PO89ZB, an inline DC filter for SMPS wall warts. Preamps, HPA, Korg NuTube, etc

OK - thanks!

My primary application will be between a Mean Well RS-15-15

and a LT3045 based regulator. Shall I be vary of oscillations?

//

It seems that the input impedance of the regulator is determined (mostly) by the cap on the input -- an aluminum electrolytic of the garden variety won't break the boundary condition.

I think it's worth reminding ourselves of something we already know: very, VERY few of the target devices mentioned in the title of this thread ("preamps, HPA, Korg NuTube, etc") feed the output of a regulated SMPS wall-wart, to the input of a linear regulator. The two largest elephants in the room, namely the Nelson Pass ACP+ and the Nelson Pass B1 Korg NuTube, certainly do not. They represent the vast majority of field deployments of PO89ZB. Instead, devices designed around a regulated SMPS like ACP+ and B1 Korg, tend to have a passive RCRCRC filter at the DC input, and no linear regulator at all.

With no downstream regulator, there's no possibility of regulator instability (oscillation), and the Middlebrook stability criterion simply doesn't apply.

Agreed, the conditions under which it would appear to be important would be those for which the input impedance of the receiving stage is very, very low. The electrolytics shown ESR around 45 milliOhms.

Nevertheless, it's important to remember that the value of Mark's device is that it reduces the unintended "conduction" in BJT devices which is created by those nasty di/dt spikes radiating all over the chassis, causing gastric distress for the listener. The first time I used an SMPS for an HPA I was delighted that worked. After listening I was in listener agony which went away when I switched to a linear supply.
 
Last edited:
I don't know what people prefer, but as far as I am concerned...

For decades I went a pure way, that is no (in fact one per channel for safety) DC coupling cap in my entire HIFI system, no active pre as no need for gain, no op amps and no SMPS etc. A bit stupidly hardcore wire with gain without even considering applications, more based on experience on products I gave a listen to, not realy comparing apples with apples.

In 2002 I built my DAC and at the end I had to admit that the best output stage, that was still convenient and sounding extremly good, was not tubes or solid transistors or output transformers… but, shame on me, then new op amps (AD825) biased in Class A (back then I took what I had lying around and that was, shocking, 2SK170!) And, I still have that DAC and output stage, I resisted ripping its 2SK170 off.


Meanwhile I learned a bit, became more flexible, and today I listen to music through a B1 Korg preamplifier that has many caps in the signal path, a useless gain (as I attenuate to amplify again but at the end have much lower voltage at the output than at the input). But it sounds great. My last RIAA has tiny LDOs (LT3045 etc.) and op amps…

Further, I tried many PS for the little B1K, linear regulated and SMPS, and at the end the difference were quite small thanks to the layout designed by Papa. I believe I have come to a point where a good small wall-wart SMPS coupled the Mark’s excellent filter and going into Papa’s 2 PI filters with upgraded PS caps is probably 99% of what is achievable and improving on that is getting meaningless as possibly other areas should be addressed first with much more sonic benefits.

So today I believe there is no rule, it all depends on application, and also nothing is fixed for ever as parts improve massively so what was true yesterday may not still be true today. An op amp such as OPA 1656 is doing an amazing job in absolute terms… A good SMPS with an excellent filter is doing an amazing job, and for that great thanks to Mark. It is very useful and an eye opener to me. Indeed, I am now even considering a good SMPS + filter for a Class D amp (also something I would have discarted not long ago by principle), and so far the good SMPS is already doing an acceptable job for ridiculous money, but of course the only question is : can it deliver asolute sound quality and compete with the best ? If yes, then I will even think about a SMPS possibility for my big main amplifier, a hopefully very high quality amp by Fred (FSSA-2). With a filter. I am still more familiar (and confident) with big solid tranny and CRCRC filters… but who knows what decision I will have to take should the SMPS + Filter sound better, as that’s the entire point at the end. Time will tell…

Bottom line for my own case : only idiots don’t change minds, and always stay openminded as at the end only the result matters, and that is very application / layout dependant, and also time dependant as parts/units can get better. I am today unable to make a clear cut and say linear is always good, SMPS is always bad, I even have probably SMPS beating whatever at least in one application. Argh ! But the nit sounds indeed better, so…

Claude
 
The difference was quite small you say but was it in favor of the linear PSU? Just yes/no will do.

I agree with what you say and it is not about topology but in the case of SMPS I have yet to meet the first low power device that benefits from it. Regardless which brand or who designed it which, as we can see, creates strong bias. There are many reasons to mention which are in favor of SMPS but clean power is not one of them (otherwise this filter would have no purpose...).

Since Covid started I have built around 12 PSU's all linear/LDO with various regulators. These were to replace the existing SMPS of DACs/streamers/SBC etc. as I noticed stray RF and mains pollution which is old news I guess. In all cases the results were better. However I found out I have to put much care in GND/PE scheme as leakage is an issue also in LPS. I can not get away with floating GND with many a transformer. Those that did not satisfy at first often needed to be referenced to PE to have optimal results.
 
Last edited:
Quote "The difference was quite small you say but was it in favor of the linear PSU? Just yes/no will do."
=> NO. (short version). It was overall in favour of the SMPS.

Longer version... Jean-Paul, as you can read more in details in the B1K thread, the huge linear lab PS and the small SMPS wall-wart were quite close on my modded B1K (with uprated PS caps). Again, no argument that junks comes out of SMPS, as can be measured, but Papa seemed with its radical PS filtering integrated in the B1K (and perhaps this device intrinsecly) to have dealt with mostly. Back to the comparo, the linear performed a bit better but only in the bass, but at the cost of homogeneity and treble opening etc.. Overall the SMPS realised the best compromised to my ears, more musical and airness, that was very unexpected. But indeed still a small compromise somewhere as not best scoring in all areas. In fact each PS or PS + filtering combo seems to add something and lose something in absolute terms... so far. I left it with the best sounding compromise, low ripple SMPS, wondering if one day I could gain a little bit in the bass... and then completely forgot about it.

Now that, triggered by this thread, I added Mark's filter in front of 2 SMPS for different devices, the SMPS on the B1K has all the benefits of the big linear (with or without additional caps) and some more: it is no more a matter of compromise in regards to what I had tried, it clearly now scores better and very high in all areas. Low ripple slightly overdimensioned (read better quality) SMPS with Mark's filteri is getting very very good (measurements seem indded to indicate that, but at the end my ears matter... to me 🙂.

So perhaps Mark's filter was what small SMPS needed for us music listener? Therefore I find it deserves a special mention. Not that it is a world of changes, but the long waited additional step. And now I am wondering about large size SMPS...

So I started to be a small believer (one that is now opening an eye), and from where I come from (I am tweaking a Class D amp with SMPS for a friend, I personaly still discard this technology for myself!) that's a majour step.

Last but not least, I have on all my devices PS (but on the B1K as useless) a separate passive main 230V filter module. These are Conrad kits I use since decades with 2nd order slope and cut off at 1.6kHz or so. That didn't make a change in the B1K PS case, admittely perhaps the junk that its SMPS sputs out is filtered before entering my other devices PS.

That Conrad filter is at the input, before any PS. However, Marks filter acts at the output of the SMPS, it seems to make things much cleaner and the traces seen here on the oscilloscope indicate indeed a very clean DC output. OK, that doesn't say it all re dynamics etc., a battery also has a very clean output and yet can have dynamic / impedance issues with no other help (caps or else) at its output, but still nice curves and listening experience, so...

Let us know how you find your SMPS and filter experiments, interesting new field to play with!

Sorry for the long reply

Claude
 
So YES a low noise linear PSU is better but as the device designed by your father uses heavy filtering the SMPS is not an issue. Still the filter by Mark is needed if one measures the SMPS.

As expected really 🙂
 
Last edited:
Papa's design is indeed probably such that it tends to make PS qualities less relevant, probably due to the strong double PI filter, large caps and possibly other things due to circuit itself and its parts. I understand that's what he wanted.

Whatever, at the end still the SMPS won vs linear, so it has also some intrinsec sonic qualities, but wasn't perfect. Now the SMPS + Mark's filter is really close to the perfection (the way I hear and imagine it), anyway enough for me to close that topic on that specific B1 Korg unless someone comes up with a revolutionary finding... Note I am not outruling that a very fine tuned linear PS wouldn't get there aswell, but please note my reverse of thinking now: the SMPS +filter is my reference unless proved otherwise, and given the result even if it gets beaten it will very likely be marginal or academic at least on B1 Korg

My surprise was on the other set up, to see that a cheap SMPS + filter gave also VG results on the first test, with as reported a basic set up without serious onboard filtering, admitelly around op amps that do have some PS rejection qualities. So the filter is doing a VG job.

It hurts me, but it appears quality SMPS + filter is a qualitative alternative choice to a VG linear PS on some applications... that's my conclusion. Indeed, a filter à la Mark is key, but then so are caps in a linear PS, so each has its particularities and can be good or average ... what matters to me is the overall sonic result and in my mind SMPS+filter is now an option to try (together with top ranking linear PS).

Give SMPS + Mark's filter a try on one of your PS sensitive devices (say not B1 Korg) and let us know, alqays interesting!

Meanwhile I an curious enough to have build my filter for a power amp's SMPS, that is now testing SMPS qualities for power application, someting I didn't even consider a few weeks ago. New field for me, starting with Class D as their high speed current needs could indeed advantage SMPS and make sense. Who knows...

Claude

Claude
 
Yes I will of course stay busy with PSU design (and linear versus SMPS) and a PO89ZB PCB is on its way. Measuring and listening fill follow. It will be tested on sources that have SMPS as a B1 Korg will never enter my home.
 
Last edited:
The other elephant in the room, with many hundreds of DIY units built and well-loved, is the Nelson Pass Amp Camp Preamp+ (written ACP+). LINK 1 ... LINK 2

The "+" Plus means, not only is it a preamp, it's also a headphone amplifier. Indeed it is a Class-A headphone amplifier with 0.15 amperes of bias current in its output stage.

Schematic is attached below.

_
 

Attachments

  • ACP_plus_schem.png
    ACP_plus_schem.png
    181 KB · Views: 806
I have built 6 of these for my 2 power boxes which contains 3 each of the RS-15-15 Meanwell SMPS units.

Using ohm meter:

- All 6 measured the same dc resistance in-out
- 4 measured 1,2M between in and in-ground, 1 measured 960k and 1 measured 550k.
- 4 measured 2,3M between out and out-ground, 1 measured 1,2 and 1 measured 1,5.

These feed my soekris DACs for +/-5V analog and +5V digital.

The inductors do have a marking - is that significant - I did not align all the same way.

My SQ impressions:

- Darker sound (good, probably less rubbish upwards involved)
- More harmonics and texture/layering. I like when this happen, you actually take away something (noise) and get better back rather than adding something to get a thrill.

I'm satisfied - be it imagined/placebo 🙂 I will measure one day... maybe tomorrow.

Thanks for the effort and result Mark!

//
 

Attachments

  • filt.jpg
    filt.jpg
    306 KB · Views: 707
Last edited:
TNT, I think you'll probably get even better results if you use the Kemet electrolytic capacitors called for in the Parts List (post #1 of this thread). Your photo contains clues which suggest your capacitors are different:

A. On your capacitors the white stripe indicating the negative terminal, has black chevron symbols ">". On the Kemet capacitors it is a solid white bar.

B. On your capacitors the text "50V" is next to the white stripe. On the Kemet capacitors the text "470 uF" is next to the white stripe.

C. Your capacitors don't appear to include the KEMET logo​

I suggest you buy at least one of the Kemet capacitors, and then use your most sensitive test instrument to measure the ESR of the Kemet capacitor AND the ESR of your capacitor. I predict you'll find that the Kemet's ESR is significantly lower. Which means the Kemet will do a better job of shorting the (unwanted!!) high frequency noise, to ground. And that's the goal: to short high frequency noise to ground.

If the Kemet does measure significantly better than your alternative cap, then you can decide whether or not you want to retrofit Kemet capacitors into some or all of your existing PCBs. Or to inventory a bunch of Kemets in anticipation of future builds.

The instrument I happened to use for this project was an ET4510 LCR Meter, but if you've got an HP or other sensitive piece of gear, by all means use that.

Click on the image below to remove the foreshortening distortion, and click again, on the white "X" at lower left, to see the image at full size.
_
 

Attachments

  • electrolytic_for_filter.JPG
    electrolytic_for_filter.JPG
    436.5 KB · Views: 672
Perhaps it is not just the stated ESR at a given frequency (with other fixed parameters), but an entire ESR enveloppe given multiple parameters (heat, voltage), or at least a given ESR curve vs frequency (at given constant parameters) that are relevant, who knows...
 
Maybe there is no big difference between manufacturers (??) but the data suggests there certainly IS a big difference between the product lines sold by each individual manufacturer. Here is Nichicon, who serves as an example.

Those three-letter prefixes sure do seem to separate capacitors into different functional groups, including low-ESR versus high-ESR. But maybe I have gotten it wrong after studying the datasheets of 20 or 30 different Nichicon product families ...?

_
 

Attachments

  • capacitors.PNG
    capacitors.PNG
    467.8 KB · Views: 592
Yes, "maybe" you got it wrong... but then it must be "a very shy low volume coming from the back of the schoolclass" maybe 😀

What fab little gift to the community, thanks again Mark for all the trouble you went picking the right combo with components

Claude