boiler pressure, which this thread is really about. And I agree that .05 bar in boiler pressure correlates with nearly a degree and a half F., and that is most certainly detectable.

[John]

It is our feeling that these are not minor flavor changes. We have learned, just this summer, about just how sensitive this process is. We were getting lots of variations in the cup that we thought was the natural result of the variability in the beans from one shot to the next or one day to the next. What it was was the small variations in brew parameters that we didn't imagine would affect the flavor, and that we therefore paid no attention to from shot to shot or day to day. Now we have serious repeatability.

We are not imagining things - nor exaggerating - when we say that the flavor change of .05 bar on the M3, 100 rpm grind speed on the M3 grinder, or 1-2 degree change on the Sivetz roast temperature on one of three beans in a blend - is a very serious flavor change. Any single one of these can make or break the flavor in a cup. And any one of these is repeatable.

I consider myself an average espresso lover and thus assume such minute subtleties would either be beyond my abilities to detect or acceptable variance. To put it in basic terms by way of example... I like red grapes more than white grapes, but I'll happily eat either for lunch. But if I'm really in the mood for grapes and we only have bananas in the house, well, I'm disappointed. I am struggling to grasp if the differences between such minute pressure changes would mean nicely ripe bananas, hard green bananas, or simply red grapes / white grapes. I don't mind a little serendipity in my espresso as long as it's within an acceptable range. In contrast, based on the writings of some professionals, I imagine them flying into a rage if the temperature is off by 0.4F from the previous extraction. That's not me.

More seriously and directly to your point: Is the tight control you describe within .05 bar of a fixed pressure, or do you intentionally vary pressure through the extraction to produce a "pressure profile"? Jim Schulman postulated on the positive effects in Received wisdom about brew pressure (excerpted below) and I'm wondering if his findings are consistent with your own.

Dan I think that the fruit tastes analogies might work, but I think to refine the enquiry.

But first let's go back to the ability of you to detect the flavor variations. As I have stated, no one has yet to fail to identify them. You might not, but I would find it hard to believe that someone with your interest in flavor would fail. A .1 bar change was even clearly noticed by a young woman who doesn't drink or like coffee. We didn't think to try her on .05.

The difference is not like red or white grapes, which if both of fine quality give equal pleasure. It is more akin to the difference between wonderful grapes and adequate/slightly inadequate grapes of any kind. Then within further revisions of pressure one will find a whole panoply of different flavors - some appealing and some not appealing and they all are difficult to describe, but clearly different.

Bananas. Are you familiar with the possibility of finding a very fine ripe banana that tastes and smells perfectly as you imagine a ripe banana should and then another time finding a ripe looking banana that lacks all that quality. It looks the part, the texture is a little tiny bit off, but it just doesn't deliver the banana it promises. Almost tasteless and a bit mealy. This might take .1 bar.

Limes. Sometimes a lime is really brimming with limey sweet/acidic juice. More often today one finds limes which look like limes but deliver less juice and the flavor is certainly citric, maybe something of lime, but hardly satisfying.

Finally. I don't expect that any others have experienced precisely what we have - in fact I would be shocked! No one should expect to be able to hotrod a home or commercial machine in their spare time and have something that is on a par with what we did. Think F1 car versus very nice saturday night modified stock racer. While no one but home espresso fanatics were paying much attention to espresso quality, Versalab threw over 15,000 hours and $140,000+ into creating the most nutso espresso machine imagined. We are not stupid. If we could have accomplished 60% of the M3 by tricking out an available machine we would have. But we felt we saw too many problems with existing equipment so chose to go our own way. Then our development continually revealed new levels of clarity at each design revision. So we kept at it and at it. To some very fine levels of detail.

We presently use a fixed pressure brew.

[barry]

when you changed the brew pressure, did you change the grind as well?

[AndyS]

I was confusing the brew pressure with the boiler pressure, which this thread is really about. And I agree that .05 bar in boiler pressure correlates with nearly a degree and a half F., and that is most certainly detectable.

No, you had it right the first time. John is talking about flavor changes resulting from a change in extraction pressure of 0.05 bar.

[John]

when you changed the brew pressure, did you change the grind as well?

No. We did not need to. Flow through the bisket is pretty self-compensating providing the grind and pack are good.

[AndyS]

Flow through the bisket is pretty self-compensating

That's good, John. Illy calls it the "coffee cake," most people around here call it the "puck." But I kind of like the "bisket."

[HB]

The difference is not like red or white grapes, which if both of fine quality give equal pleasure. It is more akin to the difference between wonderful grapes and adequate/slightly inadequate grapes of any kind. Then within further revisions of pressure one will find a whole panoply of different flavors - some appealing and some not appealing and they all are difficult to describe, but clearly different.

Thanks for extending my analogy, it does help. Unfortunately at some point words fail and I have to take your word for it that the difference is startling despite the minute nature of the change. I would love to participate in a side-by-side comparison someday, e.g., at the upcoming SCAA conference in Charlotte.

Just a couple closing comments... in the past reviews, I've made an effort to suggest why a given machine performs better than another. Measuring in-basket temperature is one way to offer supporting data, although the latter posts of the Scace Thermofilter Temperature Device thread pointed out the potential failings of that approach:

What about all of those other profiles that you haven't measured? Bury the thermocouple 5mm down inside the cake and you'll get a completely different profile. What is the correct profile there?

To give readers a perspective on just how much placement makes a difference, see the chart below:


Cimbali Junior - puck top and bottom temperatures

That's one thing that really drives me batty about temperature discussions. Someone will say temperature X is THEE correct brew temperature for espresso Y. In reality the coffee is subjected to a wide spectrum of temperatures throughout the extraction.

Presumably the same logic applies to pressure: Does it not vary from the top of the puck to the basket's exit? Excuse my lack of knowledge of fluid dynamics, but how could a small difference of 0.05 bar produce a taste difference when that same extraction pressure is "experienced" mere millimeters below the uppermost surface? I suppose therein lies the danger of applying my limited engineering knowledge to taste; if indeed any Joe or Jane off the street would appreciate the taste difference with ease, I can live with not understanding why it matters and simply accept that it does.

Hmm-m, I would have never said that ten years ago... for whatever reason, I'm on a "holistic" kick lately.

[gscace]

Presumably the same logic applies to pressure: Does it not vary from the top of the puck to the basket's exit? Excuse my lack of knowledge of fluid dynamics, but how could a small difference of 0.05 bar produce a taste difference when that same extraction pressure is "experienced" mere millimeters below the uppermost surface? I suppose therein lies the danger of applying my limited engineering knowledge to taste; if indeed any Joe or Jane off the street would appreciate the taste difference with ease, I can live with not understanding why it matters and simply accept that it does.

I've seen some research that shows pressure is less important as pressure goes above 9 bars, but more important if it is too low. However, as I learn more about coffee and as I'm able to refine my process control and technique I learn more about these things and I'm not sure that the methodology provided in the study I read is without fault. It's certainly fertile ground. With respect to pressure distribution in the cake, the pressure profile is less complicated than the temperature profile. Pressure drop across the cake should be relatively constant in the flow cases here, since the velocities are fairly low through the cake and since there is supposed to be some headspace above the cake, which should result in constant pressure across the upper surface of the cake. In this case, pressure drop should be linear across the resistance (the cake) and pretty simple, although I'd add the deeper baskets drop pressure differently than shallow ones, since the resistance to flow occurs over a longer distance.

WRT temperature uncertainty, John's comments are correct, and people generally sweep the uncertainties under the rug. As far as thermofilters are concerned, right now I'm makeing them with only type T thermocouples, since type T is the best one to use for boiling wire temperature and I'm using probes constructed of "special limits of error" wire. That cuts down uncertainty as much as I think is economical, since most folks can't afford calibration services and I don't think they're really necessary anyway for what we are doing.

What gets confusing when one discusses measurements is the difference between accuracy and precision. One of the usual things that affects accuracy in thermocouple measurement with cheap readout devices is the level of care in temperature compensation of the readout device. If you really want to kick ass with thermocouple measurements, you make the measurements with a high accuracy voltmeter, and sink a referemce thermocouple junction in a well-insulated bath of distilled water ice. In this case, the voltage measurements are referenced to the ice point, and you can do very good thermometry. It ain't practical in real life and it's expensive (an Agilent 3458A is around $10k these days), so we make do. The way cheap readout mfrs make do is to use a thermistor to measure the temperature of a small block of copper inside the readout device, assume that the "reference junction" is at the temperature of the thermistor, and compensate accordingly. the thermistors used for this may not be very accurate, in the case of cheap readouts, and the degree of correlation between the actual reference jcn temp and the temperature measured by the thermistor may be pretty crappy. Howeve, one can get very repeatable answers if one recognizes this and takes some practical steps to minimize the problems. Good things to do are to keep the readout device away from heat or cold sources, don't handle it when making measurements, and always use the same readout device. In this way, you'll get good reproducibility from pretty cheap gear.

-Greg