No.052 - Final Test & Calibration of my PDVS2mini's - Automated Process

I recently found a way to automate the calibration of my PDVS2mini's post production using a VB app (on Win10), GPIB and a serial/Usb connection.
Here I share in a quick video some of the details & process as well as some testing/production aids I use.
Visit my website for lots more - www.ianjohnston.com
Help me make more of these videos:
Pledge using Patreon - / iansjohnston
Donate using PayPal - www.paypal.me/...

[@TheDefpom]

KZbin never showed me this video in my subscriptions ! I just stumbled across it! Seriously impressive work Ian, my PDVS2MINI is working well, I just used it to calibrate some of my other gear.

[@Leezorc]

Brilliant! Nice work. Well done

[@alpcns]

Very impressive - and always a joy to watch.

[@oriole8789]

The only person on KZbin who owns/uses a 3458A for what it's meant to do, as opposed to so many people who get high precision gear "just because". I've never understood this seemingly common obsession with precision when a person doesn't actually benefit from it (except maybe emotionally?). I'm going to stick with my 4 1/2 digit peasant gear and use the extra funds on things that create value. Nice to see your process - wishing you success with future products!

[@bgdwiepp]

Thank you for showing this, i've been working on a telemetry system and dread the calibration process, this approach looks excellent to use

[@IanScottJohnston]

My method could be done a lot faster by calculating the offsets rather than stepping so much, but it was easy to implement as I needed those routines anyway so just left it at that. It works just fine.

[@TheBreadboardca]

Awesome video, great of you to share your test and cal processes and give a little insider look at the Voltage source, I still use your original version you sent me, I use it with rechargeable PP3's and it is still great and accurate (Best I can tell with a 7.5Digit DMM). Keep up the great work

[@IanScottJohnston]

Cheers Peter....and glad it's still going strong.

[@electronicsunleashed4112]

You are phenomenal......!!👍👍👍👍

[@cs2dsb]

You could maybe speed that search up by doing a binary search. You'd start with something way under and something way above and you'd test the half point. If the half point is too low you take that as the new lower bound and repeat, if it was too high it becomes the new upper bound. Because you are subdividing each time it should converge much quicker than a fixed step scan like you appear to be currently doing. This would be somewhere between your stepping and the PID someone else suggested. But it's likely < 10 lines of code ;)

[@ydonl]

Linear interpolation would be a similar approach to making very good first and subsequent guesses, essentially using the high and low points and slope to make a guess that's more accurate than "half". It may be that much of the code has already been written in one form or another.

[@twobob]

Everyone's already waded in but I like the Over/Under in diminishing steps from a larger seed. Nice system. While since I used VB

[@IanScottJohnston]

The reason i have never re-written the code in such a way is because the 10 to 15 min process of calibration as it is at the moment gives the PDVS2mini a good run at running through cal where I can observe 'live' the results as they happen. The run that I do is split into separate runs with a tighter tolerance and higher NPLC under which I can make sure all is ok. Call it an extended test across the devices range and also check for a good INL spec. The other option would be a 5 minute test where I would have to then take 5 or 10mins to manually verify specific or random voltages to be happy......or indeed write another routine to do that for me. Net result, I am happy with the current procedure. PS. It also gives the comms a good test.

[@twobob]

@@IanScottJohnston To be honest Ian in my head I was just effectively approximating Pi... My bad. Soak test FTW sir of course. Just rewatched SDG#099 and pretty sure I caught a later one about constant current, they both dripped respect with the kind of treatment he would proffer to Keysight gear not some kit off Banggood. You have /my/ utmost respect as a maker who has brought to market, well, anything of note. Note, not the hugest of clubs. Cap doffed from a Scotsborn lowlander. (who played guitar most of his life too)

[@IanScottJohnston]

@@twobob go to my webs!te and look under PROJECTS for the ZOOM FIRE-30 guitar amp mods.......classic example of when electronics, guitar and too much time on my hands meet......LOL!

[@twobob]

@@IanScottJohnston Eyeballs Marshall Code 100 menacingly...Oh baby. You were my kind of guitar player (glances at several partially finished data glove designs...) yup

[@405line]

Found that very interesting. Would starting off with the rolling average data from all the previous calibrations be of any use or do you already do that?

[@cellularmitosis2]

Hi Ian, can I ask how many of these you have sold? I’m curious what the market is for precision boutique electronics

[@brendanbarbour8568]

Great job Ian. The efficiency gain is very significant clearly. You mentioned a final "manual tweek" of the calibration after you've run the Auto-calibration routine; what's that about ?

[@IanScottJohnston]

For the automated process the 3458A is set up with a very small NPLC because its fast, and so the manual tweak afterwards I would increase this but which would also increase the sample rate to a couple of seconds or more. A high NPLC value increases AC noise integration time and increases measurement resolution and accuracy. Also, by doing the manual tweak without the serial comms connected I also minimize any ground loop or noise issues.

[@TheDefpom]

At 15:30 couldn’t you do a calculation ratio for the difference for the step correction so if it is out by a lot it uses larger steps, so the step size is a scaling of the difference

[@IanScottJohnston]

Yes, this is just the first working version.......and I used it to calibrate a whole batch, it's great to see it stepping to the target value and helps me identify other possible issues such as too big a step requires a slightly longer wait time on the 3458A as it takes longer for the voltage to travel further......but the plan is to modify it for the next batch.

[@largepimping]

Ian, how many PDVS2s would I have to buy to convince you to say "Welcome back to *the* workshop" at the start of the video? It's hurting my American ears. Hopefully it's obvious that I'm kidding - your videos are great and I love your channel.

[@riccardoiacob4560]

You could always tell the program to start with big bit counts (i.e. 1000) until it overshoots, then step back and divide by two (i.e. 500) and proceed like that to have a faster cal

[@IanScottJohnston]

Part of the reason for doing it this way was to give the units a good run through the steps.......just to make sure all is ok.

[@UltraOffie]

Thank You Ian. Awesome automated process... product and design. I saw this video after I Ordered my PDVS2mini. I am a new fan and subscriber. I hope you keep finding time to share you work via video. Best Wishes Stay Well.

[@IanScottJohnston]

Thanks!

[@JensPeterSecher]

Wouldn’t it make sense to use a PID algorithm instead of simply stepping one bit at a time?

[@IanScottJohnston]

Stepping method implemented in minutes, PID or other a good bit longer........I was keen to get it up & running.

[@DrFrank-xj9bc]

nice test and calibration setup. The first two steps for zero and 1V take too long, especially you could speed up 1V point to a few steps only, by calculating the gain of the D/A, and then approach this voltage to within a few bits, then iterating up and down.

[@IanScottJohnston]

Yes, that is what I have been looking to do.....this first incarnation as it sits just to prove my concept. The zero counts are always around the same value so not an issue, but certainly 1V pre-run could do with a re-write........I'll get onto it and report back....:-)

[@ivolol]

Seems like something you want Newton's method for.