Author Topic: DIY low frequency noise meter and some measurement result of voltage references  (Read 67284 times)

0 Members and 1 Guest are viewing this topic.

Offline lukier

  • Supporter
  • ****
  • Posts: 596
  • Country: gb
    • Homepage
Looks like a nice & useful piece of kit. I wonder how it compares to trusty Tek's AM502 (with Jim Williams' mod for 10 Hz - see AN124).  :-+
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 5759
  • Country: de
The amplifier(s) shown in this thread are considerably lower bandwidth (more like kHz) than the Tek AM502. They are also lower voltage noise in the LF range. However current noise can be higher and the input impedance is lower.

There is another tread on an JFET (BF862) based LNA that is more similar to the AM502 - though lower noise.
 
The following users thanked this post: lukier

Offline mimmus78

  • Supporter
  • ****
  • Posts: 652
  • Country: it
Quote from: Andreas
Andreas

Just wondering if Andreas design that I repost here is still valid ... seems quite simple design if used with a "sensible" scope.



 

Offline mimmus78

  • Supporter
  • ****
  • Posts: 652
  • Country: it
The amplifier(s) shown in this thread are considerably lower bandwidth (more like kHz) than the Tek AM502. They are also lower voltage noise in the LF range. However current noise can be higher and the input impedance is lower.

Can the AM502 be powered without the frame?
 

Offline lukier

  • Supporter
  • ****
  • Posts: 596
  • Country: gb
    • Homepage
Can the AM502 be powered without the frame?

I guess so - haven't checked on AM502 because I have the frame now (TM501) but I did that with AM503 some time ago and you'll need AFAIR +-33.5V and +11.5V and two power transistors (one NPN, one PNP). Check TM501 & AM502 service manuals.
 

Online David Hess

  • Super Contributor
  • ***
  • Posts: 9495
  • Country: us
  • DavidH
Can the AM502 be powered without the frame?

The AM502 accepts +/-35 volts and +11 volts and regulates them down to +/-15 volts using a pair of external power transistors and +5 volts so it would not be too difficult to operate it with an external power supply.
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2323
  • Country: de
Just wondering if Andreas design that I repost here is still valid ... seems quite simple design if used with a "sensible" scope.

Even my old HAMEG 412-5 does the job.
2 mV/div (maximum gain in 5mv/div range) gives 0.2uV/div referred to input.
So a LTZ1000A will give nearly full screen picture.

see also here:
https://www.eevblog.com/forum/projects/low-frequency-very-low-level-dc-biased-noise-measurements/msg658105/#msg658105

With best regards

Andreas

Edit: link added
« Last Edit: March 02, 2017, 07:33:28 am by Andreas »
 

Offline mimmus78

  • Supporter
  • ****
  • Posts: 652
  • Country: it
Even my old HAMEG 412-5 does the job.

I was thinking 2mV was something not so common for a scope.

I got op amps. So now I'm exercising myself at drawing the PCB.

I started some capacitor testing too and I found an old Nichicon 16V 2200uF that went down to 8nA in a hour (10nA in 30 minutes) all the others seems to be very leakiiiiing.

Kinda like the idea to not have to wait forever for making the measurement. Will it be enough to use just this 2200uF?
Do I need to re-tune the filter?
« Last Edit: March 04, 2017, 10:18:06 am by mimmus78 »
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2323
  • Country: de
Hello,

be carefully when soldering. (leakage may go up again).
Standard 85 deg C Capacitors show usually a lower leakage than the 105 deg C types.
from 10 quality capacitors you should usually find 2 which are usable (after 2 days at 10 V).

What is not in the simulation schematic:
I have 2 * 100nF and 2*1000uF (the more leaking) for decoupling.
Between pre-amp and 2nd stage there is also a 100 Ohms in +/- power supply.
At the input there is a additionally 3K6 resistor (with shorting switch for measurement)
to charge up the input capacitor when connecting to a LTZ1000.
Otherwise you will damage the LTZ because the heater setpoint goes too long to infinite.

If you use only 2200uF you could use 1K5 as input pull down resistor.
But in this case you will have more noise floor due to current noise of the LT1037.
With only 2200uF you loose somewhat at the 0.1 Hz corner.
(not very much since the design is robust against input capacitor tolerance).

But if you use the cirquit only for comparative measurements this should not hurt much.

with best regards

Andreas

ps: and do not forget the cookies box on your BOM

http://www.lambertz-shop.de/gebaeck/gebackmischungen/composition-1000g.html

« Last Edit: March 04, 2017, 04:36:23 pm by Andreas »
 
The following users thanked this post: TiN, mimmus78

Offline mimmus78

  • Supporter
  • ****
  • Posts: 652
  • Country: it
Yes and cookies can be used during assembling :-) too
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 5759
  • Country: de
A smaller coupling capacitor slightly increases the effect of current noise at the low frequency end. Increasing the capacitor to ground to keep the input frequency response about the same does not cause more noise:

The main effect is to increases the low frequency gain and thus more of the amplifier noise and more of the source noise is visible. It is only below the -3dB limit that the effect of amplifiers current noise really increases. In the intermediate region one even has the benefit of less current noise of that resistor. So a higher resistor values is improving the SNR ratio at the low end. The lower frequency limit is better set by a later stage or in the digital domain and the input RC slower by something like a factor of 10.  The downside of a very large resistor is, that it takes longer for the input to settle.
However such low frequency noise measurement will not be very fast anyway - for most sources it takes at least minutes to thermally settle.

With 2200 µF and 1 K the input RC start to contribute to the 0.1 Hz lower limit - so a larger resistor is a good idea, I would even prefer more like 5 K.
 

Offline mimmus78

  • Supporter
  • ****
  • Posts: 652
  • Country: it
What is not in the simulation schematic:
I have 2 * 100nF and 2*1000uF (the more leaking) for decoupling.
Between pre-amp and 2nd stage there is also a 100 Ohms in +/- power supply.
At the input there is a additionally 3K6 resistor (with shorting switch for measurement)
to charge up the input capacitor when connecting to a LTZ1000.

I've seen also a couple of diodes (they seems to go only on the LT1012).
They are for reverse voltage protection or what?

 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2323
  • Country: de
Hello,

Yes only for reverse voltage protection.
They are in parallel with the battery. (1N4007)
They short the 9V battery in case of wrong polarity.
Usually they should not be necessary.

By the way the full schematic is attached here (as filt1105w.pdf).

https://www.eevblog.com/forum/metrology/ultra-precision-reference-ltz1000/msg834013/#msg834013

With best regards

Andreas
« Last Edit: March 04, 2017, 10:07:19 pm by Andreas »
 

Offline mimmus78

  • Supporter
  • ****
  • Posts: 652
  • Country: it
So I finished assembly of my unit ... I made initial tests and it seems to works.

I used 120 ohm resistor instead of 100 for the first op-amp gain divider (did't find one 100 ohm in
my stock). And 100K on the second op-amp. This should give me 8K ~ of gain until I do not
replace those resistor with the correct value.

I feed a 1uV and 1mV sine wave to the amplifier at 15 Hz and it give me a 3.6K time amplification
factor that seems to me ok.

Unfortunately I cannot attach any LTZ1000 has they don't like to drive this 2mF cap and start
oscillating. The only buffered LTZ1000 don't feet inside metallic box and if I leave it outside the
box I get all sort of 50Hz crap.

I tested also with batteries and considering the 8K multiplication factor noise floor is very low ...

I will build a buffer to put between the preamp and the LTZ1000 ...

Anyway settling time is almost immediate (max 1 minute), but this was a very luky cap. I leaved
it at 7.3V for 10 days and I found it at this same voltage, leakage is almost zero.

Front:


Messy - Back:


1uV RMS - 15Hz


4AA alkaline battery noise

« Last Edit: March 12, 2017, 01:16:25 am by mimmus78 »
 

Offline TiN

  • Super Contributor
  • ***
  • Posts: 3952
  • Country: tw
  • xDevs.com/live - 24/7 lab feed
    • xDevs.com
Connect LTZ output thru 3-5k resistor to precharge preamp, and after it's charged (can take hours), short the resistor without breaking the connection.
YouTube | Chat room | Live-cam | Have documentation to share? Upload here! No size limit, firmware dumps, photos.
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2323
  • Country: de
Unfortunately I cannot attach any LTZ1000 has they don't like to drive this 2mF cap and start
oscillating.

Hello,

if you have C5 (2,2uF) populated this one could be the reason for the oscillation.
It has no real function and is only left there for EMI reasons. -> feel free to remove the 2.2uF.
But if you put all into a shielded box and supply by battery there should be no EMI.

In any case if connecting to a unbuffered LTZ you have to pre-charge the 2200uF by a  >= 3K3 resistor.
Otherwise you will likely change the LTZ output by overheating. (hysteresis).

with best regards

Andreas
« Last Edit: March 12, 2017, 02:38:00 am by Andreas »
 

Offline mimmus78

  • Supporter
  • ****
  • Posts: 652
  • Country: it
So I made another test having care to wait 5 minutes before to apply the jumper (that short the 3.6k resistors at input) and this is what I get. Although noise seems very low 6mV p2p at 8000x are 0.75uV.

I have to check what is real gain at 1Hz, I think this is not 8000x as calculated (or maybe this a very good LTZ1000).

This is a screen shot of what I get now.



PS: what happens when I run out of batteries with this circuit?
« Last Edit: March 12, 2017, 07:22:31 am by mimmus78 »
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2323
  • Country: de
Hello,

noise seems a bit low for a LTZ1000.
I always had something between 1 and 1.4 uV.
(the 1 uV more for quiet devices with 10 s record length and the 1.4uV more for noisier devices with 100 s record length).

What is your sample rate?
The horizontal resolution looks relatively low for me for a 10s diagram.
(those single pixel width spikes look somewhat unnatural for a 10Hz bw signal if you compare it with my analog screen shots).

Could also be too much scope noise. -> Is the 20 MHz BW-limiter on?
Or is there some averaging active on the scope?
Or is it due to the screen resolution?

Running out of batteries is something that I never tried with a LTZ1000 as DUT.

With best regards

Andreas

PS: the more I think about the screen shot the more I think that the lower frequencies (0.1-1 Hz) are missing.
Are all 6.8uF capacitors really connected?
« Last Edit: March 12, 2017, 05:49:03 pm by Andreas »
 

Offline mimmus78

  • Supporter
  • ****
  • Posts: 652
  • Country: it
Hi Andreas as I stated in previous messages I used only one cap of 2.2mF.

After reading your post I measured bandwidth at 0.1 Hz and signal was less than 50% than at 1 Hz. So this can explain measurement with this "low" value cap.

I'm now testing with a 6.4mF and now noise seems a reasonable amount (8.8mV p2p / 8400 = 1.04uV). Also waveform shape seems to not contain too much high frequency noise.

I still have some high frequency noise coming from the amplifier stage (or caps). This noise have a great improvement if I turn off all led light in the office or if I run all on batteries, but I think it's still there and visible in the chart.



As for the scope, bw limit on doesn't make any difference. As sample rate I have to check as in this scope in not shown anywhere (also did a rapid search on the manual but cannot find any paragraph relative to this but they declare up to 1GSPS). No average, no peak detection, just normal sampling. I think this oscilloscope has 640x480 lcd display ... so resolution is just this.

Inviato dal mio Nexus 6P utilizzando Tapatalk
« Last Edit: March 13, 2017, 09:14:55 pm by mimmus78 »
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2323
  • Country: de
Hello,

now the results look plausible to me.

Never thought that the input capacitor has such a large influence.
Did you measure the 2200uF capacity value?

With best regards

Andreas
 

Offline mimmus78

  • Supporter
  • ****
  • Posts: 652
  • Country: it
Andras it was 2.1mF measured at 100Hz.
Anyway you already had -3dB at 0.1Hz with 3mF in your simulation ... not much unexpected.
 

Offline mimmus78

  • Supporter
  • ****
  • Posts: 652
  • Country: it
Ok guys 23 hours passed from last post in this metrology section, time to break the silence ...

So I cleaned the board, tinned some tracks, removed the 2uF caps at input and that's the result. Pretty nice and all this high frequency noise is gone.

Multiplication factor was precisely determined to be 8720 ... so this 12 mV p2p become 1.4uV p2p in 12 seconds span.

Considering the thingy should have some noise by it self it all seems to work now.

Anyway it's not that practical. All the stuff should be put in a panettone metallic box (biscotti box was too small) including reference and all the batteries.

Any suggestions on how to improve it? It would be good if It can at least run on standard linear power supply.
It seems with this 2uF cap noise is better, but as mentioned before plain LTZ1000 circuit don't like it.
What can you suggest to try for improving it?




« Last Edit: March 16, 2017, 08:24:04 am by mimmus78 »
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2323
  • Country: de
Hello,

what are you complaining about?
A set of 9V NiMH batteries will serve you more than a working day for measurements for the amplifier.

Even the Fluke 732B has full specified accuracy only when battery supplied.
So why do you want a extra source of uncertainity?

The only thing is to find a suitable cookies box.
Note that I have a extra metallic box within the cookies box for the amplifier.

You could try a small R/C instead of the 2.2uF to keep some RF noise away from the input of the amplifier.
Perhaps R around 100 Ohms (should keep the LTZ calm, but not increase the current noise too much)
 and C some nF -> far enough away from the upper 10 Hz  bandwidth frequency.

with best regards

Andreas


 

Offline mimmus78

  • Supporter
  • ****
  • Posts: 652
  • Country: it
Hello,

what are you complaining about?
A set of 9V NiMH batteries will serve you more than a working day for measurements for the amplifier.

Even the Fluke 732B has full specified accuracy only when battery supplied.
So why do you want a extra source of uncertainity?

The only thing is to find a suitable cookies box.
Note that I have a extra metallic box within the cookies box for the amplifier.

You could try a small R/C instead of the 2.2uF to keep some RF noise away from the input of the amplifier.
Perhaps R around 100 Ohms (should keep the LTZ calm, but not increase the current noise too much)
 and C some nF -> far enough away from the upper 10 Hz  bandwidth frequency.

with best regards

Andreas
Don't get me wrong Andreas. Your design is fantastic and just the fact that I was able to get those results for me is astonishing and prove how good is it.

What I'm trying to learn now is how to minimise all those susceptibility effects on the stuff I'm dealing now (that's LTZ1000 circuits and this pre-amp).

I know running on batteries is the best technical solution to get the best from this circuits, but I'm also trying to understand how all this stuff influence the thingies and how to gain a certain degree of immunity.



Inviato dal mio Nexus 6P utilizzando Tapatalk

 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 5759
  • Country: de
For the noise meter without the 2 µF to GND, it might be worth adding something to block RF signals. This could be a choke / ferrite bead and maybe 100 pF-1nF to GND. With a 1 nF capacitor one might already want a series resistor in the 50-100 Ohms range to keep sensitive sources (like the LTZ1000 circuit) happy.

I would guess that a lot of spurious signals today are coming from RF signal, like those from mobile phones. So a proper metal case is a good idea.

Another important point is that an OPs output should not be directly connected to an output - it usually needs some kind of output series resistor.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf