Humidity sensor on exhaust - other gas sensors ? - Page 9

Discuss roast levels and profiles for espresso, equipment for roasting coffee.
GDM528
Posts: 830
Joined: 2 years ago

#81: Post by GDM528 »

Esteve wrote:Have you tried without the pump?
Airflow resistance is way too high for the IKAWA exhaust to reliably push air through the narrow tube, especially with two cigarette filters in the path. It feels a bit clunky to use a pump in the presence of IKAWA's powerful fan, but the pump is really tiny, low power, and enables so many other useful attributes. I'm hoping this catches the attention of someone with a drum roaster, as the tubing is about the same size as the hole one might drill to install a BT probe. The vacuum pump is strong enough to pull air through the tubing even if a thermocouple wire is threaded inside the tube, to make a combined temperature/gas probe:



I also tried sticking the tube into the same bag of greens that I'm roasting, and the TVOC reading was about 30,000. Probably just a coincidence, but that happens to be about the same reading I get when roasting the greens...
How is the sensor tray? Did you 3d print it or is off the shelf?
The custom-designed sensor tray was resin SLA printed - just a temporary solution until the final configuration of sensors is figured out. BTW, it turns out the finished SLA print is a significant source of VOCs. The cleaning process for SLA resin prints involves soaking in alcohol, which slowly outgasses over many days. When I convert to an FDM-friendly design, I now know to check what sort of emissions the filament produces.
One thing I didn't like when I had plastic tubbing was that I could smell plastic when roasting, even if my tubing was rated for well above the temperature produced by the roaster. Do you notice that?
After several runs, no detectable discoloration or odors from the silicone tubing, including empty-chamber runs. Perhaps your tubing had some residue inside it? When I first assembled the system and powered it up, the TVOC readings started off very high (30,000+) and took several hours of operation before it settled down to the low 1,000's.

User avatar
drgary
Team HB
Posts: 14294
Joined: 14 years ago

#82: Post by drgary »

GDM528 wrote:I'm hoping this catches the attention of someone with a drum roaster, as the tubing is about the same size as the hole one might drill to install a BT probe. The vacuum pump is strong enough to pull air through the tubing even if a thermocouple wire is threaded inside the tube, to make a combined temperature/gas probes.
Gary,

If someone with a drum roaster doesn't start adding to this thread, you're welcome to use my North 1Kg propane roaster as a test bed.

Also, is there a way to dampen the fan noise so that one can still hear FC?
Gary
LMWDP#308

What I WOULD do for a good cup of coffee!

GDM528
Posts: 830
Joined: 2 years ago

#83: Post by GDM528 »

drgary wrote:If someone with a drum roaster doesn't start adding to this thread, you're welcome to use my North 1Kg propane roaster as a test bed.
Awesome. I'd suggest for the sake of any other drum roasters that might be gas-curious, to post the journey here. For example, could you post a photo or two, of possible places to located a 1/4" silicone tube?

I can think of two concerns (there's probably more) for adding gas analysis to a drum roaster:
1) The roasting gas sample may be too concentrated and overwhelm the VOC sensor. That may require some sort of room air dilution widget.
2) Propane combustion gasses leaking into the roasting chamber, and swamping out the roasting gasses. I'm presuming the gas burner combustion fumes are aggressively isolated from the beans... are they?

As Esteve has demonstrated, the IKAWA system makes gas sampling super easy. So it's about time they finally added it to their Pro100x.
Also, is there a way to dampen the fan noise so that one can still hear FC?
I'm guessing you mean the air blower in the IKAWA roaster, which completely drowns out the sound of that little vacuum pump. Unfortunately, like 99% of all the appliances with electric motors, the fan unit is hard-bolted to the chassis. That turns the entire roaster into a sloppy audio speaker, with various resonances that drown out the cracking sounds. Just spit-balling, but perhaps wrapping the exterior of the IKAWA with sound-dampening mats will attenuate some of the higher frequencies that compete with the cracking sounds - will also put you in the running for an ugly roaster contest. Alternatively, some noise-cancelling headphones are tuned for attenuating fan noise (e.g. airliner jet engines), but still let higher frequencies through.

User avatar
drgary
Team HB
Posts: 14294
Joined: 14 years ago

#84: Post by drgary »

GDM528 wrote:I can think of two concerns (there's probably more) for adding gas analysis to a drum roaster:
1) The roasting gas sample may be too concentrated and overwhelm the VOC sensor. That may require some sort of room air dilution widget.
I wonder if this could be accomplished by having the gas sensor far enough downstream in the exhaust pipe. For my roaster, that pipe connects to a cyclone fan.
GDM528 wrote:2) Propane combustion gasses leaking into the roasting chamber, and swamping out the roasting gasses. I'm presuming the gas burner combustion fumes are aggressively isolated from the beans... are they?
Many roasters don't isolate the roasting chamber very well from the gas burners, especially those like mine with perforated drums. I don't know if that's a large problem if you have enough air circulation near the burners so that your flame is mostly blue.

Otherwise, is there a way to set up the sniffer to isolate the unwanted gases?
Gary
LMWDP#308

What I WOULD do for a good cup of coffee!

GDM528
Posts: 830
Joined: 2 years ago

#85: Post by GDM528 »

Repeatability test:

Ran three consecutive roasts with the gas sampler shown in post #78 to look for drift and/or excessive noise in the readings. Probably not a concern for just doing first-crack detection, but consistent readings would be essential for comparing different greens, roasting profiles, etc.

Used Mexico greens previously described, with a linearly declining RoR profile taken from the profile calculator described here: IKAWA inlet profile calculator "C7 170 262 D800" describes a moderately curved profile that's intended to come just shy of a Full City roast level - no second crack, but close.

I had let the gas sniffer run for over an hour on room air prior to use, so it was pretty thoroughly purged and was reading TVOC levels close to zero at the start of the test.

Preheated the IKAWA Home once with an empty chamber, then started logging the data. First run was another empty chamber to get baseline gas readings. Then ran three consecutive 65g batches of greens as fast as IKAWA's cool-down cycling would allow (about two minutes). As usual for the IKAWA, all the BT curves laid right on top of each other, so each batch saw the exact same thermal profile.

Chart's a little busy, but color-coding is your friend. I inserted a spike in the BT curves to show when I heard first crack for each batch:



The word "residual" comes to mind. Significant amount of carryover from the previous run. Kind of a bummer for reading the nuances of the curves in the early phases of the roast without having to insert empty-chamber purge runs in-between, which seems unreasonably burdensome. All is not lost however, as there may be some ways to accelerate the purge between runs with the sensors' on-chip heaters.

The carryover doesn't impede first crack detection and other milestones, however. Dry-end, first-crack, development, and impending second-crack are all still apparent in the RoR curves. For the next chart I've ditched the TVOC and favor of the better-behaved eCO2 curve. The eCO2 curve is still technically a VOC reading with some on-chip processing to guess at what the actual CO2 reading might be.



Water, Gas, and first crack all landed within 15 seconds of each other. And that's with the roughly six-second delay associated with the gas collection and sensing system. Also, the smoothing methods applied to the RoR data can also cause a bit of a shift in the timing. They're all close enough to each other to call them the 'same' IMHO.

I'd point out that the sound of a crack is just one bean, and therefore less than 1% of the batch - whereas the gas readings represent a 100% sample size. So to be fair, audible first crack sounds are a pretty good indicator of actual roast progression ;)

GDM528
Posts: 830
Joined: 2 years ago

#86: Post by GDM528 »

drgary wrote: I wonder if this could be accomplished by having the gas sensor far enough downstream in the exhaust pipe. For my roaster, that pipe connects to a cyclone fan.

Many roasters don't isolate the roasting chamber very well from the gas burners, especially those like mine with perforated drums. I don't know if that's a large problem if you have enough air circulation near the burners so that your flame is mostly blue.

Otherwise, is there a way to set up the sniffer to isolate the unwanted gases?
So: gas flame / heat shield / perforated drum? Yikes, that sounds like a lot of mixing. Burning propane produces water vapor and CO2, and the high level of heat applied might completely swamp out the same gasses produced by the roast. Burying the sniffer tube right into the bean mass may help, but it may still take some fancy-pants math to separate roast from the burners - especially if you're modulating the gas input.

If the gas doesn't have any leak-detection additives (food grade?), the VOC sensors may still produce useful results. I ran a quick test with a small blowtorch and the VOC readings didn't budge, whereas the eCO2 reading ramped up. That result is specific to the Sensirion SGP30 sensor that I'm currently using, so other VOC sensors may react differently - but now I know what to look out for.

GDM528
Posts: 830
Joined: 2 years ago

#87: Post by GDM528 »

In my service as a warning to others, I can report on how to save $80.

I searched for reasonably-available true CO2 detectors. The detection response time for nearly all of the sensors is way too slow, on the order of 60-90 seconds. However, there is one sensor, the Sensirion STC31, that is very fast: about 2 second response time. And Sparkfun sells a QWIIC board that makes it plug-and-play: https://www.sparkfun.com/products/18385

This is a full-scale detector, meaning it can read up to 100% CO2 concentrations. The sensor outputs a 16-bit measurement, so about 15ppm precision - about 500x its stated accuracy. I observed roughly 1,000ppm noise in the CO2 in room air readings, so this sensor operates well above the other sensors I'm evaluating, which typically have an upper limit reading of only 1,000ppm.

Going into this experiment, I had imagined the roasting greens to be like little grenades of water and CO2 that explode at first-crack. So I figured the STC31 would easily be able pick up all the CO2 outgassing during the roast.

The chart below shows two roasting runs: first run empty to create a baseline, and the second run with greens in the chamber. Just prior to starting the runs, I exhaled into the sensor pickup tube to show it correctly reading the CO2 in breath, and then tested a jar with vinegar and baking soda to create a CO2 saturated environment - those readings were as expected and show the sensor was operating properly before starting the roasting runs. The CO2 was compensated for pressure, temperature, and humidity. To show there actually were greens in the roasting chamber, I've included curves for the moisture release, which were spot-on for showing drying and first-crack.



Nada. Zero. Zilch. Bupkiss. Revising my "grenade metaphor" to just "water balloon". My favorite theory is the IKAWA floods so much air through the roast, it massively dilutes the CO2 from the greens, to the point where my own breath is a greater health risk than the roasting exhaust.

I'm eager to hear any suggestions for how I might extract useful readings from the sensor. I speculate it might be better suited to a larger capacity (electric) drum roaster with the sensor tube stuck into the roasting beans. For my IKAWA-centric purposes, I'll retire the sensor for now. The good news is this makes a homebuilt roasting gas analyzer much lower cost.

GDM528
Posts: 830
Joined: 2 years ago

#88: Post by GDM528 »

Moisture sensor steam-off!

I've managed to accumulate three different humidity sensors, so it was time to compare how they perform.

Sensirion SHTC3
https://www.adafruit.com/product/4636 or https://www.sparkfun.com/products/16467

Sensirion SHT31
https://www.adafruit.com/product/2857

Bosch BME688
https://www.adafruit.com/product/5046 or https://www.sparkfun.com/products/19096

They're all mounted into the same fixture so I can log their measurement data from the same air stream. In the photo, they're mounted on the underside of the 3D printed bracket and fed by a micro-pump that pulls air from the roast exhaust and pushes it over each of the sensors lined up in a row. On the top of the bracket is a wireless microcontroller/display, and a thermocouple reader. I just stick the silicone tube into the IKAWA exhaust port to gather a controlled sample of the exhaust fumes.



Here's the data from roasting a batch of Guatemalan greens. The absolute humidity curves were all normalized and level-shifted to conveniently place them on the chart. In the raw data, the BME688 readings were about 1% below the Sensirion sensors.



Good news: essentially the same results from each sensor. Scrutinizing the data, the SHTC3 appears to be the 'perkiest' of the bunch, responding sooner and faster. The SHTC3 is also the cheapest of the bunch, only $7-11 when mounted onto a plug-and-play board.

The BME688 looks to be the slowest, but also produces smoother data - I suspect this is because the chip is 2x oversampling the humidity data. The BME688 is also the most expensive, $20-25, but it also includes a sophisticated gas sensor system that provides yet another layer of interesting data about the roast - topic for another post.

Despite the speed difference between the sensors, they all lead me to the same observations about moisture release during the roast. All of the sensors were within 15 seconds of identifying audible first crack. They all show the same wiggly nuances that perhaps I/we will understand someday. So, you can pick your favorite sensor, and everyone's results should reasonably agree with each other.

User avatar
drgary
Team HB
Posts: 14294
Joined: 14 years ago

#89: Post by drgary »

Gary, thanks for your terrific research on this.
GDM528 wrote:My favorite theory is the IKAWA floods so much air through the roast, it massively dilutes the CO2 from the greens, to the point where my own breath is a greater health risk than the roasting exhaust.
There's one nitpick. The same amount of roasting effluent goes into your house even if the IKAWA is huffing a lot of air with it. Someone on a user thread noted that they had roasted without exhaust and found that it irritated their eyes and breathing. The risk isn't the CO2, it's the other stuff. And with the amount we're roasting at home in an IKAWA, it's probably only an irritant for most.
Gary
LMWDP#308

What I WOULD do for a good cup of coffee!

GDM528
Posts: 830
Joined: 2 years ago

#90: Post by GDM528 »

drgary wrote:The risk isn't the CO2, it's the other stuff. And with the amount we're roasting at home in an IKAWA, it's probably only an irritant for most.
Agreed, I confess I was being playful with the relative CO2 percentages. Less toxic than my breath can still be quite hazardous, and for yet other reasons.

Despite all of IKAWA's marketing that shows their roaster free-standing on a countertop, I always operate my IKAWA under a range hood with the exhaust hood fan turned all the way up. IKAWA's FAQ also recommends venting after a couple back-to-back roasts.

And to reiterate an earlier observation: the VOC readings I've been getting are very high... If nothing else, this exhaust analysis exercise has bolstered my respect for the health aspects of coffee roasting - and I'm gonna keep venting the exhaust out of my house.