Using a SMPS in Audio Applications

[Amphiprion]

Recently I've been designing a linear regulated supply for op amp ckts as many have read in other threads. One thing that's been ticking me off is the gross inefficiency of these things. It sucks! But I never gave much thought to SMPS in audio apps. They always spec something like 1% p-p ripple, which I thought was totally unacceptable in audio applications. And using a post regulator after a switcher just doesn't appeal to me, especially since their RR ratio >100KHz is poor and getting poorer at a steady rate. However....

At work the other day, I put the scope on a 50W Astrodyne isolated DC/DC converter (half brick). I was amazed at what I saw. In scope mode, it had about 100mv of ripple at 200MHz measurement bandwidth, about half that with a 20MHz bandwidth limit, and virtually NOTHING, maybe 1mv ripple, with a 10kHz bandwidth limit. (Those limits are the only options on my scope, a Fluke 199C). Looking at the waveform, it had noise bursts every 2.5 microseconds (switching frequency of 400kHz I guess). These bursts were very, VERY noisy, and died down within about 1uS. Putting the scope in spectrum analyzer mode, however, showed most of the noise up in the 10-30MHz bandwidth, which I thought was interesting - nothing really at 400kHz. Now I don't totally trust the spectrum analyzer of this scope, as it once told me a high quality 16MHz crystal was operating at 16.7MHz (though it did show a nicely decreasing sequence of harmonics). But I do trust it to get things generally in the right range I guess. (Putting it back into scope mode and telling it to measure frequency gave a dead on 16.0MHz for the crystal).

Anyway, I was wondering what I need to look at with SMPS to determine suitability for audio. Do I care anything about noise >20kHz? >100kHz? Up to the GBW product for the opamps I am using? How much can be dealt with using bypass capacitors of a wisely chosen size?

I have a smaller +/-15V Astrodyne coming in early next week for the same project and I am sure going to measure the hell out of it. These are just a little pricey at 30-40 a piece, but man are they tiny, and the ripple in the 20KHz bandwidth could effectively be nothing.

Also, if you SMPS guys have rec's for companies other than Astrodyne, I would be totally open to suggestions. That's the brand we frequently use at work so that's what I can easily get experience with without spending my own money.

Also, I should mention I am very much being an engineer and not an audiophile about this - I do care very much about sound, but if the measurements tell me this is satisfactory, I am going to run with it.

[Amphiprion]

Also I should mention this DC/DC converter (24->24V used mainly for isolation but also for its wide voltage input range) was powering a slew of other DC/DC converters - two Astrodyne 5V modules that needed to be isolated from one another for different parts of the ckt, and a 500V Ultravolt DC/DC converter module. So those might have been throwing noise back on their inputs, adding to whatever noise the 50W astrodyne was putting out. I will try to measure converters on their own in the future.

Also, the +/-15V module I am getting this week is a 5 watt model, FDC05-24D15. The spec page is here:

http://datasheet.astrodyne.com/FDC05.pdf

[Dennis H]

QSC's best power amps use SMPS. They're lighter, electrically quieter and more expensive than the models with linear supplies. You can download the schematics. I know the PLX series use SMPS as do several others but I'm not sure which ones.

http://www.qscaudio.com/support/technical_support/schems1.htm

Edit: the smaller models have a single rail voltage and the bigger ones have dual rail voltage, switch to the higher voltage when more power is demanded.

[Amphiprion]

I am mostly concerned about how the high frequency noise might interact with opamps, since their PSRR up high is not good. In a power amplifier like QSC, where the overall closed loop response is going to roll off rapidly >20KHz, an opamp stage might have a closed loop response that runs up into the MHz range. I'm wondering how noise in the hundreds of kHz to 10-20MHz range might interact with them.

But I was shocked how dead silent the Astrodyne was up to 10KHz.

[Dennis H]

SL uses a 10 nF cap to ground physically located right at the opamp. Intuition says that should kill the RF before it gets into the opamp but RF is a strange and mysterious thing to me so we'd better hear what JonMarsh has to say.

[Amphiprion]

RF is strange to me too When you add in ESL effects of the cap it gets more interesting too. Would a 10nF NP0 1206 ceramic be better than an 0805 or 0402 X7R? Are NP0's of any value in bypass apps, or should I go with the low path length? The questions that plague me....

[joecarrow]

As far as I know, the NP0's main advantage is that it's more temperature stable. Since it's being used for bypass and does not participate in an oscillator or a critical filter, then the variation in capacitance with temperature won't matter. These days most SMT capacitors have such low ESL as to make smaller parallel bypass caps fairly superfluous. I think you'd be hard pressed to find a useful difference between these two caps for this application, other than the 0402 being a royal pain to hand solder.

Definitely pay attention if JonMarsh drops in on this thread. He's the guy to go to, without a doubt.

I think that a good approach would be to try to clamp down on the noise as close to its source as possible, so as to minimize antenna effects, and also to bypass again at the opamps. The main thing to worry about (keeping in mind that I'm just a MechE with over-the-shoulder exposure to thsi stuff) is if the noise introduces intermodulation distortion or subharmonics (beat frequencies from oscillation).

The gain of the opamps tends to go down with frequency, so that can help as well. If the opamps need any special considerations, then it should be listed in the data sheet. I remember our guys (when I was at the place with electronics) were working with opamps that were so sensitive that they needed unconnected traces surrounding the inputs to block any fiberglass borne currents. When the new EE started looking at the board design, he though all of the non connected traces were sloppy mumbo-jumbo... until he read the data sheet.

[Dennis H]

I think SL also uses a big cap to ground right where the PS plugs into the PCB. Belt and suspenders anyone?

[noah katz]

Check out www. coldamp.com

[Amphiprion]

The two main problems with non-NP0 ceramics that I am aware of are the non zero temperature coefficient, as well as being microphonic/piezoelectric in the 30Hz to 30kHz region. Any ceramic doped with barium has these properties IIRC.

[mefistofelez]

Dear Mark,

have you made any progress/conclusions?

Kindest regards,

M

[Amphiprion]

I just started playing with the +/-15V Astrodyne today after work. It's a little noisy, I'm not too impressed at first. But the noise again is almost all at very high frequency. I tried a few caps to bypass the output pins but none of them really gave more than maybe 15% reduction in noise. I should have the chance to stick it on a PCB with an MLCC later this week, assuming my PCB's and other parts get here Friday as scheduled. I'm also wishing for a better spectrum analyzer than what the Fluke has to offer, but I don't think I'm going to be able to get any department to loan my group the dough. One of these days my group will start landing its own SBIR grants and we'll be able to buy our own toys instead of mooch.

[Amphiprion]

I have the Astrodyne and all the analog electronics populated on the bare bones PCB I just received Friday. It is a very simple circuit - a whopping three dual op-amps. The circuit consists of two gain stages preceeded by a simple RC filter @ 300Hz to de-noise the signal. One gain stage is inverting and the other non, to amplify a -72mV signal to a 0-5V range. Reason for the two gain stages is to use the first non inverting section to present a very high input impedance to the source, while the inverting gets me a positive voltage. Signal is hit then by a 4th order LP filter @ 100Hz. Then hit with a difference amp, followed by a summing amp. This basically makes a feedback loop around a constant voltage power source with a voltage adjust pin that also has an Iout monitor pin to turn it into a constant current source supply, that varies output voltage from 0-6kV to maintain a set current level between 0-5mA. Major concerns are how exactly this will interact with the HV supply and whether or not it will be stable. The breadboard version was very encouraging.

Anyway, I took the advantage to use some reasonably hifi parts instead of cheap crap. AD712J, mylar caps, 1% MF resistors. Not truly hifi, but at least there aren't ceramics in the signal path or uA741's anywhere Anyway, once I get the darn thing working (if it works) I'll play around with how the astrodyne works. I might bypass all the low pass elements and look at what the ckt does with a signal generator of the 20-20kHz bandwidth.

Anyway, work comes first with this guy, but I'll be playing around with it to evaluate the supply too.

[engr_dave]

a linear regulated supply for op amp ckts as many have read in other threads. One thing that's been ticking me off is the gross inefficiency of these things

I must be missing something here. Are you running off of batteries? Otherwise, does it matter?

Assuming that you don't have to live with large input voltage variation (110Vac line at my house is very stable) an ordinary transformer-fed linear regulator supply should be at least 60% efficient. If you're feeding a half watt or so to a handful of op-amps @ +/-15vdc, you're only wasting a few tenths in the supply. Doesn't seem to warrant designing an SMPS, especially considering their attendant noise, PSRR, and RF demodulation issues you are correctly concerned about.

[joecarrow]

Considering that he's doing this at work and only bringing audio into it as a side thing, it wouldn't surprise me if this was going into some sort of portable device.

How about it, Mark? Is it something we'd drool over?