Mini SMPS Mic Power Supply

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

Bo Deadly

Well-known member
Joined
Dec 22, 2015
Messages
3,266
Location
New Jersey, USA
Here's a mini SMPS mic power supply that I just made:

LilMicPS.png


So this is a Mean-Well APC-16-350 that you can get on Mouser for $7 USD with a custom PCB mounted on top. Overall dimensions are 90 mm x 40 mm x 53 mm.

It makes +18V, -18V and +48V. However, the 48V is relative to chassis ground which is connected to the negative output of the SMPS whereas the +-18 is relative to a simple virtual ground. But this is all that is necessary to bias a bipolar amp and differential line driver.

Here is the schem:

Screenshot%2Bfrom%2B2019-01-16%2B01-41-03.png


Preliminary tests with a simple THAT 1510 / THAT 1646 circuit seem good to me:

Screenshot%2Bfrom%2B2019-01-16%2B01-34-10.png


This plot is of a test tone attenuated 60dB and then amplified 60dB by the THAT 1510 circuit. The test tone peak is 0 dBFS which for my Motu Traveler is +12dBu.

Hopefully someone finds this useful.

At least it should prove without a doubt that SMPS are perfectly suited for pro-audio applications.
 
If your virtual ground shouldn't be connected to ground and pin1 of balanced in and outputs, will the THAT chips work as they should?
Did you measured ripple and noise at +48V output intended for microphone phantom powering?
Did you measured also EMI from SMPS? It is intended for powering LED strings as I know, not for powering a sensitive audio circuits. Can you open it and make a few photos?
Why you use the capacitor multiplier topology, what's the maximum current into virtual ground which will be tolerated?
Shunt regs will work much better here, IMO.
 
moamps said:
what's the maximum current into virtual ground which will be tolerated?
+1. That arrangement is not going to work very well since practically no ground current can be tolerated. A better approach is to use an 18V SMPS plus a LT3757 to generate a negative rail, and a boost converter to generate 48V.
 
moamps said:
If your virtual ground shouldn't be connected to ground and pin1 of balanced in and outputs, will the THAT chips work as they should?
Yes. That frequency response plot does not lie. That is real. I used blocking caps on the output of the 1646. And of course you need blocking caps on the input for 48V anyway. Pin 1 is connected to chassis. So pin 1 / chassis and the virtual ground are only AC coupled through the 470u.

moamps said:
Did you measured ripple and noise at +48V output intended for microphone phantom powering?
Yes. That plot is with phantom power on. I used blocking caps in my attenuator so that I could run the test with phantom on. There is a 100R / 100u RC on the amp board.
moamps said:
Did you measured also EMI from SMPS?
There actually is a small peak at 1.5 kHz that you can see in the plot. That is from the SMPS. If I move the PS outside the enclosure, that peak goes away. However, the enlcosure that I'm putting this into is really small. The PS was ~5 inches away from the sensitive bits. Also, I'm wondering if it's just the SMPS throttling because the load is very low. I still have another 1510 (this is going to be for a dual pre) and a power LED to add so if I see that peak shift up I'll know I can probably get rid of it by adding a dummy load.

[update: yup - after adding load to the SMPS, that small peak is gone.]

moamps said:
It is intended for powering LED strings as I know, not for powering a sensitive audio circuits. Can you open it and make a few photos?
Those Mean Well supplies are known to work well with audio. I have used the Mean Well LPC units in other projects and they work very well. They cannot be opened. They are in sealed plastic housings.
moamps said:
Why you use the capacitor multiplier topology, what's the maximum current into virtual ground which will be tolerated?
Shunt regs will work much better here, IMO.
As stated previously, there can be very little DC current whatsoever with this scheme. It could tolerate a few mA in practice. It would throw off the Vg voltage a little but a volt or two wouldn't matter.

The circuit I'm powering is just a dual 1510 and 1646. So the only things connected to Vg are the 1K input bias / load resistors and ref pin of the 1510s and the GND pin of the 1646s. That's it!

Actually there is a peak LED that dumps current into Vg when the LED turns which is technically a fail. It could pull Vg up in a bad way. But presumably one would not use this in a way that the peak LED turning on all the time.

I have always thought that capacitance multipliers yield the quietest filtering. Regulators can actually be noisier. They are high speed amplifiers after all. In this case a capacitance multiple makes for a simple and effective filter. No doubt things could be improved. One could go active and make a better Vg. This was just quick project to retrofit a simple low noise dual mic pre into an old Sony phantom power supply enclosure so I just needed a Vg to bias the 1510s and 1646s.
 
I think we had a similar discussion some time ago. The low current problem comes from insisting the oV is the same as in a split supply design. If you remove this constraint the circuit will work just as well as the split rail original. So from the 48V just create a single rail equal to twice the split rail voltage. Connect op amps across this rail and bias them from a pot divider set to half the overall rail voltage. All signals are referenced to the 0V of the 48V SMSP 0V. This means you need caps on the inputs and outputs but dc coupling is of dubious value anyway.

Cheers

ian
 
It says "VGND". That means I bond that to chassis and then to dirt, "OBVIOUSLY".

Don't tell me users won't jump to that conclusion.

When jumpered the "obvious" way, not only is the bipolar audio un-biased, we find 38V across a typically "36V process" chip. While neither 36V chips nor e-caps are instant blow-up at 38V, it is still stressful.

And as Ian says, blocking caps are now probably ESSENTIAL at the opamp output. But isn't reduction of coupling caps THE reason we went bi-polar in the first place? (Couldn't be the early audio adopters of opamps were too dumb to question conventions of DC-based Analog Computing.) You really can no-cap out of a bipolar supply amp. You can't with this affair.

Aside from electrical trouble/confusion...
> prove without a doubt that SMPS are perfectly suited

There is always doubt when bench-racing audio. The nature of the topic seems to require say five participants to have more than five opinions.
 

Attachments

  • Image24.gif
    Image24.gif
    42.3 KB · Views: 48
PRR said:
But isn't reduction of coupling caps THE reason we went bi-polar in the first place?

Definitely not. It is a hangover from the original driver of op amp development - missile servos.

Cheers

Ian
 
Every time I've tried to float the secondary of a bipolar linear supply I've run into problems. I now always tie the center tap to chassis. It's been a while since I've tried it so I can't exactly remember what problems I ran into. Isn't it generally a bad idea to have audio common floating? In a large system wouldn't that likely be a cause of a ground potential difference?
 
Gold said:
Isn't it generally a bad idea to have audio common floating? In a large system wouldn't that likely be a cause of a ground potential difference?
Absolutely that would be a bad idea. That's why this circuit doesn't do any of that!
 
squarewave said:
Absolutely that would be a bad idea. That's why this circuit doesn't do any of that!

The warning says don’t connect V ground to chassis.  What is V ground referenced
to ?  AC coupled through C11?
 
Gold said:
The warning says don’t connect V ground to chassis.  What is V ground referenced
to ?  AC coupled through C11?
Correct. DC is set by the voltage divider and AC is set by C11. So it's not "floating".
 

Latest posts

Back
Top