Measuring switch mode supply noise - HP461A replacement, ADA4895?

[Martinn]

Reading AN118 https://www.analog.com/media/en/technical-documentation/application-notes/AN118fb.pdf on low noise power supplies, Jim Williams makes some recommendations on measurement techniques.
One is using a pre-amp to see 100 uV on a scope with 100 MHz bandwith. He recommends the HP461A, which is a 100x preamp with 20 uVrms noise and 175 MHz bandwidth (see p. 17). One could buy one of those (used), but what about building this with modern low noise op-amps?
I thought about two stages 10x each with ADA4895 https://www.analog.com/en/parametricsearch/11091#/p4502=1G|4G - 1.5 GHz GBW, 1 nV/rtHz noise. 1 nV/rtHz over 100 MHz gives 10 uVrms, not accounting for the feedback resistors.
As I have no brilliant idea on how to make a virtual ground (low noise, low impedance) from a 9V battery, I'd go for a 2x 3AA or AAA battery supply.
What do you think?

- Martin

[Vovk_Z]

I guess that you should try and post here to inform us.

Power supply is a low impedance signal source, so I guess it may be quite easy to choose an appropriate opamp.

[jonpaul]

Bonjour
HP had excellent app notes on exactly this topic, in HP tech notes/ journal, perhaps 1980s.

We used  Tektronix AM501 differential amplifier, various tektronix 7000 vertical plug-ins 

 an opamp in x100 differential or x1000 can work, but use battery or we'll isolated power supply.

Certain  portable Tektronix scopes of 1970s..1990s 465, 475, had vertical CH1 output, loop back to CH2 input for sub milli volt range.

 Beware of transients thru AC coupling caps, during DUT,PSU turn on

Jon


Bon courage

Jon

[Martinn]

... an opamp in x100 differential or x1000 can work, but use battery or we'll isolated power supply.

Bonsoir,

why would you need a differential probe? I was thinking of having everything battery-powered in a metal cookie box.
It seems the old Tek units were better for low level analog, although newest Siglent/Rigol scopes also go lower noise 12 bit.

I was thinging something simple (see schematic). Not sure if the output series termination is really necessary if you have a short cable and the scope terminated, loses 2x gain.
Input protection is from Jim Williams AN, 100x gain from ADA4895 datasheet.

Thanks - Martin

[T3sl4co1l]

Just any LNA or preamp will do.  Surely you don't need LF performance for a HF measurement, right?

Like, a bunch of years ago I built up a couple of these,





As you can see, the as-built model isn't identical with the schematic, and I don't think I have an updated schematic handy anywhere unfortunately.  I also had built two variants, one with the peaking coil as shown, another in a transconductance-transresistance sort of architecture (the proper name of which I always forget).  Despite the crude layout, bandwidth is over 600MHz, with enough gain and low enough noise to get useful performance on the scope, or an extra 10dB+ SNR on the spec (granted, my spec has a relatively high noise figure so that's not hard to achieve, heh).

The gain is not precise, but that's fine as it's easy enough to calibrate.  Measure the insertion gain, and adjust as needed.  Or maybe I trimmed it to 20dB, I don't even remember, honestly...

My TDS460 as-is, isn't bad at all.  I see a noise floor of 130uV RMS in 350MHz BW.  That's more than enough for a lot of SMPS purposes.  I did a low noise project some years back which showed around 40dBuV (100kHz RBW) in the 20-200MHz range (conducted, in a bespoke setup).  Not bad considering the use of conventional controllers with no shielding.

Tim

[jonpaul]

Rebonjour a tous:


Mentioned diff amp as at uV levels the CM noise can be an issue.

The entire subject of SMPS noise is not simple as there is conducted, radiated, noise backto mains and noise at output.

The BW and typical noise levels differ eg mains conducted 150 kHz..30 MHz and radiatied from SW to GHz.

See applicable regs and stds for your region, eg FCC, IEC DE etc.

Thus is one checking mains or ripple at 2x mains F, conducted noise or radiated?

Different treatments to recue/minimize each one.

We mentioned diff amp (not probe) TEK plugin  AM501 as it IS diff and very high gain and variablew BW, but a single ended plugin is fine in most cases


Bon chance


Jon

[coppercone2]

just get a 461 they are cheap

the only bad thing about it is the attenuator switch, I hate it, its like that episode of macguyver where the German guy ties him to the swingblade interrogation machine, it also seems to remind me of a bed time activity that gets people complaining, its a preverted mechanism

its small,you can get it tuned for pulse response, and its easy to service (outside of the attenuator switch which is a bit tricky).

I want to replace mine with a sideways lever switch of some kind, so there is a independent switch to control attenuator without the power switch being involved.

Once you factor in the chassis elements, and build quality, your not gonna get a cheaper DIY amplifier IMO.

I also tried it for HF reception, and for whatever its worth, it does sound good for HF reception (acoustically). You can actually connect it to a HF radio to boost gain, with the same antenna and radio I was able to get more stations with the amp hooked up, and some stuff came in beautifully.

[Martinn]

Like, a bunch of years ago I built up a couple of these,

Hi Tim,

interesting - I am tempted to just design a PCB and order it from JLC - if you manage to find all parts at LCSC, you can just order an assembled prototype, makes life much easier. And the schematic seems to contain a number of anti-oscillation fixes already.
However, I don't even have a signal generator to test it (I do have a spectrum analyzer). RF generators from Rigol and such start at $2k+, but have all kind of modulation options I would not need. Maybe an ERASynth Micro https://www.crowdsupply.com/era-instruments/erasynth-micro? Or some mini-nano-pico-VNA?

But right now I have a box from JLC with five other projects I need to handle first... maybe I'll be back later.

Thanks, Martin

[Martinn]

just get a 461 they are cheap

True, amazingly (though at least 3x the price here in Europe). Probably the least expensive HP device you can buy.
Although I'd miss the experience of designing a low noise amplifier myself.

[coppercone2]

for some reason I think for a lab equipment the old one might be more robust and stronger though, they have big resistors in there and what not, and it has a transformer, switches, etc (that makes price shoot up)

[Martinn]

I guess that you should try and post here to inform us.

Here we go. After an itermediate discussion on termination https://www.eevblog.com/forum/projects/series-coax-termination-for-output-really-necessary/msg4601593/
I built it it as shown here https://www.eevblog.com/forum/projects/series-coax-termination-for-output-really-necessary/msg4608472/#msg4608472
As I halfway expected I got a nice 310 MHz oscillator after first startup, exchanged the feeback Cs to 4 pF and all was fine.
Bandwidth (measured with a VNWA) is 110 MHz for the 10x output and 88 MHz for the 100x output. Low frequency cutoff is 1.8 kHz.
RMS noise is around 2 mV (input terminated with 50 Ohms, so 25 Ohms total), a bit over the 1.4 mV estimated from an LTSpice noise simulation.
There is some peaking (+2 dB at 40 MHz), but assuming flat response and 88 MHz bandwidth, 2 mV noise translates to 2.1 nV/Hz^1/2, compared to 1.2 nV from the noise simulation (opamp has 1 nV/Hz^1/2).

So I think I am covered for switched mode noise for now. Out of curiosity I built one of Jim Williams' LDO noise setups (LNA+100 kHz filter) and on top a AM502 just popped up on ebay, which is on its way to me now (Jon - you mentioned the AM501 - did you mean AM502)? I also considered a 7A22, but as I already have two AM500 chassis and no 7A22 was in sight, I went for the AM502.
As someone mentioned I could have bought a HP461A (€125 in Germany), but this was more fun. BTW the HP461A is specified at 40 uV input noise, so 4 mV at +40 dB, but 150 MHz bandwidth (compared to 2 mV/88 MHz for my design).
I also attached a spectrum analyzer noise measurement - noise marker shows -14.77 dBuV/Hz. I am not very familiar with spectrum analyzer noise measurements, so I best leave this uncommented... first of all I find the unit strange (I changed it from dBm/Hz). Shouldn't it be dBuV/Hz^1/2 - the conversion from dBm(W - a power unit) to dbV uses a square root... (U=sqrt(P*R))

- Martin