View Full Version : PSU Assessment method


Andrew L Weekes
13-01-05, 05:23 AM
Something that might be of interest to all you budding tweakers out there.

I've just recently re-built my measurement kit for PSU testing with excellent results.

I use a good quality PC soundcard, with a low-noise, high gain buffer amplifier to protect the inputs and increase the measurement resolution.

Another interesting use for it though, is to listen to your PSU's!

This is an interesting assesment method, that tells you more about the likely impact of the PSU, sonically, than many other methods. The buffer amplifier can be used to drive a pair of headphones (or fed into an existing amp) and you can probe bits of circuitry and hear whats going on.

You'll hear hiss, hums, buzzes and even the audio itself in some circumstances. They will tell you a lot about what's going on on the power rails of your device. You'll obviously notbe able to hear the supersonic stuff, but it should prove enlightening all the same!

It's easy and fun to construct and demonstrates some great build techniques that are easy, yet can achieve great results for the little effort involved.

The following posts will give more detail, so please be patient!

Andy.

Andrew L Weekes
13-01-05, 05:28 AM
I've quickly knocked out a schematic for this, as it's pretty simple I strongly recommend the 'dead-bug' technique for assembly, no only is it quite fun and quick, but will give excellent results.

Here's the basics for the one I constructed: -

Solder the INA217 to the copper-clad first - I bent all legs, except pin 5, parallel to the board first, then formed pin5 to solder to the ground plane - a tiny dab of cyanoacrylate adhesive may help hold the IC it in place whilst doing the rest of the assembly, although I didn't do this.

Next install the decoupling to pins 4 & 7 - add a good, low inductance at HF, cap close to the IC supply pins and take it to ground via as short a lead as possible. If you have any 0805 SMD caps, these are perfect, and can be 'tombstoned' (i.e. mounted on end) between the pin and the ground place on the board. I used a small 0.01u polyester film as I didn't have much else around.

Then add the larger decoupling cap, something like a 10u electrolytic / tant (not critical) - observe correct polarity!

Next, bend pins 1 and 8 over the top of the IC and fit the 100R + 1R to set the gain (100x in this case). I used SMD's, but leaded are fine.

Now add the 10k's for the bias path at the input - I used 10k 0805 SMD's, tombstoned between the pins and the ground plane directly beneath, if you use leaded R's, try to ensure both go to the ground plane as close a possible as each other (note the way the schematic is drawn here, it's for a reason). The current flow in the ground plane could have the potential to create a differential input signal, at the resolution we have available here even small currents through a low-Z plane could be resolvable. This is where SMD's work well.

Next add the input coupling caps - I fixed mine down with hot-melt adhesive. Leave enough lead length between the IC pins and the cap to be able to fit the 1N4148 diodes. You can use a suitable non-polar cap, or just utilise back-to-back pairs, as I have.

Next fit the input R's, the output R and you're ready for test.

Andy.

P.S. Save the attached file and rename from .txt to .pdf to view it.

Andrew L Weekes
13-01-05, 05:34 AM
I use this amp as part of a calibrated measurement system, so in order to analyse whether the numbers are sensible, I fed in some calibrated signals.

The noise floor of the system is -160dBv which equates to 0.01uV rms!

I've now calibrated the system in dBV, attached are some quick plots...

First the noise floor - magenta trace is the buffer amp, input shorted, which is around -160dBv (with a little FFT averaging to lower the noise and produce a stable plot).

The red trace is the soundcard noise floor (input shorted) - bear in mind this is after cal, but without the amp, so the scaling isn't the same. The sound card exhibits some noise rise >30k, I'll look at that later, but it's hardly a significant problem at present.

Sorry about image size, hopefully they are still readable within the forum limitations.

Andy

Andrew L Weekes
13-01-05, 05:36 AM
I made up a potential divider, using a 1M and a 1R, so that I could feed in various high-level calibrated signals and get uV signals at the input to the amp.

The first was 1V rms into the divider, which gives 1uV in, which equates to -120dBV, as you can see from the attached image, the accuracy is excellent, you can read off the actual FFT values on the table.

I fed in two further low level signals into the divider, 100mV rms, which gives 0.1uV in and 20mV rms, which gives 0.02uV this would equate to -140dBV and -154 dBV respectively.

Again the results are great considering the tiny signal levels

I'm really impressed with how this has worked out, it's basic performance, for something so simple that only took an hour or so to complete, is fantastic. The slight mains pickup (50 / 100 Hz) on the measured traces is non-ideal input cabling which was long and not proper twisted pair, I need to dig out some proper screened twisted pair (I've some mic cable somewhere) to improve CMRR.

Maximum input to the amp / system is 40mV rms (-28dBv) which equates to 4V rms into the sound card, the amp has more headroom than this, running from +/-9V nominal (2x PP3 batteries), but that's the maximum input to the sound card.

Andy.

Andrew L Weekes
13-01-05, 05:37 AM
Here's the system with a -140dBv input (0.1uV RMS).

Andrew L Weekes
13-01-05, 05:39 AM
Here's a -154dBv input (0.02uV), resolvable if you look at 1kHz, above the noise floor.

System is showing -153dBv here, more than accurate enough for the purposes it's intended - most PSU's will be well above this floor.

martin clark
13-01-05, 05:41 AM
-153dBV :o

Nice one Andy. I'll have a go at this - I have long been using an old NE5532 for this purpose which works fairly well with switchable 20/30dB gain. Its a real 'instrumentation' amp - when not used experimenting it's frequently pressed into use as guitar or mic preamp!

Mind you, just coupling a pair of headphones directly to PSU rails with a cap can reveal lots - even on 'regulated' lines. My guitar playng OTOH has an appaling S:N ratio..

LesW
13-01-05, 06:22 AM
You had chance to look at a HiCap yet Andy..??

Andrew L Weekes
13-01-05, 06:39 AM
You had chance to look at a HiCap yet Andy..??

Now, why would I want to look at that ;)

LesW
13-01-05, 09:42 AM
Now, why would I want to look at that
The answer to which will be handed to the Clerk of the Court.....

Markus S
13-01-05, 09:47 AM
Slightly OT: I believe there was a thread on the Naim forum where Mr Tibbs was invited to come to Salisbury with one of your PSUs, Andy. Did that ever come to pass?

Andrew L Weekes
13-01-05, 09:59 AM
Slightly OT: I believe there was a thread on the Naim forum where Mr Tibbs was invited to come to Salisbury with one of your PSUs, Andy. Did that ever come to pass?

I don't believe so.

Andy.

Andrew L Weekes
13-01-05, 10:05 AM
The other thing you may want to add is some bandwidth limiting to the system presented here, as a wider measurement bandwidth amplifies the noise present.

The standard way of defining noise demonstrates this easily - take for example a 2nV / sqrt Hz noise figure. You'll see noise figures specified this way on op-amp data sheets, they define the noise within a certain bandwidth and can be used to work out the actual total noise, as measured on a meter, or oscilloscope. The bandwidth of this circuit is around 1MHz, from memory.

To work out the peak-peak noise in a given bandwidth, firstly we have to calculate the square root bit.

Lets say we have an audio bandwidth of 20Hz - 20kHz, the root bit becomes sqrt (20,000 - 20) = 141.35

Then multiply this by the noise figure: -

141.35 x 2 nV = 282.7 nV

This is the input noise of the amp, we know need to multiply this by the gain, assuming a gain of 100, for example, we get an output noise of: -

282.7nV x 100 = 28.27uV

As can be seen, if you re-evaluate the above, but increase the bandwidth, for example, to 1MHz, we get: -

sqrt (1000000) = 10,000

Multiply by the noise figure: -

10,000 x 2nV = 20uV

Multiply by the gain: -

20uV x 100 = 2mV

Or around 70x as much noise - quite a difference (an argument against wide-bandwidth audio systems?).

Also bear in mind the input impedance (and noise characteristics) of this amp make it unsuitable for probing high impedance circuits, and the impedance presented at the input will have a direct impact on noise floor, unless the circuit impedance is less than the equivalent noise of the circuit itself.

Andy.

Napsac
13-01-05, 10:11 AM
Had difficulty in converting your text file to pdf, do you have the scheme in another format, jpeg, photo of a sketch or something ?

Andrew L Weekes
13-01-05, 11:56 AM
Had difficulty in converting your text file to pdf, do you have the scheme in another format, jpeg, photo of a sketch or something ?

Right-click the link, then save as a file.

Rename from .txt to .pdf

You may need Acrobat Reader 6 to view (www.adobe.com)

Andy.

Mr Tibbs
13-01-05, 01:13 PM
"Another interesting use for it though, is to listen to your PSU's!"

That reminds me of when I got my SR's - I wanted to compare them (for noise) with my hicap and LT1086's, so rigged up a 'listening device'. I used a low-noise mic input on a mixing desk and monitored with headphones.

First on test was a basic LM317T reg. It was noisy as hell - Niagra Falls noisy.

The hicap was a good deal quieter - a gentle hiss.

The LT1086 was pretty impressive, being almost identical to the hicap reg's.

The SR's were dead quiet - without checking, I couldn't tell the difference between them being powered on or off.

"Slightly OT: I believe there was a thread on the Naim forum where Mr Tibbs was invited to come to Salisbury with one of your PSUs, Andy. Did that ever come to pass?"

Yeah, Paul S threw down the gauntlet. I was invited to bring the modded 102 and SR-6 Hicap. I didn't get around to it though. The Naim forum used to be quite 'enthusiast' friendly - I even have email from Naim to back that up. Things have changed since then.

Mr Tibbs

James
13-01-05, 01:30 PM
"The Naim forum used to be quite 'enthusiast' friendly - I even have email from Naim to back that up. Things have changed since then."

Ain't that the (sad) truth. :(

I'm surprised the E-I (http://naim-audio.infopop.net/1/OpenTopic?a=tpc&s=67019385&f=48019385&m=3631920723) and E-II (http://naim-audio.infopop.net/1/OpenTopic?a=tpc&s=67019385&f=48019385&m=6041955595&r=6041955595#6041955595) threads are still there.

Oh well, never mind.

James

Andrew L Weekes
14-01-05, 06:46 AM
That reminds me of when I got my SR's - I wanted to compare them (for noise) with my hicap and LT1086's, so rigged up a 'listening device'. I used a low-noise mic input on a mixing desk and monitored with headphones.

I remember you doing that!

It's an interesting thing to do with devices that place a more variable load on the PSU, you can hear music on the PSU rails sometimes.

Sometimes it's distorted horrible sounding music.

Hmm......

:cool:

Andy.

laverda
14-01-05, 07:42 AM
Sorry to intrude:-

I'm surprised the E-I and E-II threads are still there

Think Positive James, Positive.

I would imagine hundreds of budding fellow DIYers are following your superb PFM Speaker project and are amazed at your DIY skills. I can't help but think they can only sound superb and they look the business to boot.

Others and myself have gleened much from all you Guru types via these pages and wish to learn more.

James, keep it up

Back to the thread then

Carry on gents.

Graham

chiily
14-01-05, 09:27 AM
Originally posted by laverda

Others and myself have gleened much from all you Guru types via these pages and wish to learn more.

Graham

Seconded. I now have a collection of boxes that produce the most amazing sounding music, my grandkids adore it!

I would never have known how or bothered before finding PFM and the knowledgable people here.

Hat's off to you all...

Andrew L Weekes
14-01-05, 10:35 AM
Seconded. I now have a collection of boxes that produce the most amazing sounding music, my grandkids adore it!

Yes, you shouldn't lose site of the fact that despite the title of this thread, the best way to assess your PSU is to use it, and listen to music.

It is after all why we do it!

Andy.