Rob Hoos: Thermocouple lesson - Page 2

Discuss roast levels and profiles for espresso, equipment for roasting coffee.
User avatar
millcityroasters
Posts: 253
Joined: 9 years ago

#11: Post by millcityroasters »

wrakocy wrote:Been meaning to ask you guys about those 2mm probes for ages Steve.

I see you've got three different lengths (100mm, 75mm and 50mm) for sale on the site. Which would you recommend for BT and ET placements on a TJ-067?
Call me Tuesday and I'll walk you through your options.

User avatar
Almico
Posts: 3612
Joined: 10 years ago

#12: Post by Almico »

I'm much happier with a 6.4m probe in my Artisan 6. The 3mm is way too responsive and just bounces all over the place.

It might be the way the probe talks to Artisan through the Phidget 1048, but with a 3mm probe I get huge BT peaks and valleys. Readings do not even follow the ET.

The buffer from the 6.4 is just right for my use. The Artisan 6 is responsive enough.

Advertisement
User avatar
MaKoMo
Posts: 846
Joined: 16 years ago

#13: Post by MaKoMo »

Almico wrote:It might be the way the probe talks to Artisan through the Phidget 1048, but with a 3mm probe I get huge BT peaks and valleys. Readings do not even follow the ET.
Could it be that your ungrounded 3mm probe is broken and turned into a grounded one, or that it is originally grounded. What you experience might result from ground loop issues. The Phidgets is known to be relatively noise free on a proper setup.

User avatar
johnny4lsu
Posts: 775
Joined: 12 years ago

#14: Post by johnny4lsu »

I have 2mm probes and they aren't noisy at all.

User avatar
EddyQ
Posts: 1035
Joined: 8 years ago

#15: Post by EddyQ »

I'm going spew some of my electrical engineering knowledge. I don't know how significant this is, but with sampling systems, the analog signals being sampled should not contain frequency content above the nyquist frequency, which is half the sampling frequency. So, if you sample at 3 second intervals with Artisan, then your sampling frequency is 0.33 Hz. Your fastest analog signals from the thermocouples should then not exceed 0.167Hz. This is approximately at a rate of change of 1.6 degF/second maximum, which likely is slower than a thin thermocouple response.

For analog signals that exceed nyquist frequencies, the sampled spectrum folds back to a lower frequency spectrum and you won't know what the real waveform is. Basically, a digital smoothing algorithm would likely result in errors if sampled at 3 sec with use of fast changing thermocouples. In fact, if the noisy analog thermocouple changed up and down at the 3 second sample rate, you will get an offset error as big as the peak of the noise. So, perhaps a fat BT probe would be best if used with software than digitally smooths the sampled temperatures.

This said, I have been smoothing with my thin thermocouple with no issues. So, maybe the rapid changing portion of our analog signals are very small compared with the slow changing signals we care about. Maybe my Amprobe meter does some filtering before sampling. Something to be aware of though. It is always good to turn off smoothing once in a while and make certain the smoothed curve falls where you think it should on the noisy non-smoothed curve. . .
LMWDP #671

User avatar
Almico
Posts: 3612
Joined: 10 years ago

#16: Post by Almico »

MaKoMo wrote:Could it be that your ungrounded 3mm probe is broken and turned into a grounded one, or that it is originally grounded. What you experience might result from ground loop issues. The Phidgets is known to be relatively noise free on a proper setup.
The 3mm is brand new. I'm not sure if it's grounded or not, but I did some experiments this morning. This is my set up:



My normal probe is the lower one sticking into the roast chamber.

I added a second probe mounted to a bracket inside the roast chamber on Extra1 to see the differentiation.



I'm able to move the lower one in and out to match the reading on the one mounted inside.

The 3mm is hooked up to BT (Blue), the 6.4mm is connected to Extra1 (Black)

For the first roast the 3mm was inside:



That big wiggle between 2 and 4 minutes is the issue. There was no changes made to any settings. Then the probes diverged, culminating in a 30* difference by roasts end.

But in the next roast I put the 6.4mm inside and this was the result:



Very strange. No fluctuations.

These are my settings:


User avatar
MaKoMo
Posts: 846
Joined: 16 years ago

#17: Post by MaKoMo »

EddyQ wrote:I'm going spew some of my electrical engineering knowledge. I don't know how significant this is, but with sampling systems, the analog signals being sampled should not contain frequency content above the nyquist frequency, which is half the sampling frequency. So, if you sample at 3 second intervals with Artisan, then your sampling frequency is 0.33 Hz. Your fastest analog signals from the thermocouples should then not exceed 0.167Hz. This is approximately at a rate of change of 1.6 degF/second maximum, which likely is slower than a thin thermocouple response.

For analog signals that exceed nyquist frequencies, the sampled spectrum folds back to a lower frequency spectrum and you won't know what the real waveform is. Basically, a digital smoothing algorithm would likely result in errors if sampled at 3 sec with use of fast changing thermocouples. In fact, if the noisy analog thermocouple changed up and down at the 3 second sample rate, you will get an offset error as big as the peak of the noise. So, perhaps a fat BT probe would be best if used with software than digitally smooths the sampled temperatures.

This said, I have been smoothing with my thin thermocouple with no issues. So, maybe the rapid changing portion of our analog signals are very small compared with the slow changing signals we care about. Maybe my Amprobe meter does some filtering before sampling. Something to be aware of though. It is always good to turn off smoothing once in a while and make certain the smoothed curve falls where you think it should on the noisy non-smoothed curve. . .
Thanks Eddy for doing the math. According to my experience every meter (incl. the Phidgets or PIDs abused as meter) sample at a higher rate (just because I had the specs open, the new Fuji PXF is sampling at a 50ms rate while the previous PXG at 500ms, while the spec of the Phidgets 1048 states 25 samples/s). Therefore the relatively low sampling rate in Artisan (which can be maximally 2 samples per seconds with a 1sec rate + oversampling) should not play a role here, right?

Advertisement
User avatar
MaKoMo
Posts: 846
Joined: 16 years ago

#18: Post by MaKoMo »

Almico wrote:The 3mm is brand new. I'm not sure if it's grounded or not, but I did some experiments this morning. [cut]
Your second graph is just black for me, but the first one shows initially high noise on the 3mm probe. Your ET seems to suffer as well from this noise (based on the computed ET-RoR) at the beginning. If the ET shows less noise without that 3mm probe it could be that the 3mm is (or got) grounded and connects your meter to the ground of the roasting machine producing ground loop issues. If the ET noise without the 3mm probe is the same, the noise might stem from some EMI effects (which the 3mm probe might pick-up more than the 6mm probe). Does your burner/fan changes a lot more in this phase of the roast than during the rest of the roast?

Just trying to think along...

User avatar
Almico
Posts: 3612
Joined: 10 years ago

#19: Post by Almico replying to MaKoMo »

The 2nd graph shows black to me in Safari, but looks fine in Chrome. I'm not sure why that happens, but it's a regular occurrence for me on H-B. Some images are viewable, some are not.

Is there a way to test if my 3mm probe is grounded? Can I do a continuity test between one of the leads and the outer probe?

I use an air roaster so there is some EMI because of the fan. I have all the SS probe sheaths grounded to the machine and it seems to help. But there was no change in fan during that 2-4 minute section, but the burner was increased twice at 0:50 and 1:50, but no change was made at 3:00. But what's really strange is that the delta ET at the 3:20 mark increases <8*/M and the BT jumps 90*/M. Can't figure out why that happens, and it only seems to happen with the 3mm probe when it's inside the roaster, not the lower position.

Update: I reloaded the 2nd graph as a jpg and can see it now.

User avatar
Almico
Posts: 3612
Joined: 10 years ago

#20: Post by Almico »

I just did another roast with the TCs in the same configuration as graph #2: 3mm probe lower (Blue), 6.4mm inside (Black)

Interesting that the 3mm probe responded quicker to the charge, but when I bumped the heat around 2:50, the 6.4mm probe responded, but the 3mm did not.

I think this is more of a location thing than the thickness of the probe.