Temperature calibration

Hi all,

Is there a way to modify the interval between the temperatures used for the Temperature Calibration ? At the moment, for instance, our RT measured in eVOLVER is 23-24 C. The second temperatures is then between 35 and 36 and so we get up to 48 C for the last one. 48 C is actually pretty warm, and the machine has sometimes a bit of a hard time to stably get there, at least with the temperature of the room, affecting the accuracy of the calibration. Also, we always work with lower temperatures for the experiments (25, 32 and 37 max…but mostly between 25 and 32). So we were wondering if there would be a way to use 3 temperatures of the calibration that would be closer to the range of temperatures we use, as it would improve how good our calibration are (provided that things may not be 100% linear). At the moment, for instance, after the calibration, when we set the machine to a physiological temperature for our cells (30 or 32) we easily measure a 0.3 to 0.5 difference between what evolver measures and what we measure in the culture using high precision PT100. We can deal with that by setting the machine at 32.5 for instance, but that should not be necessary and we think that calibrating closer to the range of temperature we actually use would be sufficient.

Anyway we can modify that ?

Thanks !


Hi Damien,

I believe that is absolutely possible, and shouldn’t affect the calibration that much, since the temperature calibration is pretty much linear. It would require you fiddling with the electron app though (I believe this is the file you should look at). I’m not entirely sure what you should change there, but it looks like you could restrict the deltaTempRange variable. @heinsz what do you think?

Hi Marco,

Thanks ! At this point obviously, we are really playing around with getting really as optimal as possible. And I think the biggest issue is that with the temperature of the room our machines are in, the last temperature measurement and stability may not be as good as for the lower ones. And I thought this may explain the 0.3-0.5 difference in temperature we get when we go back to normal temps, if the last point is not quite right. So getting to slightly lower temps to have the machine working in a more comfortable zone in terms of heating may help…but that’s just a thought.


One thing to keep in mind is that the vials in the center are going to be warmer, since they are surrounded by hot sleeves on all sides. That might explain some of the differences.

Hi Marco,

Yes that we notice when doing the calibration. But once the calibration done, it yakes care of the bias. What we see is when switching everybody to 32, we have this difference between the evolver readings and our measurement accross all the 16 vials. The room is pretty cold (imposed because of other machines) and I think when calibrating, the system has a hard time to get to the last setpoint, altering the measurement that will be used for the calibration…but that’s just an idea and I was thinking of testing that by slightly lowering the target temps of the calibration. But again, this is just an hypothesis…


Sorry I’m late to this conversation. What Marco said is exactly right - you can change the deltaTempRange variable to change the way the set points are determined during a calibration. After you change this variable you would need to rebuild the application using yarn. If there isn’t a guide on this I’ll make one next week.

Would you mind sharing your fits? The temperature calibrations should be pretty much linear and work regardless of room temp. Even if the room temperature changes or can’t go as high because the room is cold, the PID should keep the vials at the set point. We have done calibrations in our cold rooms before with no problems that I’m aware of, but maybe @cmancuso or @bgwong can elaborate on that.

Hi Zach,

Thanks ! I will get the fits and post them. But they look pretty good…I just thought that as we find systematically a 0.3-0.5 C delta between what eVOLVER measures (after the calibration) and what we measure in the vials, something with the calibration must be slightly off. Or do you think that we should not try to get any more accurate ? I only thought we should try to improve it as I believe you mention something more like 0.1 C error in the paper…



Hmm, sorry to hear that, its puzzling. To clarify: the temp in the vials is consistently lower than expected temp or consistently higher than expected temp?

If consistently higher, then it could be a waiting time during calibration thing: water hadn’t heated up during calibration so it’s overdoing it during longer experiments.

If consistently lower, that’s harder to parse out. I’ve not experienced this myself, but if you changed the splash guard since calibration, maybe the local temp near the thermistor is now much higher than the temp on the sleeve? Or maybe convection in the room changed, achieving the same thing?

The +/- 0.1 I believe was a precision, not accuracy, but that’s tied to the PID algorithm, so it should be independent of the conditions. The accuracy would be affected by any variable that changes the temp difference between the media and the location of the thermistor (same as with an incubator).

1 Like

And yes, calibrating over a slightly smaller thermistor value range is a good option for temperature, since it’s linear and you want to make sure temps are stable before measuring.

OD is better to calibrate over a wide range, since it’s not linear.

Hi Chris,

Thanks a lot for your input. Again, we could really work with what we have, but we want to get things as optimal as possible from Day 1 so that it rolls in the future.

The temp in the vials is always slightly lower than the temps measured by the thermistor. The one thing you mention is actually something we are about trying. The room is used by other machines and there is quite “strong” AC, creating some airflows in the room, which may affect the calibration itself…and may also slighlty cool down the sleeve (and overall the vials), while locally at the level of the thermistor it remains a bit higher. So early next week we were planning on killing the AC during the temp calibration and then checking what we reach when putting our temperature at fixed physiological values, all this without any airflow. This is planned. If this helps, we will simply isolate the eVOLVERs for the airflow !

It is true that the stabilisation of the temperature for the last point of the temperature calibration takes forever in our conditions. The first 2 are really quick…but the last one…

I will keep you posted when we have tried without AC !

Thanks, again, for all the support !


1 Like

Yeah, it should be something you can rely on, so I agree it important to get this right. I wonder if changes to the layout in the fynchbio layout also potentially have affected the temperature gradient. Spacing between sleeves, anodized vs painted sleeves, splash guard design, could all theoretically affect convection rates in a breezy room. We should double check on our systems too.

1 Like

I totally agree with all of the above. A couple more points:

  • Be careful where the temperature probe is in regards to the magnets, stir bar. The magnetic field messes with the reading of the sensor. On the other hand, if not fully submerged, that would impact the readout also.

  • We have noticed that if you use a thermal couple, the readings aren’t accurate enough. Though I think this isn’t an issue because you are using the same probes as us.


Hi Brandon,

I will test stopping the stirring when doing the measures. But if one consider the stirring is somewhat stable, it should be also taken care of through the calibration I guess.

For the temperature measurements of the vials, we actually use a very high precision, multi-probe systems with really accurate PT100 (actually I think it’s overkilled of the calibration as the temperature resolution of this system is like 0.01 C or something like that)…we already had that in the lab so at the end we saved a couple of $$ by not buying the Fisher sensors.

My feeling is that the air convection in the room is a large part of the problem as it is quite strong. I just moved one of the eVOLVER in an office, with no air movement whatsoever (except for that of my brain (-: ). I am doing a Temp Cal tomorrow to check if that helps. I will keep you posted. If so, that simply means that when installing a machine in such conditions, one wants to protect it a bit from air flows.



I will test stopping the stirring when doing the measures. But if one consider the stirring is somewhat stable, it should be also taken care of through the calibration I guess.

Sorry I was unclear, I mean to just say that make sure the probes don’t get too close to the magnet, this would cause the sensor give a bad reading.

Hi guys,

Sorry to bother again but we are really struggling with the temperature calibration. Help needed ! (-:
So I isolated the machine in an office (so no airflow whatsoever) that is at 20 C RT. I used shorter probes to make sure they are fully immersed, and thinner so that they don’t touch the stir bar. So best possible conditions. I checked that these probes give the same results (at 0.01 C they are identical…pretty good probes !!).

The problem is the following: the first measurement is around 22-23, which I can understand although the room is at 20-21 because of the machine, the stirring etc…Then for the + 10 temperature, I would expect 32-33…but I am, once stabilized (overnight, orange circles on the software are full), at 38 C, so rather +15 than +10. Then I switch to the last temperature, and after 10 hours, the orange circles on the software are more or less 1/2 way and have stopped progressing for several hours…the temperatures are at 51 so rather +13 than +10…I do the measurements again, enter the calibration in the machine. Then I start looking at the correspondence measured by eVOLVER / measured by probe and it’s a bit all over the place, with sometimes eVOLVER and measure being close, sometimes eVOLVER is higher, sometimes eVOLVER is lower than with the probes, with differences ( in both directions) that can go up to almost 1 C…

Is it possible that the machine has a very hard time to get a stable temperature when it’s at more than 50 C, making the measurements really variable and so the calibration quite imprecise from one sleeve to the next and over time ? We never managed to actually get the software to consider the last cal temperature stable (at least based on the orange circles).

It would be great (if it is easy to do), just to try and solve this issue, if we could get some sort of non-official version of the software that sets 3 pre-defined temperatures rather than +10 (which is not + 10 in our hands but more like +15). Like RT, 30 C and 40 C. Because the machine does stabilize without a problem around 40 and we may then have better and more robust calibration. Or indeed restrict the range (but I am not sure how to rebuild the app myself…not an expert). Right now the range is 0,1000 with 3 steps, and it is indicated that it’s around 0.02 C per a.u…I am wondering if actually this is not quite right, that is each a.u. is more than 0.02, explaining why we have up to +15 C between points, and we struggle with the last point which never get high enough.

Thanks a lot !


Is it possible that the machine has a very hard time to get a stable temperature when it’s at more than 50 C, making the measurements really variable and so the calibration quite imprecise from one sleeve to the next and over time

This is precisely what is going on, we don’t actually expect it to reach the final temperature. So the last measurement is to determine what the temperature ceiling is for that sleeve, since each one is variable. I mentioned this when you were in Boston but might have been missed in the loads of information presented. The calibration only depends on what values are entered and the measured arbitrary sensor values. The +10 C is inaccurate (probably like what you said, +15) but that doesn’t matter since all it’s doing it setting it at a reasonable range for you to measure via the thermometer, by hand.

In summary, even if it doesn’t complete the progress bar for the 3rd time point, that is OK. Just go ahead and measure the temperature and enter whatever values they stabilize at. The calibrations shuold work just fine.

Hi Brandon,

Thanks a lot. Yes I may have missed that point, sorry about that. But that’s what we do at the moment, measuring the last point as high as it can get. But the calibrations we get then give us weird results afterwards once we set the system to more normal temperatures (that is 32 C. We then get differences between the measurements of the machines and what we measure by hand of 0.3-0.4 C in the best case scenario…with the vials being generally colder but some warmer. Today I had up to 1.3 C difference). So there must be something wrong elsewhere. I will keep troubleshooting. I am sure it will eventually get sorted !



I am also happy to test if reducing the calibration temperature range as Zach suggested makes it work better for us. Zach if you have a moment to write up how to rebuild the app once modifying the range (I would try 0,700 instead of 0,1000), that would be awesome. Then I can play around and report if it helps at all for us.

Thanks again for the support ! We will get there (-:

PS: We changed a number of PDs to improve the OD90. New cal next week and I will report.

I’ve also seen this issue pop up recently. I am not sure where the issue is coming from. It might be time to tweak the PID controller variables on the Arduino code again. Some of the software might have changed within the last year such that we need to revisit that again to re-optimize.

I don’t have immediate time to do it right now, but if you want to try yourselves, the Arduino code is there. Just follow the tutorial and upload to the U2 controller. (When opening the hood, should be 2nd controller from the right to left. U1 should be stirring)

Sorry I was offline for a few days traveling. Let me know if you guys have other questions about the electron application. I’ll make a guide this week and post it here on how to rebuild it for future users.