EEVblog Electronics Community Forum
Electronics => Metrology => Topic started by: iMo on September 04, 2023, 12:35:57 pm
-
While discussed with Jaromir during our regular summer sailing the south Pacific on my yacht, Jaromir motivated me to build a simple single opamp noise jig for comparing my 399 samples I collected. While he is usually using in his designs 100 opamps in parallel, my poor man's one is with a simple OP27 I found in my junkbox. Basically a Sunday afternoon project.. :D
The "noise indicator" works at first glance, I've named it an "noise indicator" in order not to start a long discussion on such details like amp noise floor, noise bandwidth, filter shape, etc.
The simulation with parts I took off my box shows aprox 0.1-10Hz bandwidth, with 2500 amplification at 1Hz.
Actually a noisy LM399H (I selected as a noisy in past while measuring it with my 34401A) shows around 6.4uVpp noise while observing 10secs run with 10secs time base on my analog o'scope.
Below the schematics and a shot of the prototype I've been playing with.. Powered off a 5x 18650 cells in series.
PS:
- the D2 is the parasitic diode inside the LM399.
- R9 is a protection resistors, normally shorted with a jumper during a measurement, jumper opened when messing with samples (limits the 1mF input capacitor discharge).
- R6 is the internal o'scope impedance.
- the 7815 is powering the 78L05, the output of the 78L05 is the ground.
- the Rheater is for simulation only, the max current is higher when cold.
- the V2 is for simulation only.
PPS: the noise of the LM399 should be roughly
e_399 = sqrt (6.4^2 - 0.24^2) = 6.39uVpp in 0.1 to 10Hz band, imho..
-
Nice little project in a cute Maxim's cookie box :-+
Why did you chose a gain of 2500?
C5 as polarized electrolytic is a bit weird, but fine if the DUT noise is low and cap is placed according to the offset voltage - I am sure you did :popcorn:
Do you have bode and noise plots?
PS: en = 0.24 contributes less than 0.1% to the overall noise (x2 ~10%, x7 ~1%, x20 ~0.1%)
-
You could just DC couple into the OP27 and move the cap onto the feedback loop, and run it off a single supply and not have big discharge current to worry about.
The large 980uF coupling cap and 1k2 AC load added to the reference IC, I'm not sure if it affects it. The 10k will add noise.
Something very interesting I do is connect an audio amplifier and listen to the noise. I have heard WiFi/cellphone birdies and chirps, microphonic parts, demodulated AM radio, outright EMI as well. The unanticipated antenna. I think it's important compared to looking at a scope trace, as far as identifying the noise type first. Then go on and measure amplitude.
-
I would use the 0.1 to 10 Hz test circuit shown in the LT1001 or LT1007 datasheet (1) for better comparability to other noise measurements. It should be scaled to the OP27 input noise resistance, but the LM399 has much higher noise so this is only a minor issue.
(1) Analog Devices uses the same circuit. Texas Instruments does not show their test circuit, but they lie in their datasheets anyway, so ignore them.
-
This has been a Sunday's afternoon project, and frankly, I am pretty happy I've built this tool finally, as the watching the slowly dancing turquoise colored beam in a dark room is hypnotizing :)
The 2500 gain - rather intuitive decision, as the 2k7 cathode resistor is attenuating the 7815 noise aprox 2500x (considering aprox 1ohm 399's dyn resistance), and also I wanted to avoid potential instabilities with higher amplifications in a single stage with that construction. It works rock stable even with the box opened, a nice clean signal. I may try with larger amplification later on. My crt vintage scope has got 2mV/5mV/div and the signals with the "good" chips fits into 16mV and with "bad" ones into 40mV full screen with that amplification nicely.
So far I've seen two really bad chips - one MAC199 creating a nice telegraph signal with ~10-15mV jumps between three levels, and, a new 1399 doing almost the same (here I do not want to judge as the wiring is lacking the compensation RC for the 1399 and also the current is not set for 3mA). You can identify a "bad" chip sending you a Morse coded message even on the 100mV/div scale easily.
Otherwise the LM399s and good MAC199s behave all almost the same, some produce short popcorn shots which are clearly visible as well.
PS: and yes, the 10k protection resistor (which gets shorted with the jumper during the measurement) there is not an optimal solution, as you have to mess with the jumper and usually you do it wrong way.. :)
PPS: added more info into the schematics
PPPS: the ADR1399 does not work properly at 1mA and without the snubber, it seems..
-
You could just DC couple into the OP27 and move the cap onto the feedback loop, and run it off a single supply and not have big discharge current to worry about.
The large 980uF coupling cap and 1k2 AC load added to the reference IC, I'm not sure if it affects it. The 10k will add noise.
Havinng the AC couling only in the feedback loop is not that practical, as the capacitor would have to be hughe, setting the lower frequency limit together with the small (e.g. 22 ohm in this example) resistor of the FB divider. The protection diodes between the OP-amps input would still allow for a charging spike when connecting.
-
FYI - these values below should provide you with the gains 5000 and 10000 in 0.1..10Hz (-3dB).
With noise the "amp gains" have not be extremely precise as people usually mess with the noise in log-log scale..
-
You could try a pair of diodes across the positive opamp input as an alternative method for power-up protection. Diodes and cap will delay reference, yet only for a short time.
Regards, Dieter
-
You could just DC couple into the OP27 and move the cap onto the feedback loop, and run it off a single supply and not have big discharge current to worry about.
The large 980uF coupling cap and 1k2 AC load added to the reference IC, I'm not sure if it affects it. The 10k will add noise.
Havinng the AC couling only in the feedback loop is not that practical, as the capacitor would have to be hughe, setting the lower frequency limit together with the small (e.g. 22 ohm in this example) resistor of the FB divider. The protection diodes between the OP-amps input would still allow for a charging spike when connecting.
I thought getting rid of the capacitor, maybe going to two op-amp stages for the bandpass filtering, would be better, from the old app note:
Ref. "775 Nanovolt Noise Measurement for A Low Noise Voltage Reference, Jim Williams" AN124 (https://www.analog.com/media/en/technical-documentation/application-notes/an124f.pdf) or EDN colour version (https://www.edn.com/wp-content/uploads/9.3.09_MS_Williams_Characterizing_noise_voltage_reference_ICs.pdf)
He hand-selected for <5nA leakage 1,300uF 30V wet slug tantalum Vishay XTV138M030P0A (https://www.vishay.com/docs/42077/xthklmv.pdf) (discontinued was $400 last going for $721).
"... There may be a more palatable alternative. Selected commercial grade aluminum electrolytics can approach the required DC leakage although their aperiodic noise bursts (mechanism not understood; reader comments invited) are a concern."
I have no idea, have not checked noise generation in electrolytic capacitors, or what OP is using.
-
Ok, I've increased the gain to 10000 as in the above sim, also I've increased the input RC resistor to 1k5. Looks stable :phew:
The better 399 shows ~60mVpp, the Morse one ~120-140mVpp.
-
Ideally the protection should work both ways: protect the OP-amp and protect the DUT from a sudden load spike (e.g. LTZ1000 circuits may not like the current spike and resulting drop in the reference).
A way that could work is a FET current limit to some 1-5 mA, e.g with depl. MOSFETs or back to back JFETs. It may still add some 200-1000 ohm of series impedance and corresponding noise.
-
..I have no idea, have not checked noise generation in electrolytic capacitors, or what OP is using.
I've been using an aluminum elyt 1000uF/35V ("LXV", 105C) I have on stock. Took 2 of them and let them for a night at 18V. I took one and wired it in..
The residual DC offset at OP27's output after 40 minutes is +0.89V (with gain=10000, input resistor=1k5).
The residual DC offset at ADA4523's output after 40 minutes is +0.16V (with gain=10000, input resistor=1k5).
-
It's the little known electrolytic cap (current) noise, and I would trust Jim Williams going to great length to avoid it. But who has got $500 tantalum capacitors lying around to compare?
Input impedance is the same for both circuits at 1.2kΩ
As long as you know it's the ref to blame.
-
This "noise indicator" tool will show you what a 399 is doing within a couple of minutes. All you need is an o'scope with 10secs time base. With a DSO you may do the min max math easily.
Unless you own a good LTZ/LTFLU based DMM, or such a reference with a KVD w/ a null meter, you will not get a better information about the DUT zener, imho.. (except its voltage stability)..
On the first glance the tool indicates a "smooth", "telegraph", "popcorn" 399 styles, or a mixture of all those on the screen easily.
The actual indications I see currently with the gain=10k are ~50-60mVpp (5-6uVpp) with my "good zeners", and up to ~140mVpp (14uVpp) with the "telegraphing zener", taken as the diff between max and min during the 10secs long beam run.
PS: Interestingly my "Morse zener" shows quiet periods as well - when it behaves "normally" (like for several minutes long). Then it starts with sending its encrypted message again.. :scared:
-
A forward biased:
green led ~5uVpp
1N4148 ~1.5uVpp
PS: w/ ADA4523
-
A forward biased:
green led ~5uVpp
1N4148 ~1.5uVpp
PS: w/ ADA4523
Make sure the LED, and a glass diode like the 1N4148, are shielded to keep ambient light out.
-
OK, I've found a DS-1062CA in the junkbox of a friendly HAM :) ..
Below some shots. My previous measurements indications were pretty conservative, however..
Gain 10k.
-
And more..
-
Mhm,
the 1N4148 at which current? (= data sheet current of 10 mA?)
(I thought I could make a low noise adjustable voltage source with a string of 1N4148)
with best regards
Andreas
-
@Andreas: The 1N4148 simply put into the "noise indicator's" LM399 socket :)
Therefore (10V-0.669V)/2k7=3.45mA
PS: added a jumper for 1mA/3mA selection, and added the RC snubber (4.7ohm ser 1uF) across the zener.
Below the ADR1399 (a fresh one, none burn-in), the winner for today :)
PS: the Rigol has got 12secs time base, thus you have to adjust the values accordingly.. ::)
-
.. and finally the V2 with some info - see below..
PS: ..also would be better to use an LDO, lower noise, low cost 15V/100mA regulator with 4x 18650 and a schottky instead of the 1N4002.. Any hint?
-
What really bugs me is the fact that you are using a 7805 to Generate your GND at the output.
I would have used a (TL)431 set to 5V and a biasing resistor giving the 431 a handfull of mA to work with.
If it works for you, good to hear, but I havent found any information on the current sink capability of the 7805.
-
Yep, that is a good question I waited for.. I had there a zener, I was thinking on the 431 as well, then I tried with the 78L05 I had laying on the bench, and it works, it holds the voltage well.. :D
As I've written above - the power source is not optimal in my prototype, better to use a lower noise low drop 15V/100mA regler with schottky diode for a 15.5-16.8V battery input, with perhaps the 431 set somewhere around 5V..
PS: I will try with the 78L15 (still suboptimal) and the 431 and I will report soon..
-
PS: ..also would be better to use an LDO, lower noise, low cost 15V/100mA regulator with 4x 18650 and a schottky instead of the 1N4002.. Any hint?
Hello,
I generally use LT1763 (adjustable version).
Low noise, not really low cost nowadays.
But it has maximum input voltage of 20V.
My LNA is powered by 2*9V Blocks (independant of DUT).
with best regards
Andreas
-
It could be the current sinking at the 78L05 output "works" somehow thanks to the rather large 1.9ohm shunt at the output's current limiter - see the National DS schematics below.. Experts may comment..
PS: with 3mA current through the zener the 5.00 voltage increases to 5.04V..
-
Below the V3, 78L05 replaced by the TL431 set to 5V.
No instability even with the 1000uF at its output (here one has to be careful with the 431), voltage 5.00V, no change in 3 digits when sinking 20mA.
ADR1399 still shows 2.2-2.6uVpp..
PS: finally replaced the 7815 by 78L15.. Works fine.
-
Now, I want to get an indication of the noise level and noise type upon a set of my 399/1399s..
What would be the recommended process here ?
For example:
Procedure 1.
1. place a new sample and close the box
2. wait 20 minutes until it settles
3. make a single shot indication (10secs, peak detect mode, etc.) and record the mVpp/10 = sample_uVpp
4. goto 1.
Or
Procedure 2.
1. place a new sample and close the box
2. wait 20 minutes until it settles
3. make, say, 10 single shot indications at random times (10secs each, peak detect mode, etc.) and make the average_mVpp/10 = sample_uVpp
4. goto 1.
Or
Procedure 3.
1. place a new sample and close the box
2. wait 20 minutes until it settles
3. make, say, 10 single shot indications at random times (10secs each, peak detect mode, etc.) and write down the worst (or best, or both) mVpp/10 = sample_ uVpp
4. goto 1.
or..
Any recommendation before I start to mess with my indications ? :)
-
The best results are probably from the 2nd procedure, so the average of the peak to peak values. Ideally one would remove something like the best and worst of the set before taking the average to avoid rare excursions, though this is somewhat debatable.
Anyway the noise is usually more complex than just the 0.1 Hz to 10 Hz band. This band is more chosen because it is reasonable easy to measure (e.g. such an amplifier), not so much because it is relevant for real world use. For a DMM or so it is more the noise for lower frequencies that is relevant and audio is obviously higher.
-
The best results are probably from the 2nd procedure, so the average of the peak to peak values. Ideally one would remove something like the best and worst of the set before taking the average to avoid rare excursions, though this is somewhat debatable.
Anyway the noise is usually more complex than just the 0.1 Hz to 10 Hz band. This band is more chosen because it is reasonable easy to measure (e.g. such an amplifier), not so much because it is relevant for real world use. For a DMM or so it is more the noise for lower frequencies that is relevant and audio is obviously higher.
The 2nd procedure introduces a kind of addtional filtering, imho.
My naive preference would be 3. with recording the best and worst for a given sample.. But not sure, of course.
I wonder how the big boys are doing it in their datasheets..
-
Hello iMo,
what I am doing:
1. place a new sample + cover with a micro fibre chloth (avoid local air movements)
2. wait some time until it settles (also the amplifier).
3. make 15-19 successive single shots (minimum 10 secs each, but mostly 100 secs per shot)
settings: 20 MHz hardware BW limitation, minimum 50 kSps/sec, 1 kHz low pass software filtering to filter out oscilloscope noise (it is far outside 10 Hz).
4. build average and standard deviation of uVpp and uVrms values
5. also check for "popcorn" events in the records. (here zoomed in record 16 of 19)
Of course the 100 sec measurements deliver somewhat "higher" averages than 10 sec measurements.
with best regards
Andreas
-
I wonder how the big boys are doing it in their datasheets..
There are application notes:
(I think it was LT) with following procedure:
make several (hundreds) single measurements of 10 seconds.
sort the results by peak values.
take the value where half of the results are above and below the final result (so effective the "median" value).
with best regards
Andreas
-
My Rigol's DS-1062CA (I would call it an "electrician scope" :) ) setting is currently:
DC, Acquisition: peak-detect mode (my understanding is it emphasis showing peak values), sinx/x OFF, Sampling rate - no idea, BW limit ON, time base 1sec/div (it got 12secs over the display), printing Vpp on the display in addition.
The median over hundreds of measurements is a pretty heavy low-pass filtering.. Almost any sample after such filtering will be better than the competitor's best :)
-
But it is only a "typical" value (see e.g. Data sheet of LTC6655 for the procedure)
with best regards
Andreas
-
Ok, interesting method with the 6655..
-
After cleaning up the boards with IPA, final assembly and installing foam insulation (from the bottom of the pcb and a cap for the samples) here are the first data with the brief notes..
PS: added a shot of the foam insulation used - there is a small block of foam beneath the pcb as well.
-
And the results after a weekend with the noisy refs..
For example:
..the LM399H will have equal or less than 4 to 5.5uVpp noise in 80% of 10secs intervals..
..the ADR1399 will have equal or less than 2 to 3uVpp noise in 80% of 10secs intervals..
See may above table - the top green one is the number..
::)
Some notes:
- almost all samples do "pops" from time to time (one 399AH does nice ones)
- no Morse telegraphing this time - the RC snubber did help obviously with the "Morse MAC199" and also with ADR1399 - thus my previous observation was an instability perhaps, imho
- MAC199 - all were shortly produced around 1990 by TESLA afaik (TESLA corp. died after 1990), almost unseen (the MAB399 was/is more less known), my measurements from past showed large drift (none snubber at that time, btw), and one of the samples above has broken heater. Interestingly all show 0.1-10Hz noise comparable with the LM399.. :o
For a calibration the 1N4148 is suitable (~0.7uVpp here), with an empty socket you will see the noise of the voltage regulator(s), here I see around 350-400mVpp that is 35-40uVpp (and multiply it by 2k7/1k5 divider, say 3x).
-
FYI - a DIY "ref diode" made of a 6.2V BSX zener and an 1N4148 diode in series.
The noise seen is typically between 1.8 and 2.5uVpp at 3.6mA.
PS: added a 10 minutes long run - 3uVpp
-
..based on a discussion in a different thread I've made some minor improvements which helped with filtering - see below the sim. I also added a small ferrite bead (like 3 turns in a 4mm*10mm ferrite bead or a small toroid, do not use axial chokes) - it should suppress RF noise propagating from the scope towards the inside of your cookie box..
PS: the 10uF output ceramic MLCC has got aprox 8mm long thin wire leads (such you will not get a loud music into your spectra)..
PPS: the same with 1uF MLCC in the snubber..
-
And a simulation with a pink noise file I found on the web (aprox 95MB wav file).
Looks pretty similar to the oscilloscope waveform above.
The amplitude set such it mimics my ADR1399#2 with the ACRMS (aprox 400nVrms).
The FFT (1MPts no aver, Blackman-Harris window) looks differently as we do not have log freq in the scope..
At 200Hz the sim shows aprox -116dB, the scope aprox -117dB..
PS: added a measurment of an 1N4148 at 3.5mA If.
The noise is aprox 82nVrms - perhaps a half of a good LTZ1000..
The 50Hz is now visible again at aprox -101dBV (so even the cookie box of 1.5m distance from closest active transformer is not enough).
PS: that -101dBV of 50Hz is measured at the LNA's output (LNA's gain is 60dB at 50Hz).
-
I recommend always connecting an audio amp and speaker and listening to the noise.
You can hear beeps, boops and other issues from RF getting rectified/demodulated due to WiFi, cellphone, AM radio. Also there is component microphonics, AC hum, other interference as well that can be found compared to looking at scope waveforms.
A cheapo LM386 board and speaker, battery powered does OK. Voltage gain max is 200 for that so you might need more gain than that, at the reference IC.
The LM386 is also fairly noisy and likes to pick up AM radio on its own though so make sure that does not mislead you.
-
This is with TheoB's python based "DHO-remote utility v1.4", as-is except input data divided by A= 10000.
20sec/div, Blackman, 1Mpts, no FFT averaging (needs to be added).
1N4148#2 @3.5mA quick test, not fully settled..
The noise folding message ??
Sample rate=0.005 Ms
FFT max freq=0.0025MHz
FFT min freq=0.005Hz
FFT bin noise BW=-22.8db
Noise is not accurate due to noise folding, please increase sample rate!
Vrms=9.6e-08 V
-
100sec/div, Blackman, 10Mpts, no FFT averaging.
LM399#2
PS: the transfer of the 10Mpts (bin) via LAN takes aprox 3secs..
FFT max freq=0.005MHz
FFT min freq=0.001Hz
FFT bin noise BW=-29.8db
Noise is not accurate due to noise folding, please increase sample rate!
Vrms=4.86e-07 V