It is not just the noise on the output lines. It is also the noise radiated in all directions from the supply location and the noise back into the power line.
My measurements not yet published show a cleaned up switcher can be a good approach. (Not likely to be published any time soon, rather busy.)
My measurements not yet published show a cleaned up switcher can be a good approach. (Not likely to be published any time soon, rather busy.)
I prefer to lowpass filter an SMPS output, with at least -60dB attenuation at its switching frequency and harmonics thereof. Luckily these are high frequencies so a two pole RLC filter with Q<0.5 will do it easily, using (a) capacitance values typically found in PSUs, and (b) relatively low inductances, lower than the ones in loudspeaker crossovers (microhenries not millihenries).
Doing that you reduce differential rf-noise between DC-output terminals, which on my opinion is no problem under most circumstances.
There are other noise leaks not covered by such filters like common mode noise or noise that is coupled magnetically to the environment and picked up by magnetic loops.
Not recommended for audio are all kind of class 2 ac-adapters ("wall warts") due to their hi leakage currents that produce 100/120Hz noise in sensible analog circuitry.
There are other noise leaks not covered by such filters like common mode noise or noise that is coupled magnetically to the environment and picked up by magnetic loops.
Not recommended for audio are all kind of class 2 ac-adapters ("wall warts") due to their hi leakage currents that produce 100/120Hz noise in sensible analog circuitry.
My lab has a 45W laptop SMPS powering 20W in total of white LEDs as lighting.
When I test headphone amplifier circuits on the bench, I always see extra noise on the scope when the SMPS is on.
The amp itself uses its own Tx and linear regulators.
Everything is properly cased and shielded.
I have a hard time believing in no extra noise from a SMPS.
I know that Jan's Silent Switcher is supposed to be real quiet.
And one can design circuits with high PSRR.
But PSRR at >300kHz is not quite the same as at 100Hz.
Patrick
When I test headphone amplifier circuits on the bench, I always see extra noise on the scope when the SMPS is on.
The amp itself uses its own Tx and linear regulators.
Everything is properly cased and shielded.
I have a hard time believing in no extra noise from a SMPS.
I know that Jan's Silent Switcher is supposed to be real quiet.
And one can design circuits with high PSRR.
But PSRR at >300kHz is not quite the same as at 100Hz.
Patrick
Patrick, Not sure anyone said 'no extra noise' . I just said that the THD+N over the audio band tracked almost exactly. If there was noise at several hundred kHz it was not affecting the audio band, the source or the AP that was doing the measurements. Sample of one I realise but a real comparative measurement nonetheless and posted on here. Just checked and TI don't publish PSRR at 300kHz for the LME49710 and I wouldn't want to interpolate from the 20k numbers.
I don't see SMPS as evil, but neither do I think glomming one in where one was not supposed to be is a particularly sensible idea
.
I don't see SMPS as evil, but neither do I think glomming one in where one was not supposed to be is a particularly sensible idea
.
I believe a lot of the noise comes from switchers below 100W (? Might be higher), because at or above this level the EU requires PFC.
Without power factor correction there is a lot more rectifier noise coupled through from the mains rectification.
In addition, below this level there is a tendency to use diodes instead of synchronous rectification which leads to more loss and noise. Snubber (Zobel) design is critical for most switchers and I've seen few in CFLs and LED lamps I have disassembled. I haven't stripped many wall worts.
Without power factor correction there is a lot more rectifier noise coupled through from the mains rectification.
In addition, below this level there is a tendency to use diodes instead of synchronous rectification which leads to more loss and noise. Snubber (Zobel) design is critical for most switchers and I've seen few in CFLs and LED lamps I have disassembled. I haven't stripped many wall worts.
And yet all $$$ T&M gear from Tektronix / Agilent (Keysight) etc. are using SMPS now. DC to 30 GHz+ spectrum analyzers, scopes, VNAs, etc...
It is a solvable problem. Just because your laptop supply is junk doesn't mean that all designs are.
I wanted to see what it did, does and tests and sounds. It was the chosen one on forum here at DIYaudio.
The SMPS was not inside chassis, near circuitry. The noise was wide band.... as the analyzer with LP and HP still had 10X more noise.
The cost must be competitive with non-smps costs. I'm not going to attempt desiging a SMPS of the quality I want. So, if anyone has a recommendation for +/- 60v at high current for PA, let me know and I'll buy and test and report.
Part of the problem is that the wide band noise on spkr term with input shorted is about 10 microvolts with analog PS. And, the thd+N is in the .000x %. So, little noise can be tolerated.
THx-RNMarsh
Last edited:
'5 pin' op amp PSRR is a function loop gain
since its not externally compensated, dosen't have a gnd pin - then the worst case high frequnecy PSRR follows excess loop gain curve (the rail that the intenal Ccomp is referenced to)
I had to go to the LME49720 DS for a gain&phase plot - its close to single pole below 1 MHz
since its not externally compensated, dosen't have a gnd pin - then the worst case high frequnecy PSRR follows excess loop gain curve (the rail that the intenal Ccomp is referenced to)
I had to go to the LME49720 DS for a gain&phase plot - its close to single pole below 1 MHz
Richard,
Delta. Eltek, UniPower etc all make 48 VDC Löw emi Telecom Power supplies parrallelable up to several thousand amps. All PF corrected low noise input 60 VDC ungrounded output Which can be series’s connected for +/- supplies. 50 amps a rail for about $500 apiece
The last version we built did +/- 135 VDC @400 amps < 100 mv ripple at full load. It was about $100k — considerably more expensive than commercial Meanwell type units.
Delta. Eltek, UniPower etc all make 48 VDC Löw emi Telecom Power supplies parrallelable up to several thousand amps. All PF corrected low noise input 60 VDC ungrounded output Which can be series’s connected for +/- supplies. 50 amps a rail for about $500 apiece
The last version we built did +/- 135 VDC @400 amps < 100 mv ripple at full load. It was about $100k — considerably more expensive than commercial Meanwell type units.
Last edited:
Richard, maybe you can tray the one from Hypex. Hypex Electronics webshop
I did not try it (to expensive for me) but could be good one, look for specification, protection included (even possible to add external DC offset loudspeaker protection)
BR Damir
It is a solvable problem. The process is called electronic design & engineering ;-)
I remember chatting up Jeff Rowland at an RMAF a few years ago about his new super duper $ 70k monoblock power amp. The amp uses custom nCore amplifier modules. Asking him about the psu, expecting he would also use one of Bruno's there, he surprised me saying that they developed an in-house SMPS because they had requirements for low noise etc that were not available on the market.
And look what Benchmark does with SMPS in their latest power amp. State of the art cleanliness if there ever was one.
After the SilentSwitcher I would really, really like to try my hand at a similar unit for power amps. But it would be a big project and I am not at all sure yet I have the required expertise there.
Jan
Last edited:
There are three Meanwell SMPS (SP100-27) with PFC in my system with no audio problem but I am deaf.
TV, FM-AM reception works fine though.
Wall worts and some laptop power packs can be real enemies
Bill, any link please?
George
TV, FM-AM reception works fine though.
Wall worts and some laptop power packs can be real enemies
Bill, any link please?
George
Of course, it is solvable. But, not by me. I have zero interest in designing a smps. I own a BenchMark or rather company does here in Bangkok. With a check on IC p/n for controller chip and a few other clues, I can figure the basic topology they used. But that doesnt really help me all that much. The distortion et al of the amp isnt dependant on the PS. But it is very compact and light weight ,,,, but cost more than it should due no doubt to the PS.
I'll try some of the others suggested above.
THx-RNMarsh
Last edited:
Sorry, what is it then according to your interpretation of the measurement I provided?
0.1% @ 50 Hz is better than the M2. By much.
- Status
- Not open for further replies.