Greg,
Any idea when you might have a Spaziale version of your new device?

Sorry to be asleep at the switch here. I didn't see this post until June 8. The deal with the Spaz's is that I drew up and built two laSpaz thermofilters. I used one to make sure it worked right, sent both to Terry and off they went to folks who wanted them. The groundswell of desire for 53mm thermofilters after those two left was a bit underwhelming, so I didn't invest in any parts inventory. If folks want em I'll build em, but I gotta believe that there's demand before I sink money into machined special parts. Lately I've had a couple of requests, so I'll prolly make a few specials once I get on top of the Scace2 thing, Send me a private message with yer email address in it and I'll let you know when I do.

-Greg

I only had the thermo filter for a couple of hours before the flow rate valve blocked. What is the best care and best way to clear the hole :?

from coffeed.com, here:

gscace wrote:Hi there:

Unscrew the filter / flow orifice from the plastic body. Unscrew the brass orifice cap from the filter. Rinse the daylights out of the orifice. Make sure that it is absolutely clean and that there is no loose thread sealant. Look at something light coloured thru the hole to make sure it is clear. You can use a very fine needle to gently poke at the hole if you need to, but be careful you don't enlarge the hole. Rinse / wash the daylights out of the filter, on both sides. Clean all of the old thread sealant off of the filter threads. I've used Loctite 242 (removeable, blue stuff) as thread sealant when installing orifice caps onto filters. Non-hardening pipe sealant works also. There's a bit of a fine line between too much sealant, which can come loose and plug up the hole, and not enough, which leaks. Try not to have excess sealant on the tip of the filter. Also, notice the direction of flow arrow on the filter and assemble it right. Finally, install the filter / orifice back into the plastic body. I've discovered that you don't need to use any sealant here. The plastic seals against the brass just fine. Remember that these are tapered threads. If you crank down on them you can split the orifice cap, and you can also damage the plastic body. Don't be a gorilla.

Have fun. lemme know if you have problems. I'm gonna be at a conference next week, so I'm gonna be incognito from Monday thru Thurs as I ain't takin a 'puter wid me. Lemme know if you have trouble before then and I'll write you back with helpful suggestions and crappy English.

-Greg

steve wrote:I only had the thermo filter for a couple of hours before the flow rate valve blocked. What is the best care and best way to clear the hole

PM me if you need more help, or don't feel comfortable doing it. The orifice hole is pretty small, but it's sposed to be protected from clogging by a filter. Lemme know if you want me to fix it for you or if you need any other help in getting it to work properly.

-Greg

Wanted to say first that I think the Scace Device is really a cool idea and I'm happy that he came up with it. Being able to consistently measure the temperature across machines and adjustments (even if there appears to be a lot of debate on what it might mean) is awesome.

I did have a couple of questions about the accuracy that is being discussed though. Pardon me while I work through this.

I don't have any idea what thermocouple probe Greg is using other than Type T, so used the assumption that he's using high-tolerance T thermocouples with a tolerance of +/-.4%

Assume that you would plug that into something like a Fluke 50 Series which for temperatures <100C has an accuracy of +/- [.5% +.3C]. (Which I think most people would. Decent price for pretty decent accuracy.)

If we assume that water at 96C is being felt by the probe in the device, that would mean that the measured temperature would be 96 +/- 1.18C (.004*96 + 0.3 + 0.5) or 94.82-97.18C (202.7-206.9F).

So the numbers may not mean what people want them to. (For instance, if I got a Scace Device, went home and was desperately trying to get Schomer's '203.5F' I might succeed at that at the expense of having a nice temperature, but lousy espresso.) In fact, the original 203.5 really only means that Schomer's original rig read that when he got great espresso.

The point being, that me getting 96C on my device and you getting 96C on yours doesn't mean anything more distinct that we know that we are within 2.36C of each other. (And don't get me wrong, 2.36C is pretty darn good.) But people seem to be assuming a much tighter margin of error.

Which I think means I'm really missing something. Help?

-Jesse

cdrikari wrote:Which I think means I'm really missing something.

For the record, I use the thermofilter almost exclusively for determining the thermal characteristics of an espresso machine as part of equipment reviews. The particular temperature readout, whether it is 202 or 203.5, rarely concerns me. My interest is that the thermofilter indicates the same "X" readout when subjected to "X" temperature every time. In other words, I'm after consistency, not absolute precision for the purpose of sharing precise temperature measurements that others could then theoretically reproduce when brewing espresso.

That said, if you want to compare temperatures with others, the first step is calibrating the device in boiling distilled water. The error limit is significantly smaller between say 212F and 203F than two arbitrary temperatures within the thermocouple's possible range. As to the precise error limit of a type T thermocouple between boiling water and typical brew temperatures, I humbly defer to Greg's greater knowledge and experience (note: Greg's response may be delayed because he's probably enroute to the SCAA conference in Minnesota).

Which I think means I'm really missing something. Help?

I certainly defer to Greg also but in the meantime:




All you need to know is the temperature of boiling water/condensing steam in your location at that particular time. In my case, Maryland water was 211.3 F so I made a tag for the thermofilter and adjusted the meter accordingly.

I'm completely with you on the consistency end of the scale, and for home use, you're dead on for the altitude-compensated boiling calibration as well. (Though I'm guessing a few enterprising souls have access to more extreme equipment...and have probably used it. Like Greg for instance. )

*blink* Right. Duh.

You calibrate using altitude compensated boiling point. The boiling point is right next to where you want to measure, so it's darn near ideal. If everyone passes the cal result along with data -> no issues.

I'm going to go see if I added stupid sauce to my morning coffee today. (And maybe pull a shot while I'm there.)

Cool. Thanks for providing the required calibration. (pun intended.)

cdrikari wrote:Wanted to say first that I think the Scace Device is really a cool idea and I'm happy that he came up with it. Being able to consistently measure the temperature across machines and adjustments (even if there appears to be a lot of debate on what it might mean) is awesome.

I did have a couple of questions about the accuracy that is being discussed though. Pardon me while I work through this.

I don't have any idea what thermocouple probe Greg is using other than Type T, so used the assumption that he's using high-tolerance T thermocouples with a tolerance of +/-.4%

Assume that you would plug that into something like a Fluke 50 Series which for temperatures <100C has an accuracy of +/- [.5% +.3C]. (Which I think most people would. Decent price for pretty decent accuracy.)

If we assume that water at 96C is being felt by the probe in the device, that would mean that the measured temperature would be 96 +/- 1.18C (.004*96 + 0.3 + 0.5) or 94.82-97.18C (202.7-206.9F).

So the numbers may not mean what people want them to. (For instance, if I got a Scace Device, went home and was desperately trying to get Schomer's '203.5F' I might succeed at that at the expense of having a nice temperature, but lousy espresso.) In fact, the original 203.5 really only means that Schomer's original rig read that when he got great espresso.

The point being, that me getting 96C on my device and you getting 96C on yours doesn't mean anything more distinct that we know that we are within 2.36C of each other. (And don't get me wrong, 2.36C is pretty darn good.) But people seem to be assuming a much tighter margin of error.

Which I think means I'm really missing something. Help?

-Jesse

Hi there:

Here's the late response because I don't always go through all of the forums in hb, although I'm supposed to do this one at least...

Uncertainty in thermocouples doesn't actually follow a neat percentage of the temperature. I use Type T thermocouples because they are particularly good at boiling water temperature. This causes some headaches for folks who would rather buy a cheap type K readout device, but if you want accuracy you gotta do the type T program. There's an ASTM term called "Special Limits of Error Class 1", which refers to wire with enhanced accuracy meeting their spec. I use probes that meet this spec, and they have an uncertainty in temperature of 0.9 Deg. F.

Fluke publishes an uncertainty spec of 0.6 deg. F for their series 5X thermocouple readouts. This uncertainty fits suitably with the probe uncertainty, as you will see.

Imagine that you buy a bunch of probes - maybe a couple thousand, and you make it your life's work to calibrate all of these damn probes, plotting the results on a graph. You'll find that most of them cluster around the temperature you'd expect, with less and less of them further away. About 95% of them will be within the accuracy spec of 0.9 degrees and a few will be outside of that. If you plot the number of thermometers that deviate by a certain amount away from the expected value, against the deviation amount you'll be likely to come up with that bell curve thing.

The same deal probably applies to the readout device as well in that most of them are really close and a few are not.

The two uncertainties are "uncorrelated" in that the uncertainty of the thermometer doesn't have anything to do with the uncertainty of the readout. They are two different devices and you might have two perfectly accurate ones, or you might have a probe that reads high and a readout that reads low, or any combination of things you can think of that partially cancel or don't, or whatnot.

Since they are uncorrelated, you can calculate the combined uncertainty of the system of thermometer and readout by calculating the RSS (root sum squared) value of the two uncertainty components. Here's how.. Square the uncertainty of the probe (0.9^2 = 0.81) and the uncertainty of the readout (0.6^2 = 0.36). Add them together and then take the square root (0.36 + 0.81 = 1.17, and 1.17^0.5 = 1.08). Rounding 1.08 to the nearest tenth of a degree gives 1.1 degrees, which is the combined uncertainty of the probe and readout in degrees F. That's about 0.6 Deg. C for you SI junkies.

I mentioned that the Fluke is pretty well suited to the probe in terms of accuracy spec. Notice that the squared values of probe and readout uncertainty differ fractionally more (.81 is more than twice .36) than their non-squared values (.9 and .6). If you used a readout device that had an accuracy spec of say 0.2 degrees, the squared value would be 0.04, and the combined squared values of probe and readout would be 0.85. The combined system uncertainty would be the square root of .85, which is .92. You'd spend a boatload of money to buy a readout with 0.2 degree accuracy and you'd only gain 0.2 degrees of accuracy in the system - .9 vs. 1.1 degree.

It's useful to differentiate between precision and accuracy at this point. The Fluke meter has a reproducibility spec of 0.1 degrees, and the reproducibility of the probe is prolly on the order of 0.01 degrees. So if the same probe and meter are used to measure the same constant temperature process, the repeated measurements will be within 0.1 degree. That means that if you use the same probe and readout you can transfer machine settings to the tenth of a degree level.

On the other hand, suppose you mention to your friend that you like coffee A at 200 degrees and he sets his machine up similarly using a separate Scace and readout. Now you have two uncorrelated systems with system uncertainties of 1.1 degrees F each. You handle them just as before and calculate the RSS value of the pair. (1.1^2 + 1.1^2)^0.5 = 1.55, which rounds off to 1.6 degrees, which is the uncertainty value for the brew temperature of the second machine, compared to 1.1 degrees for the first machines temperature uncertainty.

These uncertainty numbers are entirely reasonable at this stage of the game in espresso. There are very few espresso machines that have one-degree reproducibility. The uncertainty figures for the probe and readout of a single system (or pair of uncorrelated systems for that matter) are about the same size. We do know that if you have a very reproducible machine and very good technique, you can discern the taste difference between pairs of shots brewed at a half degree or so different. Note that I said discern. I didn't say that one will suck and one won't. You can just taste a difference. In this instance, using a multiple group machine with independent temperature control of brew boilers, it's likely that you will be using the same probe and readout device, so the reproducibility spec is entirely adequate.

Now for calibration - Thermocouples are used in Scace devices because they usually work or they just plain don't. The operating principle of thermocouples is pretty fundamental and generally no calibration is needed, although I've learned of a few bizarre modes of failure since I've been making these things.

You can calibrate the system as Eric S demonstrated. I posted how to do it somewhere on this site in the past, but basically the deal is to invert the Scace, and suspend it over furiously boiling distilled water that is boiling in a tall pan, like a huge steam pitcher, with the level of the boiling water maybe half way up the huge steam pitcher. You knew there had to be a legitimate reason for those huge 2 liter pitchers that get used in crappy espresso shops, right? YOU HAVE TO USE DISTILLED WATER!!! You want the Scace to be pretty close to the boiling water so that it is suspended in the steam above the surface. Here's the neat part. Steam, which is water vapor, contains a lot of heat because it's in the vapor phase. When it changes back into the liquid phase it gives off most of that heat, which is why steam burns are so destructive. In our calibration vessel the released heat from condensation keeps the region just above the boiling water at very constant temperature. So you can get a pretty damn good calibration answer - with uncertainty of a few hundredths of a degree. Once you get the temperature reading you have to correct the boiling temperature for atmospheric pressure. 212 Deg. F is the boiling point at seal level at standard barometric pressure ( 14.696 psi). The quick and dirty way to get local atmospheric pressure is to assume that your local airport is at the same pressure. This may or may not be right, but it's prolly good enuff unless you are living on your private mountain and the nearest airport is in da valley somewhere. Boiling point corrections are easily found on the internet using your favorite search engine.

An interesting result from calibration is that the system of probe and readout are now highly correlated in that only that probe and readout will produce the calibration result you obtained. There's no way to know which of the devices - probe or readout - contributed most to the difference between measured value and actual value, but we dont' really care anyway.


You can now look at the difference between the real boiling point and measurement, and apply the difference as a correction to your measurements, giving you a high degree of accuracy. If you calibrate multiple systems using the method outlined here, the uncertainty propagated by using independent measurement systems goes way down, allowing you to transfer temperature settings to the tenth of a degree level if you really wanna.

-Greg