advertisement


PSU Noise Question

N

nojobtoosimple

Active Member
Hi all,
I have a NAC72 with some RSL after-market boards fitted and powered by a FlatCap. My mind has been turning lately to improving the power supply though, honestly, I have no real complaints about the sounds I'm hearing I just fancy building something and the FlatCap looks like it might be a weak point.

Looking at the RSL cards more closely I note that they have onboard regulation using one or more LT3082s. Looking at the spec sheets for this device it quotes Error Amplifier noise of 33uV and PSRR of around -65db to -90db, depending on the frequency of the noise, in the audio band.

So here's the first question... given the foregoing, to what extent should I strive for a lowest possible noise PSU solution ? It seems to me (and please be aware that I am at or beyond the bounds of my understanding here) that, unless the PSU is delivering millivolts of noise, the input noise level will result in additional output noise that is dwarfed by the 33uV produced by the LT3082 itself. Am I on the right lines here ? If so, then selecting a PSU design simply because it should produce a couple of uV less noise than another would seem a bit bonkers.

I am (dimly) aware that there are many other factors in PSU performance that play a part in determining how well they perform in any given situation but assume that some, at least, will be rendered less relevant by the onboard regulation of the RSL cards. In this context which of these factors are likely to have a greater influence on the performance of the NAC and should therefore be prioritised when selecting components or approaches for a new PSU ?

I'll be grateful for any advice, guidance or comment offered.

Thanks,
Jon
 
You are on the right lines yes. I consider PSU noise from regulators a total non issue apart from when powering MC stages/input stage of MC stage when it then becomes critical. In all other applications the noise from any IC reg I can think of, even the oldest and crudest like LM317 or 7815, is inaudible IME.
 
Thanks Arkless.

Any thoughts on what I should prioritise ?

Naim deploy huge traffos and big reservoir caps in their bigger PSUs, Teddy Pardo and RSL rather less so; I'm not sure what Avondale do. Are these approaches reflective of the type of regulation used in these PSUs, or just different design philosophy ? That is, if I go for a simple approach (eg LM317 or 7824) should I also adopt the big traffo & caps or is there good reason to think that somewhat smaller traffo & caps will essentially perform just as well ?

Thanks again,
Jon
 
If the RSL boards have on board regulation, a FlatCap may be as good as it gets. As you correctly point out, unless the shared supply was really noisy, the on-board regulators will dominate, and the FlatCap is good enough.

The only marginal upgrade I can think of, which is firmly in the harmless lily gilding area, is to add an CRC filter network before the Flatcap regulator. The on-board regulator, and the main one in the Flatcap, both have performance that falls off with frequency, and the raw rectified waveform is a sawtooth with components to high frequencies. Adding a bit of RC filtering will eat those up.

Arkless will now remind us that even at 20kHz, the cascaded regulators will give you something like 100dB of PSRR, plus the gain circuits themselves will have some PSRR (30dB? more), and so we are looking at residual levels way, way, way under the noise floor.
 
PigletsDad said:
If the RSL boards have on board regulation, a FlatCap may be as good as it gets. As you correctly point out, unless the shared supply was really noisy, the on-board regulators will dominate, and the FlatCap is good enough.

The only marginal upgrade I can think of, which is firmly in the harmless lily gilding area, is to add an CRC filter network before the Flatcap regulator. The on-board regulator, and the main one in the Flatcap, both have performance that falls off with frequency, and the raw rectified waveform is a sawtooth with components to high frequencies. Adding a bit of RC filtering will eat those up.

Arkless will now remind us that even at 20kHz, the cascaded regulators will give you something like 100dB of PSRR, plus the gain circuits themselves will have some PSRR (30dB? more), and so we are looking at residual levels way, way, way under the noise floor.
Click to expand...

You saved me the effort though:)
 
nojobtoosimple said:
Thanks Arkless.

Any thoughts on what I should prioritise ?

Naim deploy huge traffos and big reservoir caps in their bigger PSUs, Teddy Pardo and RSL rather less so; I'm not sure what Avondale do. Are these approaches reflective of the type of regulation used in these PSUs, or just different design philosophy ? That is, if I go for a simple approach (eg LM317 or 7824) should I also adopt the big traffo & caps or is there good reason to think that somewhat smaller traffo & caps will essentially perform just as well ?

Thanks again,
Jon
Click to expand...

It's all marketing foo and makes no difference whatsoever. Far easier to sell an extortionately over priced PSU if you can give it "feel the weight of that!"....
A typical pre amp or phono stage uses 10W or less of power and a 50VA transformer is way more than adequate never mind the 500VA etc ones some of these companies use. Obviously I must be wrong though as little 'ol me couldn't possibly be right when companies like Naim, Exposure, Avondale etc etc all say the opposite:rolleyes::rolleyes::rolleyes::rolleyes::rolleyes::rolleyes::rolleyes:

As PD says CRC filtering is very effective. At the power levels of a pre you can use relatively large R's (4.7 - 47R type range depending on current draw and hence voltage drop across the R's) with caps in the 3300uF - 15000uF type range and with a C-R-C-R-C arrangement you can get all noise and mains hum down to such low levels that regulators are only really adding a low DC source resistance and rock solid stability of the output voltage.
The mains bounces around in voltage all the time and if you were to look at the output of the C-R-C supply on a scope you see no noise of mains hum but the line on the screen will drift/bounce (depending on time constants of the C-R-C etc) in step with the variations of the mains. A regulator after the C-R-C will give a rock solid line and put back the "stiffness" to the DC supply which has been removed by all those R's in series.... providing the input from the C-R-C supply to the regulator remains above the drop out voltage of the regulator.

I defy anyone to measure even a millionth of a percent difference between the output of the above with a 50VA TX and with a 500VA one! Even more so with cascaded regulators!
 
nojobtoosimple said:
Thanks PigletsDad.

What sort of C & R values might be sensible in this CRC filter ?

Cheers,
Jon
Click to expand...


Keep it quite small. Even adding 1ohm, then 1000uF between the existing reservoir cap and the input of the regulator (yes, needs a track cut and bridged by the resistor; then use the new larger-value cap to replace the 10uF tantalum cap on the input to the LM31x regulator) is powerful - it means noise and ripple is knocked -3dB at 160Hz in round numbers, maybe -20db by 1600Hz - the key midrange. Given the internal gain, therefore regulation of these regulators starts to drop c 800-1khz - that benefit gets 'pasted-on top' across the whole audio bandwidth. Job done.

That's a lot of cheap & effective pre-filtering, esp. when added to the numbers PD notes. Feel free to vary the values say +/-100% or to taste/ what you have/what fits; slightly larger resistor values means more ultimate attenuation, because that is ultimately-limited by the capacitor ESR, but the resistor will run warmer.

But since all this does is pre-filter - no need, nor point, in much more, IME.

My suggestion - use a 1 or 2W metal film resistor, like vishay pro2 series (or anything comparable); and a plain 105degC / 1000uF / 50v electrolytic cap from an industry supplier - Panasonic, Nichicon, Samwha, many others.

HTH.
 
Actually - don't take my word for it - here from Doug Self Small signal audio design: a raw psu with 4700uF reservoirs, feeding a load of 650ma (c 4- 5x what the flatcap will see, so - commensurately- would have about that much more ripple on the raw supply) and adding 2.2ohm/ variable caps to make a C- -C arrangement:

DSelf-650ma%20ripple.jpg


So you see the benefit of 'going larger' is diminishing returns, becasue the regulator itself generates some internal noise. larger value caps on teVadj pin helps, but so does some C-R-C ahead. So the values that get there, ime, are what drive my previous post, esp at the lighter likely load currents; and the thing this table doesn't show, is the way that knock-down of ripple and in-bound external sh*te is excluded, wideband, for all that follows.

Have fun!
 
Thanks, all, for that really helpful info.

I'm going to need to dig around a bit in the FlatCap to work out the details but hopefully I can get it done. In the event of confusion I shall return.:)

Cheers,
Jon
 
At least You should have no problems fitting anything you like into a flatcap as it's an empty box of nothingness o_O
 
From discussing this with Kit and trying a ZapCap I can confirm there is an audible difference between a ZapCap and a TPR4/Hicap driving RSL boards in a 72. Why? Dunno. Not a noise thing I suggest. It is a bass thing. Perhaps improved LF regulation. Gets rid of the TPR4/Hicap bass warmth/hump/bloat (pick your preferred adjective). Worse, discussing this with Kit he assures me there is a gain to had with a third level of regulation: 1) in PSU, 2) in his preamp, 3) then regulation on the cards. YMMV.

Kit recommended and Adam @hifiaf did a RSL PSM in a FlatCap, ZapCap used a smaller transformer and closer to those in the FlatCap than the monster in a HiCap. Sadly RSL don't do the ZapCap or PSM anymore.
 
This has been discussed a few times...and there is a benefit from multiple layers of regulations

I know I've linked this a few times before.. but it may have missed your search on the subject

https://pinkfishmedia.net/forum/threads/help-with-nac72-regs-wiring-please.258625/


Considering my 72 and its PS probably cost no more than £1200 total to buy/build I would happily stack it up against any new preamp upto £3k plus
And if its a naim preamp....make it £6k+
 
Well, I decided to leave the Flatcap alone and build a supply instead. Flatcap mangling may occur in a later project when I plan to build three or four different phono stages to explore that aspect of my sound preferences.

The supply is now complete and incorporates the RC filter discussed above. It sounds great and I'm very happy with it. Photo's attached (hopefully).

Thanks again for the advice.

Jon
52239989064_a1fc159370_k.jpg

52239988984_e16b541080_k.jpg

52239989029_da6a8adaff_k.jpg
 
I believe that 0V and mains earth should be separated if you adhere to the Naim philosophy of only connecting to mains earth at the source.
 
Big topic for another day. Though discussed here before.

IN any system, ideally you only want ONE bond between signal 0v, and Mains Earth.

In any systems of DIY metal boxes holding diy hifi components, you NEED each box bonded to Mains Earth.

Note these are not the same thing , at all.
 
Presumably that is the case (no pun intended) where the transformer is mounted within said metal box. If you have an off-board transformer/psu with a low voltage supply fed to your metal box, then this doesn't need to be bonded back to mains earth?
 
a.palfreyman said:
Presumably that is the case (no pun intended) where the transformer is mounted within said metal box. If you have an off-board transformer/psu with a low voltage supply fed to your metal box, then this doesn't need to be bonded back to mains earth?
Click to expand...
I would still earth the box as it could be made live by a fault in the external transformer or some other part of the system.

I did this with my head amp that has an external tx, I had to buy a length of 4 core mains lead to do it.

Pete
 
You must log in or register to reply here.


advertisement


Back
Top