Skip to content

Calibration testing results

title: Pipette calibration: tests and results description: Gravimetric calibration of micropipettes tests and results published: true date: 2023-11-03T07:14:35.061Z tags: editor: markdown dateCreated: 2023-07-28T15:16:03.573Z

Warning

This is really old information and should not be used.

Calibration: Tests and results

21-07-05 p-200 calibration results

In general, more volume was deposited when loading 5% extra, although it is less marked/safe at low volume.

All differences were ran up, showing the effect of "underloading".

However:

  • There is a tendency to deposit "over" at 150 uL. But the error is within 2% and it is reasonable. You have to check that it is not systematic.
  • Charging 5% more at 20 uL did not noticeably correct the error. Still charging less apparently.

Some interpretations:

  • There is a problem that causes it to load from less to less volume, and from more to more volume. Charging 5% more just made him cum up.
  • Overloading is the situation analogous to serial pipetting:
  • So it makes sense that loading 5% more fixed the parallel pipetting error at 50 uL (serial pipetting was 40 uL, see 21-06-15-p200-calib).
  • The tip gets a bit wet, so you always lose something:
  • So when the tip gets more wet by charging more, you also lose more from getting the tip wet.
  • Are the pipettors calibrated to compensate for this? In this case, it should happen that for every 1uL pipetted, an extra bit is added. When it goes down, it goes down too much, but usually it doesn't matter because it's used up to the second cap (i.e., it always goes all the way down). But if that were the case, reverse pipetting would always pull too far (and is a Gilson recommended practice). It remains to check.
  • It could be that n=2 is too low and is interpreting the noise.
  • In that case, you should make n=5 and better cover the volume. 5x5=25 sounds reasonable to do it once. But not twice... high fiaca.
  • To do that exp I should maybe set the extra volume that is pipetted instead of using a fixed 5%, or use a fixed 2%...
  • You should do more tests at 20 uL to find out what happens there. And the others too, actually, but since that is the one that has changed the least and the one that I have tried the least, it is worth starting there.
  • Still, people know that pipettors have more error at low volumes... but I would like to avoid systematic pipetting less (so the error is 0 on average).

It is also true that the pipette started a little higher with each pipetting (because 5% more was taken, but that extra did not go down). I don't know if this affects anything.

23-02-17 p-20 calibration results

In this calibration I changed the pipetting sequences a bit.

Conclusions:

  • Tip priming or "wetting" is now separate from the over-draw/back-pour correction.
  • Honestly tip priming did not seem to make much difference. I've yet to try disabling it to evaluate its effect in the final condition.
  • p20 works well with 1x 12 uL draw followed by 5x 2 uL pours.
  • The last pour was tested only once but also did well. End-to-end performance achieved.
  • Overal under-pour with 1.9 uL avg. May be fixed by increasing "scaler_adjust" from 1.08 to perhaps 1.10.
  • Lowering the volume to 9uL/5x1.5uL worked but with higher variability and overa under-pouring (1.4 uL avg).
  • p200 not tested, added dummy parameters for:
  • scaler_adjust
  • under_draw_correction

22-11-30 p-200/p-20 calibration results

Calibration with a sequence in constant increments: go up to the top and go down to the bottom.

https://docs.google.com/spreadsheets/d/1JNC1xFl7ZwNCNkPnPBLw8HRQ9wp5Lkq6KBYjxUF-f_s/edit#gid=0

Relevant observations:

  • The steppers were not operating smoothly.
  • I had to "fix" noise problems on the fly by filtering false limit switch triggers. Even so, some homing sequences looked weird, and paused mid-way without warnings.

Discussion:

  • p200:
  • The p200 tool seems to load less volume than expected:
    • Missing 0.2 uL every 5 uL loaded.
    • Missing 1.5 uL every 20 uL loaded.
  • The p200 tool seems to dispense more accurately, for both 5 uL and 20 uL moves.
  • p20:
  • This pipette is harder to calibrate, because the scale is relatively less accurate for these volumes. Higher errors are expected.
  • The p20 tool seems to load relatively well (10% for 2 uL, 13% for 0.5 uL).
  • The p20 tool dispense badly (37% error for 0.5 uL) or well enough (10% error for 2 uL).
    • In the 37% error case, the dispensed volume was HALF the loaded volume ¿what happened?
    • I repeated the 0.5 uL test with the same results: 9% error to load, 33% error to dispense. The total mass change on dispensing was again close to half of the loaded mass, even though the tip was actually empty at the end. Exactly the same results… ¿is the weigh wrong?
  • Errors are average of absolute value, exluding outliers, see 2022-11-30 Multitool calibration.ods.
  • No consistent evaporation errors are expected, leaving the tube for several minutes in teh scale showed no mass loss.

21-08-17 p-200 balanza robada

Conclusions

  • The s_retract seems to only slightly affect the "take less on first" issue.
  • In order not to lose more volume range than it was already losing (180/200 uL it had approx.), you should compensate by raising the pipette 20 uL after homing.
  • It can take between 1 and 3 uL less in the first (pipetting 20 uL) and usually also takes less in the second (pipetting 10 uL).
  • Change of direction pipetting shows that there is approximately 0.5 uL of backslash.
  • When ejecting volume, it always gave too much, that's consistent with this part of actionInterpreter.py:
            # Add 5% to a volume loading pipette displacement:
            if _action["args"]["volume"] > 0:
                _action["args"]["volume"] = _action["args"]["volume"]*1.05

Therefore, it was superfluous and can be taken out (and I took it out).

Test with 20 uL of "anti homing-backslash"

It didn't work.

Conclusion: the problem is in loading volume on an empty tip, compared to loading volume on a tip that already has liquid. Compensation must be made each time a new tip is used.

It coincides with the fact that I systematically have to prepare more PCR mix than I am going to use, as if the pipette always takes too much, and then drops the correct volume.

For repeated pipetting, a lot more is taken on purpose, and then only the first stop is used. It makes sense then to do the same for repeat/serial pipetting on the robot.

It is surprising to me that this also has to be corrected for "normal" (ie non-serial) pipetting.

The last thing I did, by eye, was to pipet 1uL with the p200 and something crazy happened...

I touched the button 3 times, so I took 3 uL in theory. The scale says to take only ~1.9 uL. That is to be expected given what I observed before taking 10 uL at a time.

The surprising thing is that the first two times I touched the button, it did not load anything, and only on the third touch did the tip of the tip fill with those 1.9 uL.

But it's not like nothing happened in the first two touches. I think they are necessary to "wet" the tip inside, and overcome the surface tension of the water.

That would make sense if water doesn't like to soak plastic so much (i.e. plastic isn't as hydrophilic). So it would make sense to think that the first two touches are necessary to draw enough "vacuum" to overcome the surface tension of the water and into the tip.

Once it expires, and the tip gets a little wet, 1.9 uL of water suddenly rises. But they don't go up 3 uL, why? I'm not sure either.

What I just noticed is that another p200 has the same issue: if I push the plunger up just a little bit, the fluid doesn't come up the tip. But after moving it a little more, it goes up just fine.

The same goes for the p2, it has a bit of "dead" travel where pulling the plunger up doesn't load any volume, and a tiny bit more pushes it up just the right amount. So crazy.

Now I could understand that the tip "gets wet" in another way: what gets wet in a problematic sense is not just the walls of the tip."Wet" in this context perhaps meant to wet the tip of the tip, that is, the pressure necessary to overcome the surface tension of the water (water bending?) and make it enter the tip.

I don't think it closes me completely: in theory if one pipets normally (using a tip once to pass liquid from one tube to another) this happens, and the volume loaded is less than expected.

So the pipette should always charge a little more, to charge correctly. But after lowering the pipette, would the same thing that was loaded be poured out? Or should it be less? for the subject of wetting the walls of the tip.

I mysteriously lost volume in serial pipetting when I did roundtrips, so it could be. In this case, the pipette should always be loaded with more for two reasons: (1) those 2-3 uL necessary for the liquid to enter and (2) because of what the walls of the tip get wet.

However, the p2 charges 0.2 uL without having this problem, and it is with even smaller tips. will it be corrected for that? So I don't know... I would have to try putting a p2 in the robot. For now there are two options: (1) I don't understand anything (2) the pipettes are corrected to contemplate this issue.

Conclusions

The solution is obvious: every time you put on a new tip to pipet water, you have to take 2-3 uL more at the beginning, and then lower what you want to pipet.

That way of pipetting is going to be reproducible.

The trade-off is that more solution is always taken (wasting 3 uL each time). But that was something that already happens to me with the PCR mixes.