Hello!
I've recently been interested in measuring the gain non-linearity of various op-amps and composite op-amp configurations for use in integrating ADCs.
Following the circuits and measurement techniques posted in these app notes from ADI and TI, the results I got were unusable:
https://www.analog.com/media/en/training-seminars/tutorials/MT-044.pdf
https://www.ti.com/lit/an/snaa047a/snaa047a.pdf?ts=1687173468806
The problem I found with these circuits is that they are basically noise amplifiers with a high-gain network created by the top feedback resistor and the divider on the inverting input. Instead of a clear line, I get a flood of noise that comes from the resistors and op-amp (the noise waveforms are very characteristic of band-limited noise). I don't understand how the clean lines in the app notes were obtained. I tried both a small capacitor from input to the summing junction as recommended in the TI app note, as well as a film capacitor (10s of nF) across the top resistor on the divider on the inverting input. The latter fix helped remove the noise, but the XY plot was flat even for the incredibly poor LM358. I had a couple other people repeat these measurements for me, and they had similar results.
Is there a sane way to measure gain non-linearity that actually works?
Regards,
NNNI

On Mon, 19 Jun 2023 14:52:53 -0700 (PDT), NNNI
<vivekanandottalingam@gmail.com> wrote:

>Hello!
>I've recently been interested in measuring the gain non-linearity of various op-amps and composite op-amp configurations for use in integrating ADCs.
>Following the circuits and measurement techniques posted in these app notes from ADI and TI, the results I got were unusable:
>https://www.analog.com/media/en/training-seminars/tutorials/MT-044.pdf
>https://www.ti.com/lit/an/snaa047a/snaa047a.pdf?ts=1687173468806
>The problem I found with these circuits is that they are basically noise amplifiers with a high-gain network created by the top feedback resistor and the divider on the inverting input. Instead of a clear line, I get a flood of noise that comes from the resistors and op-amp (the noise waveforms are very characteristic of band-limited noise). I don't understand how the clean lines in the app notes were obtained. I tried both a small capacitor from input to the summing junction as recommended in the TI app note, as well as a film capacitor (10s of nF) across the top resistor on the divider on the inverting input. The latter fix helped remove the noise, but the XY plot was flat even for the incredibly poor LM358. I had a couple other people repeat these measurements for me, and they had similar results.
>Is there a sane way to measure gain non-linearity that actually works?
>Regards,
>NNNI

Just signal average a bunch.

On Tuesday, 20 June 2023 at 00:56:58 UTC+1, John Larkin wrote:
> On Mon, 19 Jun 2023 14:52:53 -0700 (PDT), NNNI
> <vivekanand...@gmail.com> wrote:
>
> >Hello!
> >I've recently been interested in measuring the gain non-linearity of various op-amps and composite op-amp configurations for use in integrating ADCs.
> >Following the circuits and measurement techniques posted in these app notes from ADI and TI, the results I got were unusable:
> >https://www.analog.com/media/en/training-seminars/tutorials/MT-044.pdf
> >https://www.ti.com/lit/an/snaa047a/snaa047a.pdf?ts=1687173468806
> >The problem I found with these circuits is that they are basically noise amplifiers with a high-gain network created by the top feedback resistor and the divider on the inverting input. Instead of a clear line, I get a flood of noise that comes from the resistors and op-amp (the noise waveforms are very characteristic of band-limited noise). I don't understand how the clean lines in the app notes were obtained. I tried both a small capacitor from input to the summing junction as recommended in the TI app note, as well as a film capacitor (10s of nF) across the top resistor on the divider on the inverting input. The latter fix helped remove the noise, but the XY plot was flat even for the incredibly poor LM358. I had a couple other people repeat these measurements for me, and they had similar results.
> >Is there a sane way to measure gain non-linearity that actually works?
> >Regards,
> >NNNI
> Just signal average a bunch.

....and low-pass filter the scope inputs. If a sweep frequency of around 1Hz is used, then
maybe filter at about 1kHz. Both app notes are silent on this.

John

On a sunny day (Tue, 20 Jun 2023 00:31:54 -0700 (PDT)) it happened John
Walliker <jrwalliker@gmail.com> wrote in
<35a85f6e-8c39-4a89-99f8-8c2f8660b7ban@googlegroups.com>:

>On Tuesday, 20 June 2023 at 00:56:58 UTC+1, John Larkin wrote:
>> On Mon, 19 Jun 2023 14:52:53 -0700 (PDT), NNNI
>> <vivekanand...@gmail.com> wrote:
>>
>> >Hello!
>> >I've recently been interested in measuring the gain non-linearity of var=
>ious op-amps and composite op-amp configurations for use in integrating ADC=
>s.
>> >Following the circuits and measurement techniques posted in these app no=
>tes from ADI and TI, the results I got were unusable:
>> >https://www.analog.com/media/en/training-seminars/tutorials/MT-044.pdf=
>
>> >https://www.ti.com/lit/an/snaa047a/snaa047a.pdf?ts=1687173468806
>> >The problem I found with these circuits is that they are basically noise=
> amplifiers with a high-gain network created by the top feedback resistor a=
>nd the divider on the inverting input. Instead of a clear line, I get a flo=
>od of noise that comes from the resistors and op-amp (the noise waveforms a=
>re very characteristic of band-limited noise). I don't understand how the c=
>lean lines in the app notes were obtained. I tried both a small capacitor f=
>rom input to the summing junction as recommended in the TI app note, as wel=
>l as a film capacitor (10s of nF) across the top resistor on the divider on=
> the inverting input. The latter fix helped remove the noise, but the XY pl=
>ot was flat even for the incredibly poor LM358. I had a couple other people=
> repeat these measurements for me, and they had similar results.
>> >Is there a sane way to measure gain non-linearity that actually works?=
>
>> >Regards,
>> >NNNI
>> Just signal average a bunch.
>
>...and low-pass filter the scope inputs. If a sweep frequency of around 1H=
>z is used, then
>maybe filter at about 1kHz. Both app notes are silent on this.
>
>John

It somehow made me think on how we measured distortion (non linearity) in
large audio networks (like Eurovison network).

One side would send 1 kHz, and we used a tuned meter to listen to the
fundamental 1 kHz, then tuned to 2 kHz, then to 3 kHz etc up to several kHz.
Simple sum of all harmonics gave the distortion that had to be below some value.
Done it several times back in the late sixties and early seventies.

Low frequency spectrum analyzer must exist as PC app for soundcard?

On Tue, 20 Jun 2023 00:31:54 -0700 (PDT), John Walliker
<jrwalliker@gmail.com> wrote:

>On Tuesday, 20 June 2023 at 00:56:58 UTC+1, John Larkin wrote:
>> On Mon, 19 Jun 2023 14:52:53 -0700 (PDT), NNNI
>> <vivekanand...@gmail.com> wrote:
>>
>> >Hello!
>> >I've recently been interested in measuring the gain non-linearity of various op-amps and composite op-amp configurations for use in integrating ADCs.
>> >Following the circuits and measurement techniques posted in these app notes from ADI and TI, the results I got were unusable:
>> >https://www.analog.com/media/en/training-seminars/tutorials/MT-044.pdf
>> >https://www.ti.com/lit/an/snaa047a/snaa047a.pdf?ts=1687173468806
>> >The problem I found with these circuits is that they are basically noise amplifiers with a high-gain network created by the top feedback resistor and the divider on the inverting input. Instead of a clear line, I get a flood of noise that comes from the resistors and op-amp (the noise waveforms are very characteristic of band-limited noise). I don't understand how the clean lines in the app notes were obtained. I tried both a small capacitor from input to the summing junction as recommended in the TI app note, as well as a film capacitor (10s of nF) across the top resistor on the divider on the inverting input. The latter fix helped remove the noise, but the XY plot was flat even for the incredibly poor LM358. I had a couple other people repeat these measurements for me, and they had similar results.
>> >Is there a sane way to measure gain non-linearity that actually works?
>> >Regards,
>> >NNNI
>> Just signal average a bunch.
>
>...and low-pass filter the scope inputs. If a sweep frequency of around 1Hz is used, then
>maybe filter at about 1kHz. Both app notes are silent on this.
>
>John

Most scopes do bw limiting and signal averaging these days. One can
always do some creative graphing too.

I've tried both averaging and bandwidth limiting, but that didn't help much. The only scope I have on hand right now is a HP54600A. In two weeks I should get a Rigol DS1054Z, I'll try the measurements again.
Thanks for the tip about low pass filtering the scope input, I'll try that as well.
Regards,
NNNI

On Tue, 20 Jun 2023 04:23:24 -0700 (PDT), NNNI
<vivekanandottalingam@gmail.com> wrote:

>I've tried both averaging and bandwidth limiting, but that didn't help much. The only scope I have on hand right now is a HP54600A. In two weeks I should get a Rigol DS1054Z, I'll try the measurements again.
>Thanks for the tip about low pass filtering the scope input, I'll try that as well.
>Regards,
>NNNI

Are you picking up ambient noise? An Altoids or a Danish Butter Cookie
can would be a good shield, with some lead bypasses.

I hope you are not using one of those dreadful plastic breadboard
things.

https://www.dropbox.com/s/30c6aesh4tzjg56/Z496_Can.jpg?dl=0

I wouldn't dream of using a solderless breadboard for anything!
https://cdn.discordapp.com/attachments/910915289611378709/1120716234200203274/20230609_233035.jpg
I've been using Manhattan-style as of late. Regarding the noise, I'm not 100% sure if shielding will help, because at such high gains, you might as well be amplifying the noise of the resistors and op-amp.

On Tue, 20 Jun 2023 07:08:31 -0700 (PDT), NNNI
<vivekanandottalingam@gmail.com> wrote:

>I wouldn't dream of using a solderless breadboard for anything!
>https://cdn.discordapp.com/attachments/910915289611378709/1120716234200203274/20230609_233035.jpg
>I've been using Manhattan-style as of late. Regarding the noise, I'm not 100% sure if shielding will help, because at such high gains, you might as well be amplifying the noise of the resistors and op-amp.

Wave your hands around it and see if the noise changes. Look for 50/60
Hz too maybe. FFT it!

Nice style, similar to mine. I use gold-plated FR4, which looks good
and doesn't tarnish over time.

https://www.dropbox.com/s/opnxfnk79o5lk1s/Z466_2.JPG?dl=0

Here's one I just did, to blow up some Panasonic solid-state relays.

https://www.dropbox.com/sh/y5g5r5ul08nvwum/AABJPSW6rEFHYB3iCnmONM3ia?dl=0

SSR data sheets don't have SOAR graphs.

Looks like I can add a roughly 10 ohm wirewound resistor and have the
SSR survive shorting 35 volts for long enough for software to shut
things down. Now I need to blow up some surface-mount wirewounds.

On a sunny day (Tue, 20 Jun 2023 07:52:16 -0700) it happened John Larkin
<jlarkin@highlandSNIPMEtechnology.com> wrote in
<80d39i5f3avp9pmf0vk36ea8nbktok5olr@4ax.com>:

>On Tue, 20 Jun 2023 07:08:31 -0700 (PDT), NNNI
><vivekanandottalingam@gmail.com> wrote:
>
>>I wouldn't dream of using a solderless breadboard for anything!
>>https://cdn.discordapp.com/attachments/910915289611378709/1120716234200203274/20230609_233035.jpg
>>I've been using Manhattan-style as of late. Regarding the noise, I'm not 100% sure if shielding will help, because at such high
>>gains, you might as well be amplifying the noise of the resistors and op-amp.

Those probe ground coils have inductance and a few pF trimmer makes me wonder about RF.

>Wave your hands around it and see if the noise changes. Look for 50/60
>Hz too maybe. FFT it!

Yep

I did poke around the test setup quite a bit but I was not able to discern changes. I will try again with a shield and better oscilloscope.

On 2023-06-20, Jan Panteltje <alien@comet.invalid> wrote:
> On a sunny day (Tue, 20 Jun 2023 00:31:54 -0700 (PDT)) it happened John
> Walliker <jrwalliker@gmail.com> wrote in
><35a85f6e-8c39-4a89-99f8-8c2f8660b7ban@googlegroups.com>:
>
>>On Tuesday, 20 June 2023 at 00:56:58 UTC+1, John Larkin wrote:
>>> On Mon, 19 Jun 2023 14:52:53 -0700 (PDT), NNNI
>>> <vivekanand...@gmail.com> wrote:
>>>
>>> >Hello!
>>> >I've recently been interested in measuring the gain non-linearity of var=
>>ious op-amps and composite op-amp configurations for use in integrating ADC=
>>s.
>>> >Following the circuits and measurement techniques posted in these app no=
>>tes from ADI and TI, the results I got were unusable:
>>> >https://www.analog.com/media/en/training-seminars/tutorials/MT-044.pdf=
>>
>>> >https://www.ti.com/lit/an/snaa047a/snaa047a.pdf?ts=1687173468806
>>> >The problem I found with these circuits is that they are basically noise=
>> amplifiers with a high-gain network created by the top feedback resistor a=
>>nd the divider on the inverting input. Instead of a clear line, I get a flo=
>>od of noise that comes from the resistors and op-amp (the noise waveforms a=
>>re very characteristic of band-limited noise). I don't understand how the c=
>>lean lines in the app notes were obtained. I tried both a small capacitor f=
>>rom input to the summing junction as recommended in the TI app note, as wel=
>>l as a film capacitor (10s of nF) across the top resistor on the divider on=
>> the inverting input. The latter fix helped remove the noise, but the XY pl=
>>ot was flat even for the incredibly poor LM358. I had a couple other people=
>> repeat these measurements for me, and they had similar results.
>>> >Is there a sane way to measure gain non-linearity that actually works?=
>>
>>> >Regards,
>>> >NNNI
>>> Just signal average a bunch.
>>
>>...and low-pass filter the scope inputs. If a sweep frequency of around 1H=
>>z is used, then
>>maybe filter at about 1kHz. Both app notes are silent on this.
>>
>>John
>
> It somehow made me think on how we measured distortion (non linearity) in
> large audio networks (like Eurovison network).
>
> One side would send 1 kHz, and we used a tuned meter to listen to the
> fundamental 1 kHz, then tuned to 2 kHz, then to 3 kHz etc up to several kHz.
> Simple sum of all harmonics gave the distortion that had to be below some value.
> Done it several times back in the late sixties and early seventies.
>
> Low frequency spectrum analyzer must exist as PC app for soundcard?

you can look at captures using audacity.

also: sigrok has a pull request pending,

--
Jasen.
🇺🇦 Слава Україні

On Monday, June 19, 2023 at 5:52:58 PM UTC-4, NNNI wrote:
> Hello!
> I've recently been interested in measuring the gain non-linearity of various op-amps and composite op-amp configurations for use in integrating ADCs..
> Following the circuits and measurement techniques posted in these app notes from ADI and TI, the results I got were unusable:
> https://www.analog.com/media/en/training-seminars/tutorials/MT-044.pdf
> https://www.ti.com/lit/an/snaa047a/snaa047a.pdf?ts=1687173468806
> The problem I found with these circuits is that they are basically noise amplifiers with a high-gain network created by the top feedback resistor and the divider on the inverting input. Instead of a clear line, I get a flood of noise that comes from the resistors and op-amp (the noise waveforms are very characteristic of band-limited noise). I don't understand how the clean lines in the app notes were obtained. I tried both a small capacitor from input to the summing junction as recommended in the TI app note, as well as a film capacitor (10s of nF) across the top resistor on the divider on the inverting input. The latter fix helped remove the noise, but the XY plot was flat even for the incredibly poor LM358. I had a couple other people repeat these measurements for me, and they had similar results.
> Is there a sane way to measure gain non-linearity that actually works?

Yeah, you need to put the DUT in a Faraday cage, and additionally protect it from air currents and sound. I made a pretty good motion detector using a similar circuit, sound waves caused it to produce enormous peaks.

> Regards,
> NNNI

On Wed, 21 Jun 2023 04:28:08 -0700 (PDT), Fred Bloggs
<bloggs.fredbloggs.fred@gmail.com> wrote:

>On Monday, June 19, 2023 at 5:52:58?PM UTC-4, NNNI wrote:
>> Hello!
>> I've recently been interested in measuring the gain non-linearity of various op-amps and composite op-amp configurations for use in integrating ADCs.
>> Following the circuits and measurement techniques posted in these app notes from ADI and TI, the results I got were unusable:
>> https://www.analog.com/media/en/training-seminars/tutorials/MT-044.pdf
>> https://www.ti.com/lit/an/snaa047a/snaa047a.pdf?ts=1687173468806
>> The problem I found with these circuits is that they are basically noise amplifiers with a high-gain network created by the top feedback resistor and the divider on the inverting input. Instead of a clear line, I get a flood of noise that comes from the resistors and op-amp (the noise waveforms are very characteristic of band-limited noise). I don't understand how the clean lines in the app notes were obtained. I tried both a small capacitor from input to the summing junction as recommended in the TI app note, as well as a film capacitor (10s of nF) across the top resistor on the divider on the inverting input. The latter fix helped remove the noise, but the XY plot was flat even for the incredibly poor LM358. I had a couple other people repeat these measurements for me, and they had similar results.
>> Is there a sane way to measure gain non-linearity that actually works?
>
>Yeah, you need to put the DUT in a Faraday cage, and additionally protect it from air currents and sound. I made a pretty good motion detector using a similar circuit, sound waves caused it to produce enormous peaks.
>
>> Regards,
>> NNNI

Air convection is a classic source of super-low-frequency noise, 1/f
sort of thing. Here's a little cover over a thermocouple front-end,
which really helps.

https://www.dropbox.com/scl/fi/guzbm416437bbi3wyo1ox/L350_TC_Cover.JPG?dl=0&rlkey=jag073z02vioaxk32mg804ff9

On a sunny day (Wed, 21 Jun 2023 07:53:34 -0700) it happened John Larkin
<jlarkin@highlandSNIPMEtechnology.com> wrote in
<me369ipjpkdj87k61d74lbvvfo7128sq4g@4ax.com>:
>Air convection is a classic source of super-low-frequency noise, 1/f
>sort of thing. Here's a little cover over a thermocouple front-end,
>which really helps.
>
>https://www.dropbox.com/scl/fi/guzbm416437bbi3wyo1ox/L350_TC_Cover.JPG?dl=0&rlkey=jag073z02vioaxk32mg804ff9

https://panteltje.nl/panteltje/tri_pic/
scroll down for temperature controlled box

Anyways, it seems tritium decay _does_ depend on seasons...
hehe

On a sunny day (Wed, 21 Jun 2023 06:49:13 -0000 (UTC)) it happened Jasen Betts wrote:

>On 2023-06-20, Jan Panteltje <alien@comet.invalid> wrote:
>> Low frequency spectrum analyzer must exist as PC app for soundcard?
>
>you can look at captures using audacity.
>
>also: sigrok has a pull request pending,

Yes, I wrote some of my own but not sure anybody can still compile those, using an old version of xforms for the GUI,
https://panteltje.nl/panteltje/xpequ/index.html
that said: I have it running on my Pi4 :-)
The spectrum display is simple and not calibrated,
but you could muddify eeeeh modify it with no problem to almost any precision,
and use libfftw3
https://www.fftw.org/
I use libfftw in this:
https://panteltje.nl/panteltje/xpsa/index.html

there is more stuff...
writing your own code gives some more insight and in a way more freedom.
Same with designing your own 'tronix

On Thu, 22 Jun 2023 04:35:21 GMT, Jan Panteltje <alien@comet.invalid>
wrote:

>On a sunny day (Wed, 21 Jun 2023 07:53:34 -0700) it happened John Larkin
><jlarkin@highlandSNIPMEtechnology.com> wrote in
><me369ipjpkdj87k61d74lbvvfo7128sq4g@4ax.com>:
>>Air convection is a classic source of super-low-frequency noise, 1/f
>>sort of thing. Here's a little cover over a thermocouple front-end,
>>which really helps.
>>
>>https://www.dropbox.com/scl/fi/guzbm416437bbi3wyo1ox/L350_TC_Cover.JPG?dl=0&rlkey=jag073z02vioaxk32mg804ff9
>
> https://panteltje.nl/panteltje/tri_pic/
> scroll down for temperature controlled box
>
>Anyways, it seems tritium decay _does_ depend on seasons...
>hehe

I used to sell NMR gradient-coil drivers, which are super-stable
pulsed constant-current sources. I buried home-made manganin current
shunts in a heated aluminum block, bonded a PCB to that, and ran it at
constant temperature. Our first unit was 70 times more stable than the
Crown amp the customer had been using. We put a drawn-aluminum cover
over the heated block and decided, after experiments, to not have any
foam or other stuff inside. Still air was the best insulator.

https://www.dropbox.com/sh/8s2qbyh5emhvsda/AABG72Yx5sWCKNEvysMU_4Lba?dl=0

One problem was to keep eddy currents around the shunt from trashing
pulse flatness. And to use compound amplifiers to get around opamp
open-loop nonlinearity and thermal hooks and other effects of
amplifying the shunt voltage in one step. Another problem was how to
measure 20 or 120 amp pulse flatness to parts-per-million.

How much does tritium decay depend on season?

On 2023-06-21 10:53, John Larkin wrote:
> On Wed, 21 Jun 2023 04:28:08 -0700 (PDT), Fred Bloggs
> <bloggs.fredbloggs.fred@gmail.com> wrote:
>
>> On Monday, June 19, 2023 at 5:52:58?PM UTC-4, NNNI wrote:
>>> Hello!
>>> I've recently been interested in measuring the gain non-linearity of various op-amps and composite op-amp configurations for use in integrating ADCs.
>>> Following the circuits and measurement techniques posted in these app notes from ADI and TI, the results I got were unusable:
>>> https://www.analog.com/media/en/training-seminars/tutorials/MT-044.pdf
>>> https://www.ti.com/lit/an/snaa047a/snaa047a.pdf?ts=1687173468806
>>> The problem I found with these circuits is that they are basically noise amplifiers with a high-gain network created by the top feedback resistor and the divider on the inverting input. Instead of a clear line, I get a flood of noise that comes from the resistors and op-amp (the noise waveforms are very characteristic of band-limited noise). I don't understand how the clean lines in the app notes were obtained. I tried both a small capacitor from input to the summing junction as recommended in the TI app note, as well as a film capacitor (10s of nF) across the top resistor on the divider on the inverting input. The latter fix helped remove the noise, but the XY plot was flat even for the incredibly poor LM358. I had a couple other people repeat these measurements for me, and they had similar results.
>>> Is there a sane way to measure gain non-linearity that actually works?
>>
>> Yeah, you need to put the DUT in a Faraday cage, and additionally protect it from air currents and sound. I made a pretty good motion detector using a similar circuit, sound waves caused it to produce enormous peaks.
>>
>>> Regards,
>>> NNNI
>
> Air convection is a classic source of super-low-frequency noise, 1/f
> sort of thing. Here's a little cover over a thermocouple front-end,
> which really helps.
>
> https://www.dropbox.com/scl/fi/guzbm416437bbi3wyo1ox/L350_TC_Cover.JPG?dl=0&rlkey=jag073z02vioaxk32mg804ff9
>

IIRC there's an old PMI app note about the OP27, which they tout as
having vastly superior open-loop gain linearity compared with the OP07.

I remember it being full of plots and test circuits and stuff. I'll see
if I can find it.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

On a sunny day (Thu, 22 Jun 2023 06:36:02 -0700) it happened John Larkin
<jlarkin@highlandSNIPMEtechnology.com> wrote in
<7ri89ihdr84g40id7dkbo3vsk136rrqa2e@4ax.com>:

>On Thu, 22 Jun 2023 04:35:21 GMT, Jan Panteltje <alien@comet.invalid>
>wrote:
>
>>On a sunny day (Wed, 21 Jun 2023 07:53:34 -0700) it happened John Larkin
>><jlarkin@highlandSNIPMEtechnology.com> wrote in
>><me369ipjpkdj87k61d74lbvvfo7128sq4g@4ax.com>:
>>>Air convection is a classic source of super-low-frequency noise, 1/f
>>>sort of thing. Here's a little cover over a thermocouple front-end,
>>>which really helps.
>>>
>>>https://www.dropbox.com/scl/fi/guzbm416437bbi3wyo1ox/L350_TC_Cover.JPG?dl=0&rlkey=jag073z02vioaxk32mg804ff9
>>
>> https://panteltje.nl/panteltje/tri_pic/
>> scroll down for temperature controlled box
>>
>>Anyways, it seems tritium decay _does_ depend on seasons...
>>hehe
>
>I used to sell NMR gradient-coil drivers, which are super-stable
>pulsed constant-current sources. I buried home-made manganin current
>shunts in a heated aluminum block, bonded a PCB to that, and ran it at
>constant temperature. Our first unit was 70 times more stable than the
>Crown amp the customer had been using. We put a drawn-aluminum cover
>over the heated block and decided, after experiments, to not have any
>foam or other stuff inside. Still air was the best insulator.
>
>https://www.dropbox.com/sh/8s2qbyh5emhvsda/AABG72Yx5sWCKNEvysMU_4Lba?dl=0
>
>One problem was to keep eddy currents around the shunt from trashing
>pulse flatness. And to use compound amplifiers to get around opamp
>open-loop nonlinearity and thermal hooks and other effects of
>amplifying the shunt voltage in one step. Another problem was how to
>measure 20 or 120 amp pulse flatness to parts-per-million.

Foam does damp mechanical vibrations a bit, it also thermally isolates the circuit that is artificially kept at a few degrees above
maximum expected room temperature.

>How much does tritium decay depend on season?

Mike Playle did an analysis of the data:
https://panteltje.nl/panteltje/tri_pic/tritium_decay_experiment_evaluation_2012_to_2020.html
but concludes, quote:
"
In conclusion, while there does appear to be an annual component to these measurements,
there are reasons to suspect that it arises from the behavior of the measurement apparatus,
and so this analysis neither confirms nor refutes the hypothesis that the decay rate of tritium varies annually.
"

I did the experiment after some researcher found annual variations,
I decided to use a totally different way to measure the decay and I also see similar variations.

For what it is worth,
The logs of the temperature sensors and data are on my site so it likely is not temperature related.
Personally I do not believe in coincidences....
It is quite possible that high energy particles from space cause more ticks / light flashes in the tritium tubes in different seasons,
more decay perhaps.
It is interesting my gamma spectrometer and Geiger counter sometimes react at the same moment ...
shower of particles? or just one that went through both in turn?

I am not biased one way or the other but they still deny life of Mars even though Dr Levin's experiment was positive..
http://www.gillevin.com/

I discovered some other weird thing this morning, when my neighbor starts his old Mercedes my Geiger counter count goes
up by a few percent when exposed to the exhaust.
Wonder what he uses as fuel ;-)
Gotta measure some....

2023 06 22 06:39 9 <- normal background
2023 06 22 06:40 7
2023 06 22 06:41 8
2023 06 22 06:42 9
2023 06 22 06:43 7
2023 06 22 06:44 6
2023 06 22 06:45 6
2023 06 22 06:46 7
2023 06 22 06:47 5
2023 06 22 06:48 14 <- starts engine
2023 06 22 06:49 10
2023 06 22 06:50 11
2023 06 22 06:51 6
2023 06 22 06:52 12 <- drove away
2023 06 22 06:53 6
2023 06 22 06:54 7
2023 06 22 06:55 10 exhaust smell everywhere
2023 06 22 06:56 10
2023 06 22 06:57 10
2023 06 22 06:58 11
2023 06 22 06:59 12
2023 06 22 07:00 5 <- smell gone rest back to normal

On Thu, 22 Jun 2023 11:03:38 -0400, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

>On 2023-06-21 10:53, John Larkin wrote:
>> On Wed, 21 Jun 2023 04:28:08 -0700 (PDT), Fred Bloggs
>> <bloggs.fredbloggs.fred@gmail.com> wrote:
>>
>>> On Monday, June 19, 2023 at 5:52:58?PM UTC-4, NNNI wrote:
>>>> Hello!
>>>> I've recently been interested in measuring the gain non-linearity of various op-amps and composite op-amp configurations for use in integrating ADCs.
>>>> Following the circuits and measurement techniques posted in these app notes from ADI and TI, the results I got were unusable:
>>>> https://www.analog.com/media/en/training-seminars/tutorials/MT-044.pdf
>>>> https://www.ti.com/lit/an/snaa047a/snaa047a.pdf?ts=1687173468806
>>>> The problem I found with these circuits is that they are basically noise amplifiers with a high-gain network created by the top feedback resistor and the divider on the inverting input. Instead of a clear line, I get a flood of noise that comes from the resistors and op-amp (the noise waveforms are very characteristic of band-limited noise). I don't understand how the clean lines in the app notes were obtained. I tried both a small capacitor from input to the summing junction as recommended in the TI app note, as well as a film capacitor (10s of nF) across the top resistor on the divider on the inverting input. The latter fix helped remove the noise, but the XY plot was flat even for the incredibly poor LM358. I had a couple other people repeat these measurements for me, and they had similar results.
>>>> Is there a sane way to measure gain non-linearity that actually works?
>>>
>>> Yeah, you need to put the DUT in a Faraday cage, and additionally protect it from air currents and sound. I made a pretty good motion detector using a similar circuit, sound waves caused it to produce enormous peaks.
>>>
>>>> Regards,
>>>> NNNI
>>
>> Air convection is a classic source of super-low-frequency noise, 1/f
>> sort of thing. Here's a little cover over a thermocouple front-end,
>> which really helps.
>>
>> https://www.dropbox.com/scl/fi/guzbm416437bbi3wyo1ox/L350_TC_Cover.JPG?dl=0&rlkey=jag073z02vioaxk32mg804ff9
>>
>
>IIRC there's an old PMI app note about the OP27, which they tout as
>having vastly superior open-loop gain linearity compared with the OP07.
>
>I remember it being full of plots and test circuits and stuff. I'll see
>if I can find it.
>
>Cheers
>
>Phil Hobbs

To amplify a current shunt, I used an opamp feedback network of about
10 ohms Thevenin, to keep the Johnson and current-driven noise down.
That makes the opamp work hard making much swing, which creates
thermal hooks. So, compound amp.

I used LT1028.