Issues with OD/temp measurements

We’ve had a couple issues pop up during our last experiment(s):

  1. [OD1] Two vial that were blanked, but rose immediately (within a couple minutes) to an OD of 0.3 despite not being inoculated. However, upon looking at the vials there was clearly no growth in either one.

  2. [OD2] One vial seems to no longer be able to read OD measurements. In the txt file, data showed up as “nan”. My hypothesis is that wet tubing to connect to other vials may have damaged the electrical components based on our set up.

  3. [Temp1] Normally, we start the experiment in the “Set up” menu to prime the pumps and warm the vials. However, I noticed that even after exiting the “Set up” menu and starting the experiment, if the temperature had already reached the temperature it would be running in the experiment (set in custom_script.py), then upon starting the experiment it would plummet towards room temperature. This doesn’t happen if the temperature is not rising during “Set up” or when I restart the experiment.

Has anyone else run into these problems and found solutions?

I think your setup used 180 degree measurements which we later concluded causes a lot of noise and still has fairly low dynamic range. Since you’re close by we can just come over and help out with modifying it back to 135 degrees.

Some of the new code would help the temperature issues you have. There are bugs in the old code that causes the temperature to change. A fresh update should help everything.

We are facing similar issue with the OD measurements - OD data for some vials shows ‘nan’ throughout the experiment. At first we thought it is a hardware-related or calibration issue but it is happening to a different vial in each experiment. Any suggestions?

@anupama We did think it was a hardware problem as well and have had our sensors replaced. Are you using the original sensors that came with your eVOLVER?

Also, are you presetting the temperature prior to your experiments? Some of mine would be nan with a vial and media, and then when I restarted the experiment with that sleeve empty, they would read od values again.

Another reason that I thought it could be was that the tubing is often wet during the initial set up (not soaking but there is some liquid droplets), I was worried about parts of wet tubing coming in contact with some of the hardware. At least to me it seems as though this could explain why it was a new sensor every time and why the problem seems to hit the vials away from the pumps more often. I started drying off the tubes before hooking them up to the eVOLVER now, maybe this could help?

1 Like

Hi Anupama,

I think this maybe related to using the wrong calibration file (od 90 vs od 135). I think your eVOLVER only comes with OD 135 measurements. We recently pushed an update for new calibration code on how to handle that. @heinsz may have more info.

Try running the following:

python3.6 calibrate.py -a "<ip address> -o -f <calibration name> --param "od135"

Can you show what is displayed?

@bgwong The new calibration code is almost ready. Running what you said above looks correct for the current release, but watch the quotes.

Thanks a lot everyone for your help. Here are the new calibration curves for od135:

They are definitely better than the curves for od90 (shown below) for lower values of OD but as you can see the fit for od135 is really bad for some vials (0, 5, 6, 9).

Any ideas what could be causing this?

Hi Anupama,

Can you make sure several things:

  • All the wires (ribbon cables) are properly pushed down and are connected tightly (both ends)
  • Temperature was equilibriated (at whatever temp you want to run the expt at) when you run the experiment
  • When doing calibrations, the vials are capped with the nylon 3D printed parts. This helps the vial not wiggle around in the sleeve
  • Stir rate is set to something that doesn’t obstruct the reading (e.g. stir rate of 8)
  • There is at least 25 mL of media in your vial when calibrating.


Thanks for the suggestions Brandon. All the things you mentioned above have already been taken care of in doing the calibration. We will try again and be extra careful this time. Will let you know how it went.


We might want to try to update your GUI before doing the next round of calibration. Please send me a note privately and we can setup a time to get that updated.

1 Like

Issue fixed on release 1.1.0.

Here are the OD calibration curves for the 1.1.0 -

The issue of getting nan values is still not resolved! We ran a test experiment where we let the cells grow in all 16 vials overnight without any influx/efflux and only to measure the OD over time. Here are the results from that experiment:

For some vials the OD values become Nan even before reaching 1, which is strange!

Any idea what could be causing this? Thanks.

This is expected because the OD135 signal maxes out at around .5 to .8, with variability depending on how the vial is constructed. If it goes outside the fit value, then there are NaNs. For better signal at higher densities, I would recommend using OD90 signal.

Thanks for your response! Some of the vials start throwing nan before even reaching 0.5, for example see vial #7 and #9. Also the calibration curve for vial #9 looks very noisy, do you think changing sensors for these vials would be a good idea?

Yea I think changing the photodiode would help significantly. These have a smaller dynamic range based on the calibration plots, so this is expected (esp vial 9).

1 Like

Any idea why OD measurements starts decreasing instead of saturating after few hrs? We initially thought that the cells are dying due to some reason but then we took the samples over time and measured the OD on a cell density meter (see the plot below: the black lines are reading from the eVOLVER and pink x are readings from cell density meter)

Our initial guess was that it is because calibration function is mapping it to something random so looked at the rawOD values and found that even they were decreasing, which is strange given the real OD is increasing. Is there any explanation for what could be causing this? We were wondering what raw OD values actually represents and how can we get rid of this pseudo drop in the OD values?

This is really important for us because many of our experiments involve growing cells under different drug concentrations and dosing strategies for long time, and looking at the final OD to compare.

Did you only calibrate the eVOLVER up to the 0.8 measurement and are only using the OD_calibration_oct01 (od135 calibration)? Based on your calibration files I would imagine anything approaching 0.6 (or 0.3 in some cases) might not be mapping correctly.

We have had a lot of success using the 3d calibrations instead of od90 or od135 alone. When I did my calibrations I used OD measurements at 0, 0.05, 0.1, 0.15, 0.2, 0.25, 0.3, 0.4, 0.5, 0.65, 0.8, 1, 1.25, 1.5, 2, 3 (since normally I run from 0.1-0.2, but other people in my lab may want those higher ODs in the future).

To get the 3d fit instead of od90 or od135:
python3 calibrate.py -a <ip_address> -n <calibration_name> -f <fit_name> -t 3d -p od_90,od_135

Also, the raw OD measurements should be decreases as actual OD increases since there is an inverse relationship. You could try plotting these and check that it is always decreasing or stable, if so you might just need to do higher ODs during your calibration.

Good luck!

1 Like

Thanks for the suggestions. Yes, I used the OD_calibration_oct01 for calibration, which is for OD135 calibration and I understand that it breaks down after 0.8 (or even lesser value for some vials) but what I don’t understand why would it start dropping and then saturate to a lower value (which is within a workable range). Ideally it should be monotonic, at least the raw values. The raw values are also non-monotonic over a long time and start increasing (sorry I mistakenly wrote decreasing before) after few hrs. Here is the plot of raw OD135 values:

What worries us is that even if the final OD after 10 hrs is within a range we consider accurate (less than 0.8 or 0.5) it is not a value we can trust.

Also, I am not sure if our eVOLVER has the sensor for OD90 @bgwong ?

Interesting… yeah I was also expecting that they’d at least be monotonically increasing.

I think you should have the od90 sensor since you posted the calibrations for that sensor in this thread on Sept 27? Is there anything in the od_90_raw folder for your experiment?