EEVblog Electronics Community Forum

Electronics => Metrology => Topic started by: Crossphased on October 14, 2018, 07:30:08 am

Title: Couple Questions about Reference Divider
Post by: Crossphased on October 14, 2018, 07:30:08 am
Hello,

I've got (5) AD587KRZ burning in, they've been going for around 500 hours right now. After they've burned in for 1000 hours I'll be making a few 10V standards out of the better performing ones.

At the moment I'm putting together a circuit to divide the 10V down to 5V, for use as voltage reference for ADCs. The LTC1043 is being used. I put together a sample circuit that I'd welcome advice on:

(https://farm2.staticflickr.com/1962/31432511288_d175f6cac9_b.jpg) (https://flic.kr/p/PTzQvN)

The LTC2057 was selected for low bias current to buffer the output from LTC1043. I wasnt sure if the RC filters were necessary or not. Any suggestions or reccommendations would be appreciated.

Also I have a few questions regarding the implementation:
- would it be a good idea to trim out the offset from LT1001?
- the sampling capacitor C7 was partially wrapped in copper foil for guarding. I'm trying to figure out the best pin on LTC1043 to connect the guard to. As far as I can tell it would be pin 10 is that right?
- only 1/2 of the LTC1043 is being used for the divide by two... would it be a good idea to connect the other half in a divide by 2 configuration as well and connect in parallel to lower output impedance?

Thanks for any advice, cheers

*this was my first schematic in kicad if theres anything you would have done differently that is welcomed as well!
Title: Re: Couple Questions about Reference Divider
Post by: Kleinstein on October 14, 2018, 08:02:49 am
The buffer circuits(both at the input and output) are wrong this way. As shown the OPs will be on the edge to oscillation. The capacitive load driving circuit is usually a little different: another capacitor as direct feedback and an additional resistor for the DC feedback. In addition the  resistor at the OP't output is usually smaller, more like 22-150 Ohms.

For the driver at the output, it really depends on the ADC, if the buffer is suitable this way.

The LT1001 has a low offset to start with. So there is not much gained (lower TC) from an extra adjustment - it is likely adding more trouble.

C2,C5,C10 are a little odd. Usually it only needs 1 local decoupling cap (e.g. the 100 nF) the larger cap stage can usually be shared.
If paranoid, one could consider an extra filter (ferrite, resistor) in the supply to the LTC2057 and than a larger local capacitor.
Title: Re: Couple Questions about Reference Divider
Post by: Andreas on October 14, 2018, 10:11:58 am

- would it be a good idea to trim out the offset from LT1001?
- the sampling capacitor C7 was partially wrapped in copper foil for guarding. I'm trying to figure out the best pin on LTC1043 to connect the guard to. As far as I can tell it would be pin 10 is that right?
- only 1/2 of the LTC1043 is being used for the divide by two... would it be a good idea to connect the other half in a divide by 2 configuration as well and connect in parallel to lower output impedance?


a) no you will use the trimpot at the AD587 to trim overall value.  (and have no chance to adjust to zero without negative power supply).
b) yes connect to pin 10
c) in my opinion no: you also increase leakage currents, charge injection  etc. And you have a buffer at the output anyway.

to avoid oscillations either connect pin 6 directly to pin 2 of the OP amps or use the additional components (as suggested from Kleinstein).

with best regards

Andreas
Title: Re: Couple Questions about Reference Divider
Post by: Crossphased on October 14, 2018, 03:15:23 pm
Thank you both for your kind assistance

Is this kind of what you had in mind?
(https://farm2.staticflickr.com/1934/31440651328_8e97fab04e_b.jpg) (https://flic.kr/p/PUiygj)

Or would it be preferred to have resistor and capacitor feedback on the OP?
(https://farm2.staticflickr.com/1905/44402391265_46f08f732a_b.jpg) (https://flic.kr/p/2aDFRY6)

Title: Re: Couple Questions about Reference Divider
Post by: Andreas on October 14, 2018, 06:32:08 pm
Is this kind of what you had in mind?


Or would it be preferred to have resistor and capacitor feedback on the OP?
a) yes

b) this one is wrong:
    R? has to be connected to C3
   Time constant of  R? * C? has to be larger than R1 * C3

Solution b) also compensates load currents through R1 (if there are any). So it can be more precise.

with best regards

Andreas
Title: Re: Couple Questions about Reference Divider
Post by: Crossphased on October 14, 2018, 11:07:55 pm
Ah ok I gotcha-
I see what your saying. If time constant in the feedback loop is too fast it will constantly chase its tail trying to correct the output voltage and oscillate. Ok, changing the values now.

I started to install LTC2057, but it has an odd package. I have the 10-MSOP version, but it doesnt seem to have the footprint of other MSOP devices I've seen. In the past I've been able to fit MSOP chips on TSSOP breakout boards. The LTC2057 has tighter pin spacing though. Here's LTC2057 on TSSOP breakout board:
(https://farm2.staticflickr.com/1902/30384316707_80e0d0b283_b.jpg) (https://flic.kr/p/NhXyMr)

and here it is on MSOP-8 breakout board, you can see pin spacing  is still to tight:
(https://farm2.staticflickr.com/1971/45275204712_af214b006d_b.jpg) (https://flic.kr/p/2bYPfZN)

here's purchase bag:
(https://farm2.staticflickr.com/1977/45324459941_d54aa27da6_b.jpg) (https://flic.kr/p/2c4aGRT)

Forgive me for my lack of knowledge but is MSOP not a standard pin spacing width?
Title: Re: Couple Questions about Reference Divider
Post by: flittle on October 14, 2018, 11:26:23 pm
I think it is the .5mm standard and the boards you have are .65mm.
This should work: https://www.ebay.com/itm/4-Sets-SOT23-MSOP10-UMAX10-to-DIP10-Adapter-Board-w-Pins-SMD-to-DIP-5-95mm/132799481297 (https://www.ebay.com/itm/4-Sets-SOT23-MSOP10-UMAX10-to-DIP10-Adapter-Board-w-Pins-SMD-to-DIP-5-95mm/132799481297)
Title: Re: Couple Questions about Reference Divider
Post by: Crossphased on October 15, 2018, 06:55:41 am
I See, thanks very much. Weird that pin spacing is different for MSOP8 vs MSOP10. I guess maybe more pins desired in same amount of space? Ok well I learned my lesson from now on SOIC packages
Title: Re: Couple Questions about Reference Divider
Post by: Crossphased on October 15, 2018, 06:58:16 am
Here is AD587 data after 500 hours. Temp hasnt varied a whole lot but you can see voltage change with temperature. There is no temp correction or anything like that in the circuit. Just +15V and a decoupling cap to each reference. How does this compare to data you guys have seen from other AD587 references?

(https://farm2.staticflickr.com/1976/43516367010_4fa543505b_b.jpg) (https://flic.kr/p/29ioKPQ)
Title: Re: Couple Questions about Reference Divider
Post by: Andreas on October 15, 2018, 06:20:05 pm
Hello,

If I look at the 10 hours and the 500 hours data there is nearly no ageing drift visible.
Some results of AD587 are here:
https://www.eevblog.com/forum/metrology/ad587lw-10v-precision-travel-standard/msg1750307/#msg1750307 (https://www.eevblog.com/forum/metrology/ad587lw-10v-precision-travel-standard/msg1750307/#msg1750307)

AD587LW#01 is drifting around +0.5ppm/kHr
AD587LW#02 has nearly no drift over 4 kHrs.

with best regards

Andreas
Title: Re: Couple Questions about Reference Divider
Post by: Crossphased on October 17, 2018, 06:08:21 am
Andreas,
Thanks very much for the link. I'm trying to understand what the graphs represent- what is the time scale?
I'm not sure what the first graph represents (in post #88), are you comparing the AD587 to other LTZ references in this graph?
In the second graph it looks like you are displaying the AD587 voltages as measured by different instruments. It looks like you have a talent for building precise ADCs! Do you divide the AD587 voltage down before reading by ADC or are you measuring between 587 and LTZ?

Also another question- how do you correct readings for temperature, for comparison to each other? Or maybe your circuit is temperature compensated with NTC?

Cheers
Title: Re: Couple Questions about Reference Divider
Post by: Crossphased on October 17, 2018, 06:13:35 am
Also just finished putting the  circuit together. For the moment I replaced LTC2057 with OPA227, it will have to do until I get LTC2057 in proper package:
(https://farm2.staticflickr.com/1932/44654942344_8a7724519d_c.jpg) (https://flic.kr/p/2b31fB3)

(https://farm2.staticflickr.com/1905/31503664058_eef0bedc0b_c.jpg) (https://flic.kr/p/PZSvKG)
Title: Re: Couple Questions about Reference Divider
Post by: Crossphased on October 17, 2018, 06:25:54 am
Here is update schematic:
Again if any values look incorrect or odd please let me know, cheers!

(https://farm2.staticflickr.com/1972/30439358257_6a1f90e54b_z.jpg) (https://flic.kr/p/NnPEHF)
Title: Re: Couple Questions about Reference Divider
Post by: Andreas on October 17, 2018, 07:33:08 am
Hello,

the feedback paths of the OPs are still wrong.
-> they should happily oscillate.
(you should have only re-wired the capacitor and not the resistor).

What type of 1uF capacitors do you use (Mylar or Polypropylene ?)
Which leakage resistance do they have (or discharge time constant).

with best regards

Andreas

Title: Re: Couple Questions about Reference Divider
Post by: Crossphased on October 17, 2018, 06:57:32 pm
Hi Andreas thanks for the feedback,

The capacitors I used with LTC1043 are these: http://www.rubycon.co.jp/en/catalog/e_pdfs/film/e_mpk.pdf, (http://www.rubycon.co.jp/en/catalog/e_pdfs/film/e_mpk.pdf,) I'm using a value of 1uF.
The datasheet lists the insulation resistance as:
C≦0.33μF:25000MΩmin
C>0.33μF:7500ΩFmin

I'll measure the leakage current is tonight. I matched the capacitors, they are within .5% of each other in capacitance.

About the feedback path- I guess I am misunderstanding you. I didnt change the wiring of the feedback path, only the values of the capacitors. How would you connect and what values would you choose?

Title: Re: Couple Questions about Reference Divider
Post by: Andreas on October 17, 2018, 07:20:13 pm
Hello,

not necessary to measure, they are good enough.
(looked for me like some cheap <500s Time constant values)

for the cirquit see figure 3 here:
https://www.analog.com/media/en/analog-dialogue/volume-38/number-2/articles/techniques-to-avoid-instability-capacitive-loading.pdf (https://www.analog.com/media/en/analog-dialogue/volume-38/number-2/articles/techniques-to-avoid-instability-capacitive-loading.pdf)

with best regards

Andreas
Title: Re: Couple Questions about Reference Divider
Post by: branadic on October 17, 2018, 07:22:00 pm
Quote
About the feedback path- I guess I am misunderstanding you. I didnt change the wiring of the feedback path, only the values of the capacitors. How would you connect and what values would you choose?
Title: Re: Couple Questions about Reference Divider
Post by: Crossphased on October 18, 2018, 05:10:52 am
Thank you Andreas and branadic for your kind assistance, it is very much appreciated

I understand what you mean now, I fixed the circuit. The link was very helpful.
After fixing the feedback in the circuit I took some test measurements. First I used a power supply set to various voltages to get a feel for how the circuit behaves. The power supply wasnt perfectly stable so the measurements fluctuated a bit, only accurate to 3 maybe 4 digits. Here's the points of measurement:
(https://farm2.staticflickr.com/1903/45399310651_8b516cb6e4_c.jpg) (https://flic.kr/p/2caMknn)

Here is measurements with various voltages from variable power supply:
(https://farm2.staticflickr.com/1930/45399312691_f795d48e2a_z.jpg) (https://flic.kr/p/2caMkYx)

The op amps dont like the lower voltages without a negative supply. Thats ok I wont be using for low voltages! The OPA227 seems to have a offset throughout its range much worse than the LT1001, it will be good to replace with LTC2057.
Next the AD587 voltages were measured:
(https://farm2.staticflickr.com/1935/45399311661_f32c1394fb_z.jpg) (https://flic.kr/p/2caMkEM)

It looks like even with buffering the LTC1043 is around 400 uV short. But the OPA227 draws 100x more bias current (2.5nA) than the LTC2057 (30pA). Could this account for the discrepency?

Kind of interesting the LTC1043 output is low by 400uV, but the OPA227 is offset high by 330 uV so it almost evens out
Title: Re: Couple Questions about Reference Divider
Post by: Crossphased on October 18, 2018, 03:51:25 pm
ok just took some leakage measurements with 10V on the LTC1043 sampling capacitors. The leakage current was measured to be between .008 - .010 uA at 10V
Title: Re: Couple Questions about Reference Divider
Post by: MiDi on October 18, 2018, 04:26:15 pm
How did you measured the current - with dmm in series in uA range?
Title: Re: Couple Questions about Reference Divider
Post by: Crossphased on October 18, 2018, 07:51:26 pm
Yes that exactly
Title: Re: Couple Questions about Reference Divider
Post by: Crossphased on October 18, 2018, 07:52:40 pm
Andreas,
Is this about the performance you would expect from LTC1043? Or would you expect better performance when LTC2057 is installed?
Title: Re: Couple Questions about Reference Divider
Post by: Andreas on October 18, 2018, 08:49:59 pm
Hmm,

something is odd: 400uV is far from that what I usually have with a LTC1050 or LTC2057

With a 5V reference on input of the LTC1043.
I typically measure around 2.5V - 20uV to 2.5V - 25uV at the buffered output.
(2499.975 - 2499.980 mV)

Be carefully with EMI. If you have a switchmode supply (LED Lamps) near the setup
you can get easily errors of several 100 uV due to rectification effects
on the input or output diodes of the components.
Usually the measurement values are "touch" sensitive if you have EMI effects.

Perhaps you need a additionaly filter cap/RC-filter at the input of the first stage.

with best regards

Andreas

Title: Re: Couple Questions about Reference Divider
Post by: Crossphased on October 19, 2018, 04:08:19 am
Hmm,

something is odd: 400uV is far from that what I usually have with a LTC1050 or LTC2057

With a 5V reference on input of the LTC1043.
I typically measure around 2.5V - 20uV to 2.5V - 25uV at the buffered output.
(2499.975 - 2499.980 mV)

Be carefully with EMI. If you have a switchmode supply (LED Lamps) near the setup
you can get easily errors of several 100 uV due to rectification effects
on the input or output diodes of the components.
Usually the measurement values are "touch" sensitive if you have EMI effects.

Perhaps you need a additionaly filter cap/RC-filter at the input of the first stage.

with best regards

Andreas
I see, thanks very much for the additional information.

Ok I touched around a bit, the only thing that is touch sensitive is the sampling capacitor with the foil shield. When touched, output of LTC1043 drops by 200uV. I'm now going to put RC filter on the LT1001 input. Do you think it would be helpful to put ferrite beads on power supply lines?
Also, if nothing is connected to the LT1001 input, it likes to saturate to the positive rail. Would it be a good or a bad idea to put something like 100K resistor pull down on the input to LT1001?

Cheers
Title: Re: Couple Questions about Reference Divider
Post by: David Hess on October 19, 2018, 04:45:30 pm
The last example here (https://www.analog.com/en/analog-dialogue/articles/ask-the-applications-engineer-25.html) shows how to drive capacitive loads however I often get better results using a series RC network on the output to ground which is simply a standard aluminum or tantalum electrolytic capacitor.  The large capacitance, typically 10 to 100 microfarads but smaller is possible in some cases, combined with the output resistance produces dominant pole frequency compensation and the relatively large ESR prevents emitter/source follower oscillation so do not try this with a low ESR capacitor unless you add series resistance.  Comparators can be compensated to operate as operational amplifiers in the same way.

Quote
Would it be a good idea to trim out the offset from LT1001?

Only if it is a significant source of error.  Do *not* trim the LT1001 offset to correct for other errors in the system as this will degrade its offset voltage drift.

I would consider using an LT1097 instead because of its lower input bias current and overcompensation pin but if you already have an LT1001, just use it.

Quote
- the sampling capacitor C7 was partially wrapped in copper foil for guarding. I'm trying to figure out the best pin on LTC1043 to connect the guard to. As far as I can tell it would be pin 10 is that right?

The LTC1043 datasheet discusses this but I do not remember the details.  Follow the instructions there.

Quote
- only 1/2 of the LTC1043 is being used for the divide by two... would it be a good idea to connect the other half in a divide by 2 configuration as well and connect in parallel to lower output impedance?

This would probably be a waste of time.  I might try it with the switches operating in anti-phase.

Title: Re: Couple Questions about Reference Divider
Post by: Crossphased on October 22, 2018, 11:01:21 pm
Thanks David,
Quick question- in the article, is this the RC filter you are referring to?
(https://farm2.staticflickr.com/1902/43688384490_677cd0164b_z.jpg) (https://flic.kr/p/29yAoCs)


What I understand you saying, is if Cl has a large enough ESR, then Rl isn't necessary. If Cl ESR is too low, then add a small resistor in series with Cl? Or a shunt resistor as pictured?

Thanks very much for your insights
Title: Re: Couple Questions about Reference Divider
Post by: Crossphased on October 22, 2018, 11:14:20 pm
Over the weekend some efforts were made to reduce the error of the LTC1043 divider. I reduce the error, but I was at least able to characterize it. Addressing some of the suggestions earlier, I did a couple different things-
Power supply is a linear supply, Siglent SPD330X
Other sources of switching noise could be LED lights as Andreas suggested, which I have. For measurements all lights were turned off.
One of those clamp on ferrites was put over the wire harness to the board(ps leads and dmm leads)
An RC filter was added on the input to the LT1001

None of these things helped unfortunately.
Next the objective was to characterize the error at different input voltages. Here is the data:
(https://farm2.staticflickr.com/1908/44779858874_76593b3446_c.jpg) (https://flic.kr/p/2be3tWj)
The error seems to be approximately linear proportional to the input voltage. Would it be a valid assumption to assume an EMI induced error would be approximately constant across input voltages?
The error rising along with input voltage seems to suggest some type of leakage, not sure where though. The board was cleaned thoroughly with IPA after soldering. Also interesting is the positive offset at around 1.25V. Would that be internal leakage in the 1043? Or would that suggest the source voltage of the leakage path is somewhere between 1.25 and 2.5V?

Here's another idea, assuming the error IS linear with input voltage, would it be acceptable to add a very small amount of gain in the output opamp to account for the error? Or is this something that is better done in software?

I rather like Davids idea of using the other half of the LTC1043 in opposite phase to the currently used divider, so there's only a very small off time. I'm going to try this next.
Cheers
Title: Re: Couple Questions about Reference Divider
Post by: mimmus78 on October 23, 2018, 12:51:36 pm
I had some problems in the past with this type of linear power supply that have an MCU inside with some references design resulting in some noise increase and shifts. You should try with batteries to check if some noise comes from it.

Inviato dal mio ONEPLUS A5010 utilizzando Tapatalk

Title: Re: Couple Questions about Reference Divider
Post by: Andreas on October 23, 2018, 08:51:17 pm
Would it be a valid assumption to assume an EMI induced error would be approximately constant across input voltages?

The error rising along with input voltage seems to suggest some type of leakage, not sure where though.

Here's another idea, assuming the error IS linear with input voltage, would it be acceptable to add a very small amount of gain in the output opamp to account for the error? Or is this something that is better done in software?

Hello,

I would not assume for EMI errors to be constant (it depends on parasytic capacitance) but they should not be linear to input voltage.
So there is another reason.

Stray capacitance perhaps from the flying capacitor?
One of the differences between your cirquit is that your flying capacitor is much larger than my WIMA MKS 02 (2.5 mm raster) 1uF.
Perhaps your capacitor has a "outer foil" which is on the wrong pin.
So I would try first to swap the pins of the capacitor and then try a smaller one.

I usually handle such things in software (EEPROM coefficient).

For a fixed 5V reference out of 10V I would do a overall adjustment
of the 10V Reference (at the trim pin) in a way that the 5V fit.

with best regards

Andreas

Edit: did you measure the frequency of the LTC1043?
You should have something around 400-500 Hz.

Title: Re: Couple Questions about Reference Divider
Post by: Crossphased on October 24, 2018, 05:10:25 am
Stray capacitance perhaps from the flying capacitor?
One of the differences between your cirquit is that your flying capacitor is much larger than my WIMA MKS 02 (2.5 mm raster) 1uF.
Perhaps your capacitor has a "outer foil" which is on the wrong pin.
So I would try first to swap the pins of the capacitor and then try a smaller one.

I usually handle such things in software (EEPROM coefficient).

For a fixed 5V reference out of 10V I would do a overall adjustment
of the 10V Reference (at the trim pin) in a way that the 5V fit.

with best regards

Andreas

Edit: did you measure the frequency of the LTC1043?
You should have something around 400-500 Hz.

Andreas,
Thanks for the helpful tips. I disconnected the shield lead from pin 10, but no change. Next I shall switch capacitor leads and then reconnect the shield lead.

I hadn't measured the frequency before, good idea! I just measured it and here is result:
(https://farm2.staticflickr.com/1929/44617039105_8e19f120ef_c.jpg) (https://flic.kr/p/2aYDZhV)
Only 250 Hz. Maybe this is the problem? When I built this circuit I know put the correct capacitor value in for oscillation pin, but I will unsolder it and measure to verify.

Is there any way to measure parasitic capacitance? Thanks very much for all the help.
Cheers!
Title: Re: Couple Questions about Reference Divider
Post by: David Hess on October 24, 2018, 04:55:34 pm
Quick question- in the article, is this the RC filter you are referring to?

That is the one.

Quote
What I understand you saying, is if Cl has a large enough ESR, then Rl isn't necessary. If Cl ESR is too low, then add a small resistor in series with Cl? Or a shunt resistor as pictured?

That is right and a resistor is placed in series with the capacitor.  For a standard aluminum electrolytic or tantalum capacitor, this resistor already exists in the form of the ESR.

The idea is that the output resistance of the operational amplifier combined with the shunt capacitance lowers the open loop gain below 1 before the feedback reaches 360 degrees.  The ESR or series resistance adds some phase lead making this easier to accomplish and also prevents parasitic oscillation of the output stage.

It is the brute force way to accomplish this but at least in my experience, it actually provides better performance.
Title: Re: Couple Questions about Reference Divider
Post by: Andreas on October 24, 2018, 05:26:39 pm
Only 250 Hz. Maybe this is the problem? When I built this circuit I know put the correct capacitor value in for oscillation pin, but I will unsolder it and measure to verify.
No no problem in this case (only a much too high frequency would bother).
On the other side:
I usually measure the frequency at the flying capacitor (with some input voltage applied)
If you are putting the scope directly at the capacitor the frequency drops somewhat due to the leakage (scope input resistor).

with best regards

Andreas
Title: Re: Couple Questions about Reference Divider
Post by: Crossphased on October 28, 2018, 05:45:05 am
Time for an update-
Made some new discoveries over the weekend. I believe I have been slightly misinterpreting my measurements. I was experimenting with the LTC1043 circuit that I have seen be too low in voltage, and decided to scope the output of LTC1043. The DMM was already attached. As soon as scope probe was attached, I saw the DMM measurement drop significantly. More investigation revealed the DMM measurement dropped by 300uV when the scope probe is attached. The scope probe I'm using is 10X, 10MegaOhm input impedance. So a 10M load drops the output by 300uV. Next I realized the meter I've been using to take measurements has only 10M input impedance. So the large error I have been seeing is likely due to the DMM loading down the LTC1043 output. What I had considered to be large offset voltage of OPA227, was actually the OPA227 drawing a lower bias current than my meter! Ok I think its time to invest in a high impedance DMM!

Also some less exciting news-
Put together another LTC1043 circuit with the capacitors in anti-phase as David suggested. I immediately saw problems with this circuit, I believe I have a faulty chip or somehow damaged it. The readings were very poor. in the divide by 2 configuration I was getting readings anywhere from 80 mV to 130 mV too low.
I saw current draw on power supply go up significantly, so I decided to measure. Reading only the LTC1043 current draw, it is pulling 17.8 mA ! I think normally its supposed to pull under 1 mA. I think this may be a bad one.

I also saw the reference voltage source drop in voltage when connected to this 1043. So started probing around with the scope. Observed large switching transients on the input.
Here is the scope on the sampling capacitor, with no signal voltage applied. 800 mV transients at the switching frequency:
(https://farm2.staticflickr.com/1942/44867734284_0f454a7585_z.jpg) (https://flic.kr/p/2bmNSeA)

Zooming in a bit the transient looks like this:
(https://farm2.staticflickr.com/1927/44867733954_39eca36240_z.jpg) (https://flic.kr/p/2bmNS8U)

Next a voltage was applied to sampling capacitor, transients reduced in amplitude. Weird.
(https://farm2.staticflickr.com/1972/44867733574_a6d1dbbb92_z.jpg) (https://flic.kr/p/2bmNS2m)

Next I measured the original LTC1043 circuit, to compare waveforms. Here is original LTC1043 sampling capacitor. 35 mV P-P. much quieter:
(https://farm2.staticflickr.com/1961/44867733274_9804b49ccd_z.jpg) (https://flic.kr/p/2bmNRWb)
much lower noise

So the bad LTC1043 will be replaced.
Title: Re: Couple Questions about Reference Divider
Post by: Crossphased on October 28, 2018, 05:47:26 am
Quote
That is right and a resistor is placed in series with the capacitor.  For a standard aluminum electrolytic or tantalum capacitor, this resistor already exists in the form of the ESR.

The idea is that the output resistance of the operational amplifier combined with the shunt capacitance lowers the open loop gain below 1 before the feedback reaches 360 degrees.  The ESR or series resistance adds some phase lead making this easier to accomplish and also prevents parasitic oscillation of the output stage.

It is the brute force way to accomplish this but at least in my experience, it actually provides better performance.

Thanks very much for that David. I shall use this in my next circuit and report back the results! Your experience and insight is very much appreciated
Title: Re: Couple Questions about Reference Divider
Post by: Crossphased on October 29, 2018, 04:06:10 am
Hello Friends,

Ok I have some much better results to report. I was able to put my DMM in 10G input impedance mode, and the measurements improved significantly. Now the max error is around 90uV at 10V level, and around 50uV at 5V level. This is quite encouraging, and makes LTC1043 seem a good application as input divider for ADC. Here are the results:

(https://farm2.staticflickr.com/1938/45610007331_8eedb70316_o.png) (https://flic.kr/p/2cupd9K)

(https://farm2.staticflickr.com/1929/30669117977_a9fdac03b2_o.png) (https://flic.kr/p/NJ8fha)

So at the 5V level Andrea has 20-25uV error, and I am getting around 50uV low. That is ok with me. I am closing in! The data doesnt seem as linear this time, almost like there are two "levels" or plateaus.

I have a couple thoughts-
LTC1043 is limited by 18V max Vdd, would this be a difficult device to make out of discrete components to allow higher input voltages?
Also, is error expected to get better or worse with larger divisor ratios?  My first instinct would suggest to me that accuracy would improve with more division, but the datasheet suggests the opposite.

Cheers and thanks for help everyone!
Title: Re: Couple Questions about Reference Divider
Post by: Kleinstein on October 29, 2018, 04:56:18 pm
It is possible to build a charge pump with other, higher voltage CMOS switching chips (e.g. DG411 or similar) and a suitable clock. However other chips tend to have a higher charge injection and thus a slightly higher error is expected.  There might be newer ones with comparable performance, though usually without the internal clock.

The error has three main contributions:  Loading by leakage / input currents to the DMM/OP,  parasitic capacitance and charge injection. So one can not expect a perfect division, but a rather stable one. The charge injection may show a nonlinear dependence on the voltage. Parasitic capacitance should be mainly proportional or constant (coupling to control lines). The loading effect would be about linear in the load.

Title: Re: Couple Questions about Reference Divider
Post by: Andreas on October 30, 2018, 06:19:08 am
I am closing in! The data doesnt seem as linear this time, almost like there are two "levels" or plateaus.

The plateaus are due to the limited resolution (quantisation error) of your DMM readings.
So you will need at least 1uV resolution or better to judge linearity.
Most instruments have one additional digit resolution if you read them out by the PC-interface.

The other way would be to use a 2nd source in series to the instrument as zero Voltmeter in 100mV range and compare both partial voltages on the LTC1043.
(from input to output and from output to gnd).

with best regards

Andreas
Title: Re: Couple Questions about Reference Divider
Post by: Crossphased on October 31, 2018, 04:23:23 am
Quote
It is possible to build a charge pump with other, higher voltage CMOS switching chips (e.g. DG411 or similar) and a suitable clock. However other chips tend to have a higher charge injection and thus a slightly higher error is expected.  There might be newer ones with comparable performance, though usually without the internal clock.
Aha, I see what you mean. I did some reading on charge injection. So it seems all CMOS devices have it. I read there are some mitigation strategies for countering charge injection, but that leads to complication I guess. Is that what the "Charge Balancing" function block in the LTC1043 datasheet represents? Some charge injection countermeasures?

Perhaps there is a mechanical solution. An old friend of mine had a switching capacitor military piece of equipment. It was a step up/step down converter between 12V batteries and 24V batteries. It used a multivibrator as the switching element. Maybe reed relays could be used? Or Mercury wetted relay? Maybe it is experiment time!

Quote
The error has three main contributions:  Loading by leakage / input currents to the DMM/OP,  parasitic capacitance and charge injection. So one can not expect a perfect division, but a rather stable one. The charge injection may show a nonlinear dependence on the voltage. Parasitic capacitance should be mainly proportional or constant (coupling to control lines). The loading effect would be about linear in the load.

I see so error terms are loading, parasitic capacitance, and charge injection. The loading can be minimized with better op amp selection. Leakage perhaps maybe improved with guarding and better construction methods. An actual pcb would probably be better than the breadboard I currently have. Regarding the parasitic capacitance, would you consider this to be mostly parasitic capacitance within the LTC1043 package, or do you think a fair percentage is coupling to other things on the board? For instance if one were to make a max effort construction of an LTC1043 divider, what improvements would you make to yield the largest gains?

This actual got me thinking a little bit on ways to improve this circuit. I will need to take some more accurate measurements to characterize the shape of the error curve precisely. If the error terms are dominated mostly be linear components, then it should be able to find a "best fit" value of resistance (very high) to connect directly between input and output. So that charge lost during switching will be supplemented by current through the resistor. Would this be a valid assumption to make? The other thought I had to improve performance, would be to utilize the other half of the LTC1043 that is not being used. Use that in a multiply configuration, but with the transfer capacitor being a very small value. Probably a few pf. This then gets added on to the output.  Just some ideas I'm throwing around. I think I should do experimentation and see what happens.

Another thought I had is using a vacuum tube on the output instead of low bias current opamp. connect output of LTC1043 to vacuum tube grid, and I think then the voltage can be extracted without drawing much current. Vacuum tube voltmeter I guess.

Thankyou Kleinstein for you advice. I always enjoy your intelligent responses. Cheers
Title: Re: Couple Questions about Reference Divider
Post by: Crossphased on October 31, 2018, 05:49:14 am
Quote
The other way would be to use a 2nd source in series to the instrument as zero Voltmeter in 100mV range and compare both partial voltages on the LTC1043.
(from input to output and from output to gnd).

with best regards

Andreas
Thanks very much Andreas, I see what you mean by quantization error. If I have more resolution in the multimeter then I could see the "missing" parts of the line. But right now the lines is in between LSB steps of the analog converter. 

I'm not sure I understand what you are suggesting for the instrumentation setup. I am trying to picture it in my mind. Is this what you had in mind? I understand it would be good to compare (Va - Vc)  to (Vc - Vd), but I dont know the best method for doing that. For the "source" would it be good to use say 5V LTC6655? or maybe 10V reference through voltage divider?
(https://farm2.staticflickr.com/1951/45644385421_fd61e246d5.jpg) (https://flic.kr/p/2cxrpyr)
Cheers!
Title: Re: Couple Questions about Reference Divider
Post by: Andreas on October 31, 2018, 03:39:45 pm
Hello,

yes exactly this way.
But if you do not have a buffered input on the LTC1043 I would recommend to use a
 minimum 1nF capacitor on VA to GND to filter out EMI (from cabling) which would otherwise give offsets

with best regards

Andreas
Title: Re: Couple Questions about Reference Divider
Post by: David Hess on November 01, 2018, 12:54:06 am
Another way to design a high precision charge pump is to use optically pumped switches in the form of either LDRs (light dependent transistors) or photo FETs.  Optical coupling reduces charge injection significantly.
Title: Re: Couple Questions about Reference Divider
Post by: MiDi on November 01, 2018, 07:07:42 am
Another way to design a high precision charge pump is to use optically pumped switches in the form of either LDRs (light dependent transistors) or photo FETs.  Optical coupling reduces charge injection significantly.

I thought LDR is light dependent resistor.
The LT1043 has quite low charge injection in the range of couple pC (p. 7).
Do you have a clue how much charge injection optically switches have?
Title: Re: Couple Questions about Reference Divider
Post by: Echo88 on November 01, 2018, 08:01:32 am
I love these discussions early in the morning with a cup of coffee  :)

H11F1 claims to have no charge injection https://www.electrokit.com/uploads/productfile/41014/0900766b812cfb59.pdf (https://www.electrokit.com/uploads/productfile/41014/0900766b812cfb59.pdf)

MUX36D04 http://www.ti.com/lit/ds/symlink/mux36s08.pdf (http://www.ti.com/lit/ds/symlink/mux36s08.pdf)
The MUX device claims ~0.5pC source-drain and ~2pC drain-source charge injection-maximum over a +-10Vpp input range (Figure 12 and 13, Vin +-10V, +-15V supply). 2pC into 1µF would equate to a 2µV charge injection error.

Also a bootstrapped LTC1043 would always stay in the region of Vcc/2 (max 18V supply, no 10Vin bootstrapped possible in this case) and therefore have minimal charge injection i presume, havent build it myself.
Title: Re: Couple Questions about Reference Divider
Post by: MiDi on November 01, 2018, 10:28:39 pm
H11F-series summary:
+no charge injection
-high leakage/low off resistance
-slow
-high drive current

This seems not to be a good combination for sample & hold and therefore does not impress me so much.

The MUX36 is indeed an impressive part - have to put it onto my notepad.
Title: Re: Couple Questions about Reference Divider
Post by: Echo88 on November 02, 2018, 04:48:44 pm
Tried to simulate the H11F1 within LTSpice as a discrete substitute for the LTC1043, file attached.
Apparently the high off-state dark current is the main problem, if i have correctly simulated the effect through the use of current sources.
By reversing the current of I2 and I4 the dark current sources eliminate each others error.
Question is if the dark current stays constant over device lifetime or if the simulation isnt correct.
Thoughts about this? 

Edit: the PVA3054NS has better specs compared to the H11F1 and yields way better results in the simulation, since the leakage if much lower.

PVA3054
response time 25µs/15µs
RDSon 160R
Off-state-resistance 10GR
Off-state-leakage ~15pA typical

H11F1
response time 45µs/45µs
RDSon 200R
off-state-resistance 300MR
off-state-leakage 50nA max