Author Topic: DIY low frequency noise meter and some measurement result of voltage references  (Read 137182 times)

0 Members and 2 Guests are viewing this topic.

Offline Insatman

  • Frequent Contributor
  • **
  • Posts: 278
  • Country: ph
To measure the noise of the amplifier itself, the short circuit case is more relevant than the open case. This will especially lower the noise for the version with higher resistor to ground (e.g. 2 K / 20 K and 100 µF cap). Usually signal sources are kind of low impedance compared to the resistor at the input - if not, one has to include the reduced gain / source loading. So the open circuit test is misleading.

The gain is expected to be linear down to low amplitudes. There are quite a few resistors that reduce the gain a little, but the simulation should tell.

Amplifier noise was checked with a 50 ohm terminator at the input and the case closed (see JPG on post marked SETUP).  The scope input noise was checked without a terminator. 
« Last Edit: May 31, 2018, 03:54:28 am by Insatman »
Retired Pulsed Power Engineer/Physicist...now I just dabble in electronics
 

Offline Insatman

  • Frequent Contributor
  • **
  • Posts: 278
  • Country: ph
Based on experience from my last PCB design I decided to make a new one.  The original design had a few errors that required bodges and a couple of the component spacings were off a bit.  The new layout is much cleaner and is mechanically compatible with the original PCB design.   Gone is the ability to use a 100uF film capacitor in place of the 3300uF electrolytic, to use smaller multiple film capacitors for the large 22uF film cap or to substitute U1 (ADA4522-4) with LT1037.   Schematic and layout attached.
« Last Edit: May 31, 2018, 04:06:24 am by Insatman »
Retired Pulsed Power Engineer/Physicist...now I just dabble in electronics
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 3222
  • Country: de
Both the 1K version and 2/20K versions had -3dB points of ~0.1-4Hz.   

Settling time each time a measurement is started is long…often taking a minute or so for both op-amps to come out of saturation and settle near the baseline.

probably the additional low pass R8/C16? could be the reason.
R8 (10K) + the following 1K (R20?) resistor are larger than the 3K6 in my cirquit.

It is clear (to me) that the settling time is long for a 0.1 Hz lower frequency corner.
(Time constant 3.3 sec and at least 10-20 Tau to settle to uV level)

To measure the noise of the amplifier itself, the short circuit case is more relevant than the open case.
The relevant case is with a low noise around 10V source (e.g. 8*NiMH AA cells at constant temperature).
Otherwise you do not get the real life noise due to leakage current of the input capacitor.
A short or open will give too low values for the noise floor.

with best regards

Andreas

 

Offline Insatman

  • Frequent Contributor
  • **
  • Posts: 278
  • Country: ph
Both the 1K version and 2/20K versions had -3dB points of ~0.1-4Hz.   

Settling time each time a measurement is started is long…often taking a minute or so for both op-amps to come out of saturation and settle near the baseline.

probably the additional low pass R8/C16? could be the reason.
R8 (10K) + the following 1K (R20?) resistor are larger than the 3K6 in my cirquit.

It is clear (to me) that the settling time is long for a 0.1 Hz lower frequency corner.
(Time constant 3.3 sec and at least 10-20 Tau to settle to uV level)

Andreas

In the case of the "Orig" circuit where the LT1037 is used, a 2.5K ohm resistor is used rather than a 10K resistor.  This is noted on the schematic.  R20 is 1.2K, so the combination is 2.5 + 1.2 giving 3.7K.  Quite close to your origional circuit.   What I forgot to mention in my post is that the "New" version had significantly faster settling time, especially with the 100uF cap installed.   I'm guessing that the LT1037 just takes much more time to come out of saturation than the ADA4522-4.   
Retired Pulsed Power Engineer/Physicist...now I just dabble in electronics
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 14076
  • Country: de
Electrolytic caps have a rather long internal settling time due to dielectric absorption. So independent on the capacitance value it is just the capacitor internal that takes a long time to settle. Depending on the capacitor type this can take rather long - well in to the seconds or even minutes for the time constant. Checking the settling / dielectric absorption could be another parameter in addition to leakage to look at when choosing caps.

If it would be just the RC time-constant at the input and a later filter stage, there would be the option to speed things up, but reducing the resistor for the initial settling phase.

The LT1037 should be very fast to come out of saturation (more like < 1 µs range). An AZ OP like the ADA4522 usually takes a little longer (10 µs - a few ms) - especially the modern ones with higher GBW may be a little faster. The difference is more like the difference film cap versus electrolytic cap.

The real world test is with a low noise voltage source - especially with an electrolytic capacitor.  With a film capacitor the extra noise compared to a short should not be much, but with an electrolytic the extra noise due to leakage can be important. The open input could give a higher noise, as the resistor to ground in not shunted paralleled with the signal source. So it depends on the input capacitor if the open input noise is higher of lower than the real test with a voltage source.
 

Offline Insatman

  • Frequent Contributor
  • **
  • Posts: 278
  • Country: ph

The relevant case is with a low noise around 10V source (e.g. 8*NiMH AA cells at constant temperature).
Otherwise you do not get the real life noise due to leakage current of the input capacitor.
A short or open will give too low values for the noise floor.

with best regards

Andreas

I have ordered some NiMH cells and a charger to do this test.  They are not readily available here, so Digikey to the rescue. 
Retired Pulsed Power Engineer/Physicist...now I just dabble in electronics
 

Offline Insatman

  • Frequent Contributor
  • **
  • Posts: 278
  • Country: ph
Electrolytic caps have a rather long internal settling time due to dielectric absorption. So independent on the capacitance value it is just the capacitor internal that takes a long time to settle. Depending on the capacitor type this can take rather long - well in to the seconds or even minutes for the time constant. Checking the settling / dielectric absorption could be another parameter in addition to leakage to look at when choosing caps.

If it would be just the RC time-constant at the input and a later filter stage, there would be the option to speed things up, but reducing the resistor for the initial settling phase.

The LT1037 should be very fast to come out of saturation (more like < 1 µs range). An AZ OP like the ADA4522 usually takes a little longer (10 µs - a few ms) - especially the modern ones with higher GBW may be a little faster. The difference is more like the difference film cap versus electrolytic cap.

The real world test is with a low noise voltage source - especially with an electrolytic capacitor.  With a film capacitor the extra noise compared to a short should not be much, but with an electrolytic the extra noise due to leakage can be important. The open input could give a higher noise, as the resistor to ground in not shunted paralleled with the signal source. So it depends on the input capacitor if the open input noise is higher of lower than the real test with a voltage source.

I think you are most likely right on this.  The difference between the two circuits could simply be the characteristics of the individual electrolytic caps.  I noticed that the fastest settling time by far was with the film cap. 
Retired Pulsed Power Engineer/Physicist...now I just dabble in electronics
 

Offline Insatman

  • Frequent Contributor
  • **
  • Posts: 278
  • Country: ph
REPORT ON FINAL 10/100K:1 AMPLIFIER PERFORMANCE.

I got the revised PCB in recently from JLCPCB.  I built up the board using new components except for the Electrolytic capacitor and the largest film caps (2.2uF and 22uF) which I salvaged from one of my prototypes.  The board worked with no issues.   Performance is slightly better than my best previous prototype.  The difference is probably due to superior board layout, especially since I re-used the same Electrolytic capacitor, so capacitor leakage should be the same as the prototype. 

Noise floor tests.   Tests were done using a 50 ohm resistor across the input and also with a NiMH battery based low-noise source. 


To measure the noise of the amplifier itself, the short circuit case is more relevant than the open case.
The relevant case is with a low noise around 10V source (e.g. 8*NiMH AA cells at constant temperature).
Otherwise you do not get the real life noise due to leakage current of the input capacitor.
A short or open will give too low values for the noise floor.

Andreas

I bought 8x NiMH AA (EverReady) cells from Digikey and charged them with a new T4S Tenergy charger bought for this purpose.  I found the output voltage surprising noisy as shown in an attached scope waveform of the output on my 10K:1 amp.  I added some filtering components to clean up the output.  The battery was housed in a cast AL enclosure with BNC connector output.  Schematic for battery box is also attached.   

Noise floor of the new amplifier with 50 ohms across the input (10K:1 mode with 1K input resistance) was typically <120nV pk-pk. 
Nose floor of new amplifier using the low-noise NiMH source box was typically <150nV pk-pk.

I also measured a couple of LTZ-1000 based 10V references I have built recently.  One is housed in a battery backed Vref chassis and the other is just a PCB in a plastic bag on the bench.   Value of <1.5 uV were measured with the Vref chassis being a bit quieter as expected.

Some photos and board  As-built schematic are attached below.  This project has been a great learning experience and just another adventure traveling the rabbit hole of volt-nut pursuits. 

Insatman
Retired Pulsed Power Engineer/Physicist...now I just dabble in electronics
 
The following users thanked this post: Andreas, kj7e

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 3222
  • Country: de

I bought 8x NiMH AA (EverReady) cells from Digikey and charged them with a new T4S Tenergy charger bought for this purpose.  I found the output voltage surprising noisy as shown in an attached scope waveform of the output on my 10K:1 amp. 

I usually put the battery and the amplifier into a (grounded) cookies box. (only the BNC-line to the scope runs outside).

with best regards

Andreas
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 14076
  • Country: de
Newly charged batteries might show a little more noise and drift. It can take some time for the chemistry to settle and this might not always be smooth.  Similar it might take quite some time for the input cap to settle with the new voltage.

It is very unusual to need filtering after the battery.
 

Offline Insatman

  • Frequent Contributor
  • **
  • Posts: 278
  • Country: ph
Newly charged batteries might show a little more noise and drift. It can take some time for the chemistry to settle and this might not always be smooth.  Similar it might take quite some time for the input cap to settle with the new voltage.

It is very unusual to need filtering after the battery.

It was approximately 16 hours after charge that the measurement labeled RAW BATT was made.   I was thinking along those lines, but decided that a filter, so long as it used FILM caps, would probably be helpful.  I know it takes the output voltage a bit of time to settle  now, but that is dwarfed by the time the electrolytic cap takes to settle in the amplifier anyway.  Perhaps these EverReady cells are more noisy than most?   I don't have any experience with NiMH batteries in this realm. 
« Last Edit: June 09, 2018, 12:58:33 am by Insatman »
Retired Pulsed Power Engineer/Physicist...now I just dabble in electronics
 

Offline Gerhard_dk4xp

  • Frequent Contributor
  • **
  • Posts: 322
  • Country: de
These are two 18650 Li Batteries in series, abt. 7.5Vdc:
<    https://www.flickr.com/photos/137684711@N07/39056813010/in/album-72157662535945536/    >

0 dB is 1nV/rt Hz, the batteries hit my measurement limit. I intend to move this limit further
both by using a chopper amplifier and by using cross correlation.
There are other plots in this album that cover barefoot & filtered LT6655 and LEDs.

Other batteries:
<   http://www.hoffmann-hochfrequenz.de/downloads/NoiseMeasurementsOnChemicalBatteries.pdf     >

The worse than 1/f rise of the noise below 50 Hz goes on the pre amp that had too small a coupling
capacitor at the time the measurements were made. That has been healed with a big wet slug tantalum
in the mean time. The FFT analyzer also adds low frequency noise.

There are more papers on this subject in that directory.

The NIST paper given in the reference is interesting also, at least as long as it does deal with
voltage noise only.

12V 12AH Pb will follow this weekend when the weather is bad here, i.e. I don't need my motorbike
for driving around and can lend it's battery.

regards,
Gerhard
« Last Edit: June 09, 2018, 01:21:44 am by Gerhard_dk4xp »
 
The following users thanked this post: KE5FX, kj7e, Insatman

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 3222
  • Country: de
It is very unusual to need filtering after the battery.

Thats true.

I also think that the input capacitor was not charged long enough. (2-3 days)
I always keep the input capacitor charged to about 10V except for noise measurements.

With best regards

Andreas
 
The following users thanked this post: Insatman

Offline Insatman

  • Frequent Contributor
  • **
  • Posts: 278
  • Country: ph
It is very unusual to need filtering after the battery.

Thats true.

I also think that the input capacitor was not charged long enough. (2-3 days)
I always keep the input capacitor charged to about 10V except for noise measurements.

With best regards

Andreas

I will leave the unit's input connected to the NiMH battery (x8 in series w/ filter) for 3 days and repeat the noise measurements.

Insatman
Retired Pulsed Power Engineer/Physicist...now I just dabble in electronics
 

Offline Insatman

  • Frequent Contributor
  • **
  • Posts: 278
  • Country: ph
New 10000:1 amplifier noise measurements using NiMH battery as source.
After 3+ days biased at 10V I got slightly lower noise measurements than the previous measurements.   Prior measurements were typically <150nV, now after 3+ days of being biased at 10V, the typical number is <130nV.  This is approaching the noise floor when the amplifier is shorted by a 50ohm terminator.  Still the difference of 20nV, while significant, isn't going to affect measurements of a =>1uV very much.  Conclusion is that keeping the amplifier biased at 10V is a good idea, but not essential for making good measurements so long as the noise level you are measuring is >1uV.
Retired Pulsed Power Engineer/Physicist...now I just dabble in electronics
 
The following users thanked this post: Andreas

Online vindoline

  • Supporter
  • ****
  • Posts: 324
  • Country: us
I'm thinking that a low freq LNA is a god next project. Is there any consensus amoung the group which one to build? I'm not up to designing my own from scratch (yet!), but I'll probably lay out my own board to fit my case of choice. My use will be to measure my D.C. reference noise and compare it to the US cal club ref when it comes along my way.
Thanks in advance.
« Last Edit: June 21, 2018, 01:19:09 am by vindoline »
 

Offline flittle

  • Contributor
  • Posts: 17
  • Country: us
Does anyone have any extra LNA PCBs?  I would be happy to purchase them with shipping costs of course.
DS1054Z, HP3455A, HP3457A, Agilent 34401A, HP5334B-010-030, HP204D, EX430, Agilent 6612C, (2) Sorensen XTS15-4 /M1 /M9B, WaveTek 131, WaveTek 134,PAR 110, FG-8002,FY3200S, UNI-T61E, TEK2465
 

Offline Echo88

  • Frequent Contributor
  • **
  • Posts: 821
  • Country: de
User mimmus78 might still have a few. I got two PCBs from him which work very well.  :-+
 

Offline mimmus78

  • Supporter
  • ****
  • Posts: 676
  • Country: it
Thanks Echo88, yes I still have few ones ... and some time to ship now.

Anyway Great part of the merit should go to Andreas that published his schematic ...
 
The following users thanked this post: Andreas

Offline flittle

  • Contributor
  • Posts: 17
  • Country: us
PM sent mimmus78. 
DS1054Z, HP3455A, HP3457A, Agilent 34401A, HP5334B-010-030, HP204D, EX430, Agilent 6612C, (2) Sorensen XTS15-4 /M1 /M9B, WaveTek 131, WaveTek 134,PAR 110, FG-8002,FY3200S, UNI-T61E, TEK2465
 

Offline mimmus78

  • Supporter
  • ****
  • Posts: 676
  • Country: it
Hi guys, all pcbs based on Andreas design are gone right now and travelling across Europe.
Next time I will do some others pcb for me I will respin the boards ...
 

Offline RandallMcRee

  • Frequent Contributor
  • **
  • Posts: 540
  • Country: us
Recently made some measurements of my equipment using a Pipelie LNA (s/n 1157A). Values below are RTTI. Should be comparable to Zlymex results column 6.

Fluke 731B #1 avg = 2.5 uV p-p
Fluke 731B #2 avg = 1.7 uV p-p

Malone DMMCheck 5V output avg = 26.8 uV p-p

Ensemble of 7 LTZ1000 7.115V output = .39 uV p-p
Ensemble 5V output (custom resistive divider) = .35 uV p-p
Ensemble 10V output (opamp X2) = .87 uV p-p

My measurement setup is pretty basic. In particular, I have not shielded everything very well and I notice that readings increase when I am nearby in the room. Still, the Fluke 731B measurements suggest that I am not wildly wrong.

However it seems that my circuits do not preserve the low noise of the ensemble very well. Straightforward ratios would suggest
7V output = 1.2uV/sqrt(7) = .45 uV <ideal>    <--did ok here!
5V output = 0.39/1.43 = 0.27uV <ideal>        <-- marginal
10V output = 5V*2 = 0.55 uV    <ideal>        <-- very marginal, but good compared to zlymex table right?

C27 and C28 were bodged in after measuring worse noise figures. C27 helped about 10% and C28 another 10%. Also deleted an unused LTC1043 (left over from an experiment to get 10V using it, but replaced as shown) and gained another 5% decrease.

Any thoughts on taking this further or should I just count my blessings? The above is a result of several months of work.

Should I try deriving the 10V output directly from the 7V output?  (I won't do this lightly since preserving the tempco of the ensemble will require something like a VHP100 voltage divider, $$$)

Has anyone tried something like this? Experience to share?

Also:
LNA noise floor = .23 uV p-p measured in this setup (specs say .1uV so perhaps there is a problem here?)

Thanks!
Randy
 
The following users thanked this post: Echo88

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 3222
  • Country: de
Hello,

for the results:
how often did you repeat each measurement?
how was the standard deviation of the measurement?
how long was each measurement (10 s or 100 s)?

For 10s measurements I typically have a stray or +/-20 .. 30%.
And this under ideal conditions (DUT + amplifier in cookies box).

I usually do 15-20 repeated measurements and calculate average + standard deviation.

with best regards

Andreas
 

Offline Echo88

  • Frequent Contributor
  • **
  • Posts: 821
  • Country: de
Very interesting results Randall :-+

Could you please elaborate on your 7xLTZ-Ensemble and the 5V/10V-ensemble? Did you connect the 7 LTZ-references via the 2k-input-resistors to get the 5V/10V?
Nobody answered yet how much the soakage time of the LNA-input-caps matters/how big the noise-difference is; did you wait long enough after connecting the 7.1/5V/10V-ensemble to the LNA?
I cant say anything about the circuit and can only point to the AN159, which describes how to build a nice shielding enclosure.
I also have a 0.1-10Hz-LNA, but need to build a testbox with various voltage-references as a sanity-check.

A more detailed review of your measurements with pictures would be nice, if you have time.  :-DMM
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 14076
  • Country: de
The peak to peak measurements always show quite some scattering. It is noise one is measuring after all, so it is not predictable.
To get accurate values it takes averaging - even with 25 intervals of 10 s averaged the scattering is expected to go down by a factor of 5 only. With little time in between chances are averaging might be even a little less effective.

Looking at the RMS value is a little faster. Here the scattering is smaller to start with. However the factor between RMS and peak to peak values is not absolutely constant (due to correlation effects of things like popcorn noise). Still the RMS values gives a faster and less scattering indication of the noise.

The circuit looks reasonable. One point that might contribute to extra noise could be an interaction between the chopper stabilized OPs.

The LNA noise floor looks like a little on the high side. Is this measured open circuit or short circuit ? The open circuit noise is expected to be higher and does not apply to the later use.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf