Noise cancellation circuits for voltage references

[branadic]

Hi everyone,

we have voltage references and we have low noise amplifiers to characterize their noise, mostly multistage inverting topologies, but has someone ever tried to combine both with a sum amp to create a Noise Cancellation Circuit? Even the best voltage reference we have today is not limited by the opamps used, but the reference itself. On the other hand we have high-performance chopper amps for the sum amp too. So at least in theory it could be possible to achieve lower noise at the output, right?

I'm happy to hear about any approach tested by now.

Just in case someone argues, LNA's with frequencies down to 4 mHz were alread published, so we could benefit from that very low frequencies. On the other hand I found, while testing multiple different brands and caps for an LNA that Yageo 85 °C caps showed <5 nA of leakage after 24 h, while 105 °C suffered from way larger leakage in general.

Edit: Added a first example to give a brief idea how such NCC could look like and as better basis for discussion.

-branadic-

[iMo]

What would be the propagation delay through the LNA?

[dietert1]

There is a difference: Standard usage of a LNA is to look at the AC output, while the DC precision isn't that important. With a cancellation circuit you will have the DC output in the result.
Then you arrive at the same problem as those engineers trying to make a LP filter for noise suppression: Leakage, dielectric absorption and temperature dependence of the capacitor(s).

Regards, Dieter

[Kleinstein]

The main part of the LNA is the AC coupling as a kind if high pass filter. So the use for noise reduction is using a low pass filter.

I have a low pass fitler for the reference in my ADC boards (some 5 K and 4.7 µF). It does help noticible with the noise of the LM399 reference, though part of the effect in my case is from the 50 kHz range that would be filtered with less capacitance too. The dielectric absorption gives a little longer settling time on turn on, but not very much - at least not with a polyester cap. It could be an issue for using electrolytic capacitors. It is more the normal warm up that effects the settling.
Similar the temperature effect on the capacitor should not be that bad, except for the electrolytic ones.
This does not mean a filter with electrolytic capacitor would be bad, but it takes extra care in avoiding temperature variations and mechanical stress and finding a suiteable. It can slow down the settling quite a bit.
With many references we don't really care about the high frequency noise, the troublesome part is the low frequency part. With some ADCs/DMMs  there can be an extra sensitivity to noise at frequencies around 25 Hz or maybe 2.5 Hz from the AZ cycle. So it may be worth filtering for this frequency range. This is the motivation in my case.

[dietert1]

All pass minus high pass gives a low pass. This is used in digital loudspeaker crossovers. The all pass includes a delay line in order to get a low pass filter of good steepness. Without the delay the resultant low pass has only 6 dB/oct.
I think the idea was to make a low pass with a corner frequency of 0.1 Hz using one of the standard 0.1 Hz to 10 Hz LNAs. This will be considerably more difficult than a 13 Hz low pass.
Recently i happened to use some 1uF/100V film caps we had for filtering a ADR1399 with 16 Hz corner frequency. I got a warmup time of several hours due to DA and an extra TC of 0.4 ppm/K due to a 4 ppm temperature dependent leakage loss. Then i found a 2.2uF/400V cap that behaves much better. It's an old part from a CRT display horizontal deflection.

Regards, Dieter

[MasterT]

HPF is 1-st order.  I think, that 2-nd order Sallen Key LPF build around ada4523 would be more efficient

[n_haku]

Some time ago, I run into clever idea of reducing voltage regulator' noise by introduce a resistor in series on whitch noise current can be short to ground with transconductance amplifier. Probably, this idea also will work for ref buffer, especially if provide feedback after that resistor to dc accuracy.

[Kleinstein]

Some time ago, I run into clever idea of reducing voltage regulator' noise by introduce a resistor in series on whitch noise current can be short to ground with transconductance amplifier. Probably, this idea also will work for ref buffer, especially if provide feedback after that resistor to dc accuracy.

It is a nice idea for higher freuquency noise of a voltage regulator, but I am afraid it would not work for a reference:
The main problem is the DC accuracy. The series resistor and noise compensation current would ruin the DC accuracy, unless one would use a large capacitor to couple the correction current. Than one is essentially back at a more classical low pass filter.

[David Hess]

Instrumentation and difference amplifiers use this idea to AC couple a differential input without error prone differential AC coupling which would compromise common mode rejection in the transition band.  The single ended output is integrated and fed back into the reference input.

I do not see any advantage when the same components being used for the high pass filter could be used for the low pass filter.  Leakage is still creating low frequency noise, so the capacitor should be bootstrapped for zero volts across it reducing the leakage.

There is a DC accurate alternative where the high pass filter is AC coupled at its input *and* output which removes its contribution to DC offset, but this requires two large filtering capacitors.  Old digital multimeters used this for their DC accurate low pass filtering.

[openloop]

There's a known trick to get rid of leakage and absorption - stacked capacitors. Basically it's two RC lowpass filters one sitting on top of the other with joined input, output from the top one. Thus DC voltage across the top capacitor is essentially zero. So nothing to leak and nothing to absorb.

Thermal things are still a problem though. Mostly from the bottom capacitor.
I think smearing it with thermal grease and sticking into a tight-fitting block of aluminum should slow it down enough.

[Kleinstein]

The filter configuration used for the DMM input or in the Fluke 57xx calibrators needs at least 2 large capactors, but both are used as a 2nd order filter. So this is no extra parts needed just starting with a 2nd order filter and the capacitors are fully used.

The filter part (without active amplifier) is rather low power. So there is little need for extra thermal past or similar. Just some thermal insulation and may be additional thermal ballast would be enough to get a reasonable stable temperature. Temperature variations slower than the filter transition frequency would not matter that much. It would be mainly about filtering the faster thermal fluctuations.

Instead of using a lot of effort on a filter, there is also the option to use a second refrence and average.

If reference noise is a problem, one would in most case record data over a longer time and average - this way the DMM is doing much of the filtering. The integrating type conversion filters out most the higher frequencies above some 50/60 Hz and avearging over multiple readings filters from one over twiche the avearging time to half the reading rate. So much of the reference noise would be filtered out by the DMM. There can be a small frequency band from the auto zero mode and only integrating the input half of the time (most DMMs with multislope ADC) for reference noise passing. That would be something around 2.5 Hz for 10 PLC mode or 25 Hz for 1 PLC conversion mode. The analog filter could make some sense for that frequency window. Especially when using 1 PLC mode the cross over does not have to that low. The rest of the filtering will be done by the DMM digitally. So I see relatively little need to get a filter cross over of much below 1 Hz.

[tggzzz]

Wouldn't help much with any of a reference's popcorn noise

[macaba]

I have tried simulating a few different non-conventional filters, and the main problem comes down to exactly what Dieter said - "Leakage, dielectric absorption and temperature dependence of the capacitor(s)."

So any complex circuit (more complex than passive RC) needs to address this issue (IMO).

I attach 2 circuit ideas here as food for thought. Both of them will filter, but both of them are not good enough (IMO) for DC precision (i.e. stability over temperature) that would allow for voltage reference filtering.

Circuit 1: Directly compensating for the leakage with an error amplifier
Circuit 2: Reducing the voltage drop over the critical capacitor(s) to nearly 0V.

[MasterT]

Using supercapacitors to filter LF noise is not a panacea, 66uA leakage current is roughly translates into 66 mV (!) of noise over 1 kOhm resistor.
 I recently discovered that Nichicon has UKL - Low Leakage Current series product. About 10 times lower current compare to normal

[David Hess]

Filtering is useful for midband noise, and high frequency noise is easy to filter out, but it may be better to use multiple references in parallel to reduce low frequency and flicker noise.

[Kleinstein]

Using supercapacitors to filter LF noise is not a panacea, 66uA leakage current is roughly translates into 66 mV (!) of noise over 1 kOhm resistor.
 I recently discovered that Nichicon has UKL - Low Leakage Current series product. About 10 times lower current compare to normal

Leakage current times series resistor does not directly translate to noise. Primarily this gives an offset, which for a reference is even worse.
The leakage current has however a good chance to be also accompanied with noise as the leakage current may have instabilities (current noise) and possibly shot noise.

The circuit from macaba looks a bit suspicious: the extra correction amplifier should also effect the cross over frequency and may effectively reduce the resistor by the gain.

[macaba]

The circuit from macaba looks a bit suspicious: the extra correction amplifier should also effect the cross over frequency and may effectively reduce the resistor by the gain.

https://doi.org/10.1063/1.4870248 has more information.

[macaba]

Circuit 3: Error sensing

Using the error sense output; adjust the input DAC value within very slow digital control loop (slower than the RC constant) to compensate.

[dietert1]

One could make R2 somewhat bigger to amplify the error signal and simplify digitizing it. You get a noise measurement for free.
Probably i will try a 7 to 10V gain stage that is a Sallen-Key low pass filter at the same time, like shown in PDF.
In a drawer i found Vishay MKP1848 40 uF/900V and Kemet C4DE 380uF/400V capacitors. Both tested with less than 50 pA leakage at 20V. MKP power caps seem to be a reasonable choice for this application.

Regards, DIeter

[miro123]

I have been playing with Macaba Circuit #1. While it filter high and mid-low frequency it introduces utralow frequency that creates even bigger headache.

[MasterT]

I have been playing with Macaba Circuit #1. While it filter high and mid-low frequency it introduces utralow frequency that creates even bigger headache.

My experience similar. Circuits is dated back to 1990,
https://www.ti.com/lit/an/sbva002/sbva002.pdf?ts=1669056577129&ref_url=https%253A%252F%252Fwww.google.com%252F
I observed a "bump" on the spectrum around cut-off frequency

Using supercapacitors to filter LF noise is not a panacea, 66uA leakage current is roughly translates into 66 mV (!) of noise over 1 kOhm resistor.
 I recently discovered that Nichicon has UKL - Low Leakage Current series product. About 10 times lower current compare to normal

Leakage current times series resistor does not directly translate to noise. Primarily this gives an offset, which for a reference is even worse.
The leakage current has however a good chance to be also accompanied with noise as the leakage current may have instabilities (current noise) and possibly shot noise.

Not directly, I say roughly. The point is, if AC component of the leakage current is only 1% - flicker noise, than it still 0.66 mVrms - about x30 times higher than noise level presented from voltage reference itself, 20uVrms or so.
 It is good practise to assume that AC may reach 100%

[KT88]

The challenge with very low frequency (analog-) filters is capacitor leakage obviously.
I played a bit with leakage mitigation strategies. Here is an approach with a simple Sallen-Key filter extended with a little servo that keeps the DC voltage across C1 at near zero. Note that the servo amplifier itself needs a second buffer to mitigate leakage through it's capacitors. I haven't completely optimized the frequecy response but it demonstrates the principle.
I chose Rp values from an AVX data sheet to make it realistic. Of course other effects like temperature, mechanical stress is not modled.
However the servo would cancel slow changes of Rp of C9, C6, C5 and C3. C1 and C2 are not exposed to any DC voltage more that a few uVolts...

Cheers

Andreas

[maat]

Then you arrive at the same problem as those engineers trying to make a LP filter for noise suppression: Leakage, dielectric absorption and temperature dependence of the capacitor(s).

I can only second that. I have been there and tried this and that until I finally gave up, did the maths and realized, that one either relies on the stability of some resistor (or capacitor) ratios or the stability of a filtering capacitor, which is a fallacy, but let me explain.

While the white noise component of a reference can be filtered fairly well, once you get into the 1/f territory you will hit a brick wall. To understand this, one needs to look at the spectral properties of 1/f noise. i have attached a little simulation, that shows the 1/f noise in the time domain, its PSD and the Allan deviation (I will get to that latter).

The time domain plot and the PSD are likely the plots people are most familiar with. The PSD already shows what it happening. As you go to lower frequencies, the noise power increases linearly with f (no sh** Sherlock). At the same time (pun intended) the settling time of any filter goes with 1/f as well.  So you will always have the same noise power at your output. Simply speaking, the filter will never settle. This brings me to the adev plot. I beautifully sums it all up. No matter how long I wait, the deviation will be the same.

Filtering at frequencies below the corner frequency makes matters worse. You will have have the same noise power and on top that, there is no filter apart from maths, that has the theoretical transfer function, i.e. is flat from DC to the corner frequency and then rolls off. Every filter introduces its own 1/f noise a low frequencies. So the harder you try, the worse you will make it. This is simple physics or maths.

So is there a way to cheat fate? Yes and no. 1/f noise has an Achilles' heel. It is correlated. So in other words, it does give you some information about its future. This can be used by a predictive filter like a Kalman filter (https://en.wikipedia.org/wiki/Kalman_filter) to suppress is. The catch? - Its all digital.

The only analogue thing you can do, is to sum several references. This will get the noise floor down by 1/sqrt(N), if the noise is uncorrelated. This is important to keep in mind, because as you start summing references you will invariably introduce correlation (like thermal EMF on the board, or supply ripple, etc.), so you will hit a dead end there as well. Life is a bitch...

Basically, the only way to get rid of 1/f noise is to not introduce it in the first place. If you want low noise pick the lowest noise Zeners (by hand) and sum like 4 of them. Then call it a day. It's all mother nature will give you. Ever!

Addendum: The stuff one adds usually has some temperature dependence, like caps, which results in a random-walk behaviour. A random walk is 1/f². So things will get out of hand pretty quickly.

[KT88]

On the issue of settling time it is possible to use smaller resitors to charge the caps and let the DA settle (for still a long time though). These smaller resistors could be switched on and off with analog switches or relays.
Marcoreps used large higher voltage electrolytics in his LNA. A lot of them may give good (enough) results with leakage mitigation...

[maat]

On the issue of settling time it is possible to use smaller resitors to charge the caps and let the DA settle (for still a long time though). These smaller resistors could be switched on and off with analog switches or relays.
Marcoreps used large higher voltage electrolytics in his LNA. A lot of them may give good (enough) results with leakage mitigation...

Unfortunately, no.
Let's keep it simple and neglect pesky physical properties for a second. Assume you can sample the voltage perfectly and transfer that to the ideal capacitor of the filter. Now this "measured" value will have an uncertainty. If you would like to improve upon that uncertainty by filtering (or one may say integrating over) the source, this is where the trouble starts. This is what the adev plot is showing, that for all integration periods, the uncertainty will stay the same. You will never be the wiser. The Allan deviation does not make an assumption over the starting point of integration phase. So no matter how "well" defined your starting point may be, its uncertainty will not improve, no matter what type of filter or cut-off frequency you use. Digital or analogue - it won't help you.

There is no cure to 1/f noise.

[KT88]

@ maat: My comments and circuit were only about a (at least in LTSpice) working lowpass with leakage mitigation.
I agree on your statement about the reference source itself - it wont give more accuracy unfortunately.
If I'm not completely wrong, an increased precision should be possible (assuming the filter works as simulated).
Precision would mean in this case that a better short term stability would be achieved. This could help with some kinds of measurement like noise analysis or transfer.

[macaba]

With the 1/f corner of LTZ1000 being around 0.3Hz, there is still a practical engineering challenge to be had with achieving a DC stable (vs. temperature) 0.3Hz LPF?

[maat]

If I'm not completely wrong, an increased precision should be possible (assuming the filter works as simulated).
Precision would mean in this case that a better short term stability would be achieved. This could help with some kinds of measurement like noise analysis or transfer.

Indeed, that is quite true, although the latter could also be achieved using digital prost-processing (assuming the transfer standard does not have a similar 1/f component). It is fairly straight forward. Noise analysis is a little more tricky due to the convolution involved and the assumption, that the DUT is most likely not showing white noise.

With the 1/f corner of LTZ1000 being around 0.3Hz, there is still a practical engineering challenge to be had with achieving a DC stable (vs. temperature) 0.3Hz LPF?

Do not forget, that the 1/f corner frequency moves up proportionally to the white noise component. The formula is f_c = h_{-1}/h_0. The h_α is the power law coefficient as defined in [1]. h_0 is the coefficient of white noise. h_{-1} the coefficient of flicker noise.
So if you push down the white noise spectral density (in V^2/Hz) by a factor of 10, the corner frequency moves upwards by a factor of 10. The LTZ1000 is spec'ed at something around 50 nV/sqrt(Hz), so lets assume a filter should get this down to 1.5 nV/sqrt(Hz), this would mean that f_c moves to 0.3 Hz * (50 nV/sqrt(Hz)/1.5 nV/sqrt(Hz))^2 ≈ 300 Hz. This is way more modest...

[1] J. A. Barnes et al., "Characterization of Frequency Stability," in IEEE Transactions on Instrumentation and Measurement, vol. IM-20, no. 2, pp. 105-120, May 1971, doi: 10.1109/TIM.1971.5570702.

[dietert1]

1/f noise is just a model like a gaussian error distribution. Technical noise doesn't have those infinite tails.
I suspect that quite often people use the term "1/f noise" as an excuse for deficiencies of their setups and they don't really know what is noise and what are processes caused by ambient conditions, like EMI. The noise level may be much smaller.
When i look at the daily average difference of two voltage references and i see seven consecutive days with 0.01 ppm agreement, i suspect the setup suffers from ambient processes, not from noise. I think it's worth looking at measurements.

Regards, Dieter

[Kleinstein]

A major part of the low frequency noise is popcorn type noise with more or less discrete jumps. This is especially true for the LM399 reference.  This noise part should not go up idefinitely to lower frequencies, but more like have a maximum at the typical rate of the jumps and than less noies at even lower frequencies. This would still be rather low and had to get with an analog filter. Ideally one would also do more than simple linear filtering, but more like identify the upper and lower levels and use a fixed mixture as the reference value, not the ratio as is comes in in that patricular timeframe.

On the issue of settling time it is possible to use smaller resitors to charge the caps and let the DA settle (for still a long time though). These smaller resistors could be switched on and off with analog switches or relays.

For the settling of the DA it does not really matter if the filter is switched to a different mode. The problem are the long internal time constants that than produce an appearend leakage current. The main idea for the LNAs is to keep the input capacitor charged to about the right voltage also when the actual amplifier is not used, maybe a week before use.

One way to reduce the effect of DA is to use low loss capacitors, so ideally PP dielectric, though they are relatively large in size and small in capacity, so that leakage/bias from the amplifiers can become an issue. Motor capacitors are available with some 50 µF for a still moderate price - just a bit bulky. Polyster capacitors may also be OK, though they need about 10 times longer for the DA to settle. The settling time should be about the filter time constant times the Capacitor DA values (with it uncertain way of measurement) divided be the requited accuracy. So for an accuracy target in the ppm range and DA of some 300 ppm for PP capacitors with would be a few 100 time constants waiting time.

The circuits that keep the main filter capacitor at  near zero voltage can reduce the effect of DA due to the smaller and relatively constant voltage. However there is still the settling of the capacitors used for the voltage at the foot point. DA of this 2nd capacitor gives drift in the voltage there and this couples through the main filter cap to the signal. The drif part is still attenuated by the comparable range cross over frequency. So the drift compensation circuit also helps with the DA settling.

[KT88]

On the issue of settling time it is possible to use smaller resitors to charge the caps and let the DA settle (for still a long time though). These smaller resistors could be switched on and off with analog switches or relays.

For the settling of the DA it does not really matter if the filter is switched to a different mode. The problem are the long internal time constants that than produce an appearend leakage current. The main idea for the LNAs is to keep the input capacitor charged to about the right voltage also when the actual amplifier is not used, maybe a week before use.

You missed one important point of the circuit I provided: leakage current caused by DA is provided by the servo after the current doesn't saturate the servo output anymore.
C1 and C2 won't be charged at all and thus won't cause any leakage even by DA...
The faster C3, C5, C6 and C9 are charged the faster the circuit will be settled.

[macaba]

KT88 - nice circuit idea. Perhaps you could add DA-emulation network, in the simulation, on each capacitor?

ADA4523 has a 1/f corner of around 20mHz (actual measurement, not hypothesis) so not much point in having a filter cutoff below that. ADA4523 is still a good choice though, the two other devices that are better have their own issues (ADA4528 max supply too small, OPA189 has anecdotal reports of poor behavior on inputs).

[DavidKo]

Similar approach is used in AN177 from Renesas (page 30). They subtract voltage of low noise reference from DUT. Maybe such an approach can be combined with battery source instead of reference and use pot to tune the voltage to get the necessary voltage to subtract (I have thought about PWM tuning of battery voltage, but I do not know how much noise will be there from PWM modulation - frequency should be >>10Hz, easy filterable, probably will avery in <10Hz).

To get subtraction working man do need low noise reference which can be subtracted from the source. Long term stability will not be so much important. To use the same reference, but inverted you need to shift the signal between each another for a several "periods" before summing to remove the DC signal.

I have thought about using chopper (switch between positive and negative value), make the sum of positive and negative value, you will get the average signal between two following parts, but the signal will be different to noise itself, but for the characterizing it can make sense. Such an approach can induce the chopper noise, period should be short enough compared to the signal change, but not too short to get reasonable signal.

[maat]

A major part of the low frequency noise is popcorn type noise with more or less discrete jumps. This is especially true for the LM399 reference.  This noise part should not go up idefinitely to lower frequencies, but more like have a maximum at the typical rate of the jumps and than less noies at even lower frequencies. This would still be rather low and had to get with an analog filter. Ideally one would also do more than simple linear filtering, but more like identify the upper and lower levels and use a fixed mixture as the reference value, not the ratio as is comes in in that patricular timeframe.

Almost  The PSD is constant at low frequencies and rolls off with 1/f² at high frequencies. The Allan deviation shows a characteristic bump. I have attached a simulation result using a continuous-time Markov chain. I also added the Python code to play with. Put both files in same folder and install the dependencies. Have fun.

Edit: I forgot to mention the following. The plots show different burst noise spectra for varying parameters. τ_1 and τ_0 are the lifetime of the upper (1) and lower (0) state. In those plots τ_0 = 1 s and τ_1 is changed. The maximum noise is found for τ_0 = τ_1, obviously, because this will yield the maximum transitions per unit of time.

[KT88]

@macaba: A crude DA simulation could be quite simple...just add a current source and apply a linear slope. Yes that's not the real deal but it would demonstrate the function of the servo...
One thing I forgot is to bypass C1 and C2 at start-up...
Btw. this servo approach should work for high pass filter as well.

[miro123]

So is there a way to cheat fate? Yes and no. 1/f noise has an Achilles' heel. It is correlated. So in other words, it does give you some information about its future. This can be used by a predictive filter like a Kalman filter

Last month I'm busy with implementation of Kalman Filter on multi-reference board. Unfortunately no time to document or share it. i have learned something from HW V.1 =- I need voltage sources with different properties. (long,shor stability noise TC etc) The Kalman filter work much better - get best of all sources. Some old notes I put them  here https://www.eevblog.com/forum/metrology/lm399adr1399/msg3759236/#msg3759236
I've been playing with supercaps too. Aside the leak and DA they TC is huge. I suspect their capacitance is strongly temperature dependent since C*U=const. As alredy Math siad you best voltage source. I V2 I want to combine the same LM399s and ADR1399 - and maybe  supercaps as popcorn less source to Kalman filter. Hopefully those supercaps does not exhibit any non-linear behavior /e.g. hystreresys etc/ and they fit in simple mathematical model.

[iMo]

A major part of the low frequency noise is popcorn type noise with more or less discrete jumps. This is especially true for the LM399 reference..

What about a "popcorn detector" - something hw-wise simple, counting the "edges" of the jumps. Thus no need to use a DMM. Let us assume the edges of the "jumps" are "fast", something with a

high_pass_filter ->amplifier->comparator.. 


PS: something like this (4uV jumps @7V, buried in some 20+uVpp signal, 1-100Hz):

[Kleinstein]

Chances are it would need 2 comparators to detect the jumps up and down sparate and than some flip-flop to remember the current state.  If this effort is wirth it depends on how good the detection works / how domninant the popcorn noise is.  For references with rather prominent popcorn noise like he LM399 it may be possible to get low enough an error rate to get an improvement and not extra noise from wrongly detected jumps. It also gets more complicated if there are multiple levels, or multiple jumps in one direction possible, so more long lived states involved.

From what I have seen so far the LTZ1000 usually does not show that much popcorn noise and in this case trying to identify jumps may do more harm than good. With my  JFET based reference It looks more like multiple levels (e.g. multiple long lived states) and thus no easy solution.  So it really depends on the reference to start with.

[Gerhard_dk4xp]

In a previous live, 1978 or so, I stumbled across a RCA application note
that did about that.

Cheers, Gerhard

[iMo]

Chances are it would need 2 comparators to detect the jumps up and down sparate and than some flip-flop to remember the current state..

I've added a schematics above - the edges (100ns fast let say) are easy to get off the noise mess, in the first iteration we do not need to detect the current state, but rather to look at a 399 sample - how it "pops". You may divide the number of edges by two and you get the number of pops/time. Thus a simple popcorn indicator..

[Gerhard_dk4xp]

I still have the data book. 1974. I still was in school then.

ed: someone was so nice to scan it:
<       https://www.mikrocontroller.net/attachment/179195/RCA_ICAN6732.pdf         >

[David Hess]

Attached are some photographs of flicker noise that I tracked down inside the analog channel switch of a Tektronix 2230 analog and digital storage oscilloscope.

The best physical model I found for flicker noise is trapped charge with multiple time constants, so identical to popcorn noise, but at a lower level and with multiple time constants all added together.

[mawyatt]

When RCA first introduced the CMOS based Op-Amps way back, many had serious Popcorn noise. We utilized RCA CMOS Logic and Analog right from it's introduction and "experienced" this effect first hand 

This was eventually traced to surface contamination and significantly improved with a "cleaner" process & packaging, and even chip redesigns, and likely initiated the Burst Noice App note mentioned, since RCA CMOS was well known to have this problem!!

Best,

[dietert1]

It's hard to believe the problem hasn't been solved up to now. These stories are ancient and one would assume that semiconductors are near perfect nowadays. Maybe high precision analog is similar to a moon rocket in that one needs to start from the basics over and over.

Regards, Dieter

[iMo]

The 1/f and popcorn problem was solved already..
..Before the Big Bang..

[David Hess]

It's hard to believe the problem hasn't been solved up to now. These stories are ancient and one would assume that semiconductors are near perfect nowadays. Maybe high precision analog is similar to a moon rocket in that one needs to start from the basics over and over.

Popcorn and flicker noise *are* solved problems.  We know what causes them and we know how to minimize them.  Some devices, like MOSFETs, have higher flicker noise because of how they are constructed.  We also know how to remove flicker noise with chopper stabilization.

[iMo]

..Popcorn and flicker noise *are* solved problems.  We know what causes them and we know how to minimize them..

We perhaps know how to "minimize them" but the stuff has not been understood yet. Both exist in any systems you may imagine and most probably are related to fundamental properties of our current universe..

[David Hess]

..Popcorn and flicker noise *are* solved problems.  We know what causes them and we know how to minimize them..

We perhaps know how to "minimize them" but the stuff has not been understood yet. Both exist in any systems you may imagine and most probably are related to fundamental properties of our current universe..

Here we are talking about transistors and PN junctions where charge trapped in defects causes both.

[dietert1]

A nice introduction to 1/f noise is http://www.scholarpedia.org/article/1/f_noise. Must be 2007 or so.
Elsewhere i found that electromigration at bond connections (metal to semiconductor) also generates flicker noise. But all this does not predict the noise level. It isn't enough to observe noise above white noise below a certain corner frequency. That may be any low frequency process and whether it produces 1/f noise over several decades needs to be checked.
One example i have is the observation of an ADR1399 evaluation kit i got recently and prepared well enough for noise levels near/below datasheet specs. But only during the night. During daytime, when the photovoltaic generators 3 m above our lab are working, noise level increases by a factor two - depending on how bright the sun shines. There are no two levels or so. Maybe i need to log the generators, there used to be ethernet connectivity.

Regards, Dieter

[iMo]

As a student at uni I worked in a dept. where they made 1/f noise research (80ties). I built a chamber (a bigger shoe box dimensions), made of a 30mm thick soft iron sheets (welded together), inside walls covered by a copper foil. There was a single hole in it, around 5mm dia, just for the coax (and a low noise preamp inside I built for them as well). Samples and amplifier powered by large stack of batteries placed inside. I can remember the top cover (a top hatch with a handle) was so heavy, that some colleagues had hard time to operate the chamber. I can still remember the sound of the cover when the cover hit the box (the touching surfaces had nice fine mirror finish, thus the box was almost airtight). Luckily the measurements were pretty long so no serious muscular diseases were acquired..   They made subsurface zeners measurements with the chamber as well.

[RoGeorge]

Reading through this thread, it seems the capacitor's imperfections are introducing some limitations when used to separate out the DC, or when used to implement low frequency filters.

I wonder why not using transformers instead (to remove the DC), and coils instead of capacitors for filtering?

[Kleinstein]

Inductors are also not ideal (unless you can use superconductors) with series resistance. The large inductance values would need iron cores and there one can have some kind of delayed hysteresis leading to similar and possible worse effects the dielectric absorbtion.
To avoid the trouble from DC current one would more combine an inductor with a capacitors instead of a resistor. Still inductors in the kH range are rare (and large). So one would still need a relatively large capacitor, but it could still be an option, as the inductor could live with a little more leakage.
So inductors are usually not used for low frequencies. They get more popular in the MHz range, buit not in the mHz range.

[miro123]

The link provide by Dieter gives the fundamental.  http://www.scholarpedia.org/article/1/f_noise
Mathematicians says that you cannot reduce the 1/f with simple integrator or differentiator.
Engineers knows that soldering of high value RLC components around  voltage reference makes thinks worse.
@Diether - I was trying to characterize several LM399 and ADR1399 - I sow that pop-corrn characteristics of those devices has changed over time. I have not idea is there any long term correlation with external factors  - e.g. ambient temperature, humidity - The artificial noise is reduced using batteries and optical interface

[dietert1]

Yesterday some Wima MKP4 10uF 250V arrived here and the one i tested first is showing 7 pA residual current at 14.1 V.
In comparison to the 70 pA input current of the OPA189 used as buffer, that isn't bad. These capacitor aren't that large. One could use two of them with 5 KOhm resistors to implement the Sallen-Key lowpass and errors should stay below 1 uV.
The ADR1399 evaluation kit already got a temperature isolated "shoe box" around it. That will likely become a thermostat when ready. Total consumption could be 1 W including the 600 mW of the ADR1399 heater.

Regards, Dieter

[RoGeorge]

The large inductance values ...
... in the mHz range.

For very big inductances, maybe a multiplier (with gyrators) to emulate a big coil. 
Never tried, but it should work.  Same idea might be advantageous for capacitors, too, using a small capacitor multiplied instead of a leaky electrolytic.

Even more, either inductors or capacitors, they are not used here as energy storage, but to form filters.  They can do the d/dt operation for voltage or current, and thus they can form filters.  However, nowadays number crunching and ADC/DAC are cheap, might make sense to go digital and filter there.

Putting aside how practical will be to implement any of that, either analog or digital, for active noise cancellation, two things are needed:
- 1. to decide what to subtract from the input mix of Vref+noise (usually using RC filters)
- 2. to correct the output (usually an adder)

However, point 1. takes time, so we will also need
- 3. a delay line for the original signal, so the extracted Vnoise and the Vref+noise are in sync in the cancellation adder from point 2.

Speaking while writing this, once we have a delay line "component" we can use that to implement filters, and get rid entirely of big capacitors or big inductors.

I have zero hands-on experience with voltage references, and to implement such low frequency filters I would be tempted to go digital to process the noise (before adding it back to the analog circuit for cancelling).


Heaving an analog delay line would help filtering out the shot noise, too. 

We can detect when a jump in voltage occurs (by comparing the current value with the most expected averaged value), and when a jump in voltage occurs, we disregard the current (wrong) voltage, and instead, copy to the analog output a previously known good voltage (we take the previously known good voltage from a tapped delay line, where each delay tap is connected to the input of an analog multiplexer, and we decide which input is considered as correct, or as heaving no pop-noise in it at that moment in the tapped delay line).

[iMo]

The issue here is the amplitudes of the jumps are small, like 2-4uV for 399 and even much less with 1000. And the jumps are buried in white and 1/f noise too. While looking at the measurements above made by a 7.5digits dmm you clearly see the jumps, but that "background" noise is filtered out by the dmm. Not having the dmm you would not see the jumps in the background wide band noise. Therefore, for example, my above idea with amplifying the "edges" of the jumps would not work in practice..

Delay line - that is my concern too - you are summing up two signals where one of them will be delayed by the high pass filter in the LNA (on the picture in the first post)..

What would be the propagation delay through the LNA?

[RoGeorge]

A voltage reference is a very particular case of signal, where the time lag between the physical live/noisy Vref and the clean output Vref does not matter.  In fact, the output of a Vref should never change (ideally).  We know it is expected to be constant, and that is the extra info we must take advantage of.

That is why, if we have long term memory (just like the measuring instrument has had while measuring/averaging for those 7.5 digit plots) we can filter out any pop-noise before it arrives at the output, here by outputting a recorded value of the Vref, recorded from a recent period of time when no pop-noise was present.

About the instrument having many digits, true that, but to detect pop noises we only need a similar high resolution measurement, but without the need of high accuracy, while a proper 7.5 digits instrument has to have both.  So our task is simpler, we only need to see sudden changes.  Should be doable.

As a side note, I'm aware that plenty of smart people thought a lot already about how to get a clean Vref, and low hanging fruits are long gone, so the only hope from now on is to try very different/stupid approaches, like the proposed digital filtering + analog delay line. 

[iMo]

..About the instrument having many digits, true that, but to detect pop noises we only need a similar high resolution measurement, but without the need of high accuracy, while a proper 7.5 digits instrument has to have both.  So our task is simpler, we only need to see sudden changes.  Should be doable..

I jumped on that idea in my earlier post above - to detect the sudden changes. My idea has been to high-pass the signal and to extract the "change" - represented by the rising and falling edges of the popcorn jumps. It works great when the background noise/signal contains lower spectral components only, ie. like frequencies limited to 1kHz max. You set the high-pass to 100kHz and you get perfectly every edge, even though when the signal amplitude to the jump amplitude ratio is X hundreds. But how to do it practice? When you limit broadbanded vref signal with a low pass filter, for example, you limit the jump's edges amplitudes as well (the edges are "high freqs") and you are not able to extract the edges that way - as they are not steep edges anymore.. And without limiting it you'll get a broadbanded signal and you are not able to detect the jump's edges as the signal contains unlimited number of "edges"..

PS: an example below - Vref signal contains max 1kHz (1-1000Hz sine generators used), the total background noise is 400uVpp, the popcorn jumps are 1uV on 7V with 1sec duration and 1us edges, every change/edge has been detected properly.

[MasterT]

The large inductance values ...
... in the mHz range.

1. --- For very big inductances, maybe a multiplier (with gyrators) to emulate a big coil. 
Never tried, but it should work.  Same idea might be advantageous for capacitors, too, using a small capacitor multiplied instead of a leaky electrolytic.

Even more, either inductors or capacitors, they are not used here as energy storage, but to form filters. 


2. ---- We can detect when a jump in voltage occurs (by comparing the current value with the most expected averaged value), and when a jump in voltage occurs, we disregard the current (wrong) voltage, and instead, copy to the analog output a previously known good voltage (we take the previously known good voltage from a tapped delay line, where each delay tap is connected to the input of an analog multiplexer, and we decide which input is considered as correct, or as heaving no pop-noise in it at that moment in the tapped delay line).

1. I think using small capacitors or gyrators is bad idea. Since more OPAmp amplification required and consequently more OPAmp noise getting in the path.  Capacitors served exactly to this purpose, as energy reservoir, the bigger energy - less entropy.

2. Reminds me median filter, where highest / lowest values are kicked out of the consideration. Median is the best (IMHO) filter to process spiky interference - noise.
Seems, we are arriving to the point that "classic" linear filter is not fit for reference voltage, doesn't matter if its 1-st order RC or 2-nd Sallen-Key.
 Going "median" way requires S/H circuits - to keep "most reliable value" till analog processing /evaluation new data completed. Processing may include linear continuous filtering over some periods of time, like below average pop-corn frequency, than comparator new/ previous values and some adjustment circuits if new value accepted as "noise-free" but slightly off compare to current.

[dietert1]

In my understanding the edges to detect result from the integration of flicker noise events. And a low pass with lower corner frequency supresses most of them, but will eventually show a similar picture from more rare events of higher amplitude.
So one can use an analog low pass filter to limit noise bandwidth in oder to simplify subsequent digitial processing for the purpose of noise cancellation.

Regards, Dieter

[DeltaSigmaD]

Passive noise filter for references

Active noise filter circuits substitute reference noise by noise of the operational amplifier, which has much lower noise. However, there is also a theoretical possibility to filter output noise of the reference with a passive filter while a low output impedance is provided. The attached reference filter circuit has a 2nd order Bessel characteristic with 1 Hz cut-off frequency and 0.25 Ohm series resistance. Such circuit could provide a reference voltage almost free of noise above the cut-off. This could be very interesting for data acquisition, e.g. noise measurement, in this frequency range.

The practical realisation would be not easy. The total leakage current of the capacitors should be reduced to less than 400 nA. Active circuits to reduce the 0.25 Ohm output impedance add OP noise, so that the passive filter is useless (I tested several topologies). The inductor can generate several kinds of noise. First of all, the inductor is working as antenna for EMI, and magnetic shielding of the filter is required. Second, Barkhausen noise is generated if the inductor current or the surrounding magnetic field of the inductor is changed. Third, ferrites generate noise due to loss mechanisms of the ferrite. A larger part of this inductor noise could be suppressed by the passive filter itself.

There is no other filter which can - theoretically - obtain lower noise (excluding superconducting circuits). Has anyone experience with the noise properties of inductors?

[dietert1]

As far as i understand, interaction with external fields is proportional to stray inductance. So one could try a toroid. For example a toroidal common mode choke with both coils in series, e.g. SCR-020-0R55A250J. Its two coils add up for 0.4 Ohm.

The capacitors may be more difficult.

Regards, Dieter

[Kleinstein]

With the relatively low resistance from the inductor(s) is can be attractive to use electrolytic capacitors or super capacitors, possibly even without extra compensation for the leakage. There is the DA part, but even that is partially attenuated.
For the damping part is could be possible to use a resistor parallel to the inductor instead of the extra RC element.
I would consider even more inductance. With a somewhat larger core one can get to the 1 H and maybe even 100 H range. An iron core (e.g. torroidal transformer) may not be that bad.

For the noise, there can also be Barkhausen jumps caused by mechanical stress / noise. The jumps are small and may be small enough to not be a serious problem.
The mechanical stress can limit the symmetry and this way cause more leakage inductance in a toroidal transformer than one may think.

[DeltaSigmaD]

Thanks for the information on mechanical stress. That's one source for possible external influences more to be considered.

A damping resistor parallel to the inductor leads to a filter transfer function which has a bandpass contribution. The load step response shows therefore strong oscillations or, with increased damping, the filter order degrades to 1st order. The filter topology shown in the picture can have oscillation-free step response by choosing e.g. critical damping characteristic.

[MegaVolt]

I'll bring up the topic.

A standard voltage reference noise meter contains: series capacitor - 1-5 kΩ resistor per ground -> amplifier.
And we trust this RC circuit to separate AC from DC. Those. on R only alternating current on C only direct current.
Let's swap them. Let's put R first and C on the ground. And we will take the signal from C. It should have much less noise.
If that doesn't work then the noise measurement circuit shouldn't work either
What am I missing?

[Kleinstein]

The AC coupling circuit does not care much about a little (e.g. 100 µV) range residual offset. As long as it is reasonable constant the leakage of the capacitor is not that bad. The search for low leakage is mainly be cause leakage often also comes with extra noise, not because of the DC leakage on it's own. Similart amplifier input current by itself is only causing a little extra leakage and only the current noise that usually comes with the bias current is really a problem.

The the low pass filter leakage add offset and this can be a problem.

How much lower the noise with the filter actually is, depends on how the reference is used / measured. With the high end references one mainly cares about the low frequency part and measures / compared the reference over some time, e.g. with a DMM. The measuring instrument already does quite some filtering.  Much of the noise a simple filter can remove may be suppressed by the DMM anyway. Still many DMM as not perfect in noise suppression - e.g. the usual 10 PLC AZ mode lets some of the 2.5 Hz frequency range noise pass even with averaging many readings. So the filter can still help, but that effect can be more limited than one may think.

[miro123]

I think that the tread is moving in magic circle.
I think that we must put the requirement on the table. I have learned /from my little experinace/ that Vref requirements are contradictory
 - low white noise
 - low low mid-low frequency - 1..0.1Hz - creates top on low, and ultra low frequency.
 - high stability
 - low TC
To make it clear - Low noise high stability and low TC Vref does not exist - come compromise should be made. This statement is based on traditional voltage reference design- the modern digital approaches are not taken into account
As far as I understand this tread discus reducing white noise and low refuency noise in lets say 1 min base - as result of it other parameters get worse - eg. stability TC mid low freqency noise

[David Hess]

Low drift and low low frequency noise are essentially the same thing because at low frequencies they are indistinguishable.  A reference with one effectively has the other.

High frequency noise can be ignored because it is both easy to filter out, and integrating ADCs will ignore it anyway.  It matters of course for a sampling ADC.

Flicker noise is especially pernicious because its amplitude increases below the low pass filter's cutoff frequency, which is why it is better to avoid generating it than to try and filter it out.