Measuring less than 1mV noise/signals on an oscilloscope

[mawyatt]

When working on "low to medium-field" MRI systems, operating around 5 to 15 MHz, ca. 1990, I used the late, lamented Sony 2SK152 JFET, which had a better-than-the-average-FET ratio of Y_fs/C_iss.
Unfortunately, it was intended for low-noise vidicon preamplifiers, which market disappeared.  A specialty manufacturer, InterFet, produced a similar part for replacement.
The original Sony part claimed 1.2 nV/Hz^1/2 (typ) at 1 kHz, when operated at 10 mA.  The highest-current bin, 2SK152-4 (27.4 to 42 mA) had higher transconductance, with slightly lower noise voltage.
The Sony data sheet shows the noise voltage flattish up to 1 MHz.  At 100 MHz, in a tuned, noise-matched circuit, they claimed 1.8 dB noise figure (typ) at 100 MHz, at 10 mA.
https://www.datasheets360.com/pdf/5688259163679467641

Sony had great speciality parts back then, and one the best bipolar processes. They made a bipolar DAC used in their studio monitors for RGB generation that was better for RF than any other DAC around. We often used those but kept quiet about it!! One of our customers would tell their favorite contractor (not us) everything we did, so needed to be careful around them!!!

Best

[Alex Nikitin]

When working on "low to medium-field" MRI systems, operating around 5 to 15 MHz, ca. 1990, I used the late, lamented Sony 2SK152 JFET, which had a better-than-the-average-FET ratio of Y_fs/C_iss.
Unfortunately, it was intended for low-noise vidicon preamplifiers, which market disappeared.  A specialty manufacturer, InterFet, produced a similar part for replacement.
The original Sony part claimed 1.2 nV/Hz^1/2 (typ) at 1 kHz, when operated at 10 mA.  The highest-current bin, 2SK152-4 (27.4 to 42 mA) had higher transconductance, with slightly lower noise voltage.
The Sony data sheet shows the noise voltage flattish up to 1 MHz.  At 100 MHz, in a tuned, noise-matched circuit, they claimed 1.8 dB noise figure (typ) at 100 MHz, at 10 mA.
https://www.datasheets360.com/pdf/5688259163679467641

Unfortunately, the InterFET IFN152 has a significantly larger input capacitance, around 15pF (the original 2SK152 had 8pF) so the equivalent charge noise is  double that of the original SONY device. Fortunately, there are more recent devices that are even better than the original 2SK152 in this respect, for example the classic (now also obsolete though I have some) BF862 from NXP with 0.8nV/rtHz at 100kHz and only 10pF input capacitance, and some current low noise JFETs from ON Semi, for example the CPH3910 with ~1nV/rtHz noise and only 6pF input capacitance, about twice as good as the original SONY device in the charge noise respect.

Cheers

Alex

[LooseJunkHater]

Was building a simple EMI enclosure yesterday from a steel cookie tin + aluminum tape (to see if it'll help decrease the circuit noise + RF interference) and realized that the aluminum tape isn't using a conductive adhesive Now none of the aluminum tape strips are electrically connected together.

Any suggestions for electrically connecting the individual tape strips + enclosure? I'm thinking of either purchasing some proper copper tape which is conductive on both sides, or solder paste. Conductive adhesives/glues appear to be $10USD+ for only 0.2mL...

My fear of using solder paste is some of the microscopic solder balls dislodging and causing a short in the circuit or elsewhere or just simply not being conductive after a certain period of time, when the flux dries...

[TimFox]

The good 3M 1181 conductive-back copper tape works well, but you should use new stuff each time.
Not all copper tape has this type of adhesive.

[Alex Nikitin]

https://uk.farnell.com/c/tools-production-supplies/tape/electronic-tape/copper-aluminium-tape

Cheers

Alex

[EC8010]

What were you hoping to achieve by lining the tin with aluminium tape?

The self-adhesive copper tape sold to gardeners for deterring slugs almost certainly doesn't have conductive adhesive. But does it matter? If you use it to bridge the seam between tin and lid (for example), the adhesive is so thin that the capacitance between the tape and whatever it is stuck to is pretty high and a short circuit at RF. I very much doubt that you can see any difference between tape with conductive adhesive and without, except in price. If in doubt, you can always run a seam of solder along the edge.

[LooseJunkHater]

What were you hoping to achieve by lining the tin with aluminium tape?

Improved RF (electrical interference) blocking, similar to what I discussed here on the EEVblog, with the main difference being that I'd be doing only the tin outer (for weak MI) --> Aluminum Foil (for EI) --> Air Gap --> Copper Tape (for EI) --> Plastic Enclosure + PCB

The self-adhesive copper tape sold to gardeners for deterring slugs almost certainly doesn't have conductive adhesive. But does it matter? If you use it to bridge the seam between tin and lid (for example), the adhesive is so thin that the capacitance between the tape and whatever it is stuck to is pretty high and a short circuit at RF. I very much doubt that you can see any difference between tape with conductive adhesive and without, except in price. If in doubt, you can always run a seam of solder along the edge.

I'll see if I can manage to get solder to stick to the steel enclosure + aluminum tape, but I don't have strong hope. To be continued... 

[LooseJunkHater]

The good 3M 1181 conductive-back copper tape works well, but you should use new stuff each time.
Not all copper tape has this type of adhesive.

Honestly was thinking that most/all aluminum & copper tapes would have a conductive adhesive backing, but clearly not. The more I know! Live and learn.

[LooseJunkHater]

The self-adhesive copper tape sold to gardeners for deterring slugs almost certainly doesn't have conductive adhesive. But does it matter? If you use it to bridge the seam between tin and lid (for example), the adhesive is so thin that the capacitance between the tape and whatever it is stuck to is pretty high and a short circuit at RF. I very much doubt that you can see any difference between tape with conductive adhesive and without, except in price. If in doubt, you can always run a seam of solder along the edge.

I'll see if I can manage to get solder to stick to the steel enclosure + aluminum tape, but I don't have strong hope. To be continued... 

Well, I did it; I managed to solder to the aluminum tape. Tried plenty of flux but was unsuccessful, however a random comment on YouTube suggested motor/engine oil (I used bike chain oil), apply it to the site, scrape the aluminum & apply solder. Eventually, the aluminum oxide layer is scratched away, I assume the oil prevents quick reoxidation of the aluminum, and then the solder was able to stick. The joints look fairly weak, but they should be "good enough", as I'm only connecting all of the aluminum pieces together.

[smaultre]

Take a look on Tektronix ADA400 Differential Preamplifier https://w140.com/tekwiki/wiki/ADA400 can be connected via Tektronix 1103 power supply to any modern hi-res scopes.

[LooseJunkHater]

Take a look on Tektronix ADA400 Differential Preamplifier https://w140.com/tekwiki/wiki/ADA400 can be connected via Tektronix 1103 power supply to any modern hi-res scopes.

Mostly-correct schematic of the ADA400.

Significantly more complicated than my current design or the LFLNA-80, (likely?) requires matched initial high-pass filter capacitor, a lot of internal trimmers, unnecessary gain controls. 

If I can find a 100% complete schematic (as there's a lot of things that the Github poster is unsure about), I may consider simulating it, seeing how well it works, what can be removed, and how to speed it up. The device is over $300USD on EBay, so I definitely won't be purchasing one.

Early on I did consider a differential input amplifier, but I quickly became dissuaded by the requirement for matching input high-pass filter capacitors

Edit:

Looking at this image, I could likely remove the bandwidth switch, gain selector switch, input selection (AC, DC, GND) + associated circuitry. Hmm, maybe I will consider looking more into this design...

[LooseJunkHater]

Take a look on Tektronix ADA400 Differential Preamplifier https://w140.com/tekwiki/wiki/ADA400 can be connected via Tektronix 1103 power supply to any modern hi-res scopes.

Mostly-correct schematic of the ADA400.

Some notes on the schematic:
- Uses a +15V & -5V for DC offset voltage correction, instead of a DC servo (in my design) or high-pass filtered output (of LFLNA-80)
-  DC offset voltage correction is fed into an instrumentation amplifier, which also appears to be the main preamplifier with a super-low input bias current of 20pA (max).
- No idea what Q1 & Q2 do/are for
- AD829 amplifies the signal further for the output, and uses an external compensation capacitor for stability (per datasheet, appears to be required for gains less than 20x)
- Simple overload detection circuit

The trimmers: ( = I have no idea what it's doing, as I'm not smart enough, but I do take a guess)
- RV1 ()
- C2 ()
- C3 (. For reducing noise?)
- C6 (. Not enabled for 100KHz/Full bandwidth? Can be removed?)
- C7 (. Not enabled for 100KHz/Full bandwidth? Can be removed?)
- RV5 (. Not enabled for 100KHz/Full bandwidth? Can be removed?)
- RV6 (gain fine-tuning?)
- RV8 (gain fine-tuning?)

- RV2 (External pot: coarse DC offset voltage correction)
- RV3 (External pot: fine DC offset voltage correction)
- RV4 (another DC offset voltage correction trimmer)
- RV7 (DC offset voltage correction)

Without knowing the purpose for all of the trimmers, I don't think I'd really be able to replicate the circuit.

[TimFox]

Q1 and Q2 are JFETs connected as diodes (gate to channel), used as part of the input protection circuit.
Small JFETs are often used thus, exploiting the very low leakage current at low voltages, compared with signal diodes (e.g. 1N4148).

[dietert1]

Q1 and Q2 are low leakage current input protection diodes.
And yes, the circuit becomes more complicated if you include the second input channel and one needs matched filters etc. Maybe the 300 € ebay price is justified. The device supports capacitive coupling (in case of larger DC offsets) and i guess one could increase the two capacitors C1- and C1+ in order to implement a lower bandwidth limit of 0.1 Hz if required. Differential input is nice to have for low level measurements if the DUT is more than a little test PCB.

Regards, Dieter

[voltsandjolts]

BAV199 might be a good substitute for the Q1, Q2 diode connected jfets.

[David Hess]

I always skipped the J111 to J113 because it lacks the needed specifications, but these could be read from the curves.

[Kleinstein]

The 1st amplifier is the INA111. So this sets the input bias, noise and drift performance.  It is a good FET INA, but could still be limiting when it comes to super low noise. The ADA400 is mainly a differential probe, not a super low noise amplifier.

The trimmers are not that complicated to see what they are good for:
RV1 and C2 are for the input impedance per channel  -- this may be the slightly tricky part to trim, RV1 may use a fixed resistor
C3 to trim the higher frequency CMRR
C6 frequency compensation for setting with 1:100 divider at the front
C7 input capacitance for setting with 1:100 divider at the front
RV5  symmetry in the 1:100 dividers /  DC CMRR for 1:100 setting
RV6 overall gain
RV8 gain for 0.1 gain  - there may be something wrong in the gain switching part

[LooseJunkHater]

I ended up spending a bit of time simplifying the Tektronix ADA400A schematic that was posted to Github, to improve my understanding of it (image attached). Hopefully I didn't mess any of it up.

Removed:
- Gain selection + Trimpot RV8 (now it's a constant 100x gain)
- Bandwidth selection (now it's a constant "full" bandwidth; above 100KHz)
- Overload detection/Overload range LED
- 5v Relay (was associated with gain selection)

I'm a bit confused as to what "Scale_Factor_Coding" is for, as it doesn't seem to connect to anything on the circuit and attaches to the scope? Maybe it's just for automatic scaling change on the oscilloscope? https://raw.githubusercontent.com/mattico/ADA400A-Reversing/refs/heads/master/Images/ada400a_top_annotated.jpg

Additionally not entirely sure how the In+ circuit is wired up into the circuit (related to "To + Channel" being mentioned twice in the schematic), so I may try to figure this out + simulate the circuit.

_____________________________________________________________________________________

From simplifying the schematic + Kleinstein's explanation of the trimpots, I have a better understanding of what's going on in this circuit, the purpose of the trimpots, and how it's similar & different from the circuit I made.

[TimFox]

I have used differential-input oscilloscopes, ‘scope plug-ins, and external preamps over the years.
The variable bandwidth (LPF and sometimes HPF) feature is very useful when dealing with small signals.
I would retain at least the low-pass filter feature.

[LooseJunkHater]

I have used differential-input oscilloscopes, ‘scope plug-ins, and external preamps over the years.
The variable bandwidth (LPF and sometimes HPF) feature is very useful when dealing with small signals.
I would retain at least the low-pass filter feature.

How do the low pass + high pass filter settings help when viewing small signals? Would it be comparable to the 20M bandwidth limit on oscilloscopes, removing extraneous noise?

[Kleinstein]

The scale factor coding should be a way to tell the scope what gain is set. This is likely a tektronic specific probe interface. So this is a part to leave out in most cases.

The +in part should see essentially a copy of the input circuit. It is just the other side of the differential probe.

The filters are helping in removing noise, especially higher frequency noise. For the highest gain one may really want a reduced BW. 100 kHz instead of 20 MHz from the scope alread could give around 1/14 the higher frequency noise.

As shown in the cut down schematics the 1:100 divider is in place. For a gain of 1 one would remode R25, RV5 and related parts and reduce R16. The R16 value would be a compromise between noise and the level of protection.

[TimFox]

Specifically, if looking for 60 or 120 Hz interference with a 20 MHz or greater bandwidth ‘scope, an extreme LPF of 10 kHz will make the hum components more visible, especially with a “line” trigger source.

[LooseJunkHater]

Considering the #1 goal of this thread (and project) is about "Measuring less than 1mV noise/signals on an oscilloscope" (with lesser important, but still important design goals of 0.1Hz-100KHz measurement + low-cost), I see no rationale for incorporating any gains less than 100x for the circuit, increasing complexity + price. My oscilloscope and many other modern scopes can easily see single-ended AC signals around 10mV with high input impedance's. For differential signals above 10mV, I could use the XY mode of the scope & two probes, or build/buy a dedicated differential probe.

The dedicated low-pass and high-pass filter switches I can support the idea behind; being able to remove signals below 500Hz (AC + harmonics) & above 100KHz (RF) would be nice, as both have been causing me problems in my own circuit (as I discussed earlier in the thread + showed in the oscilloscope shots). I'll consider incorporating them into the simulated circuit and/or future builds. 

[TimFox]

On a traditional analog CRO, the first thing you see when on a sensitive mV/cm range and you apply the low-pass filter is that the trace width narrows considerably due to reduction of high-frequency noise in the reduced bandwidth.

[dietert1]

That simplified circuit has a gain of about 1, as there are two 10x gain stages and a 1:100 input divider (990K+10K). As mentioned before by Kleinstein the instrumentation amplifier INA111 may be detrimental to noise performance and unnecessary if you go for a single ended input.
I would also keep some low and high pass filter option in case you need that later on.

Regards, Dieter