Using Class-D amplifier as a tracking pre-regulator for a power supply

[Kalvin]

There are quite a few cheap +50W Class-D stereo amplifiers available at the eBay. I was thinking about using one of those as a tracking pre-regulator for a simple dual-channel, adjustable, psudo-linear benchtop power supply 24V / 1A per channel. The channels needs to be galvanically isolated so that I can connect them in series or parallel as needed.

The Class-D amplifier would be driven by a low distortion sine wave signal, and the amplifier's output would be driving two 1:1 transformers. At the secondary side of the transformer there will be basic diode rectifiers, filtering components and a simple adjustable linear regulator. The sine wave frequency will be around 20kHz in order to keep the transformer and the filtering component size as small as possible but high enough in order to reduce audible noise. The transformers will provide galvanic isolation and they will also block any residual leakage current from the switching mode power supply driving the Class D amplifier.

There will be an optical feedback from the output to the volume control of each channel which will keep the transformers output voltage 3V higher than the adjusted output voltage ie. tracking pre-regulation.

I haven't been playing yet with the Class D amplifiers, so I do not know what to expect in the terms of switching noise and other possible problems when using the Class D amplifier driving transformers. The transformer itself should attenuate the switching noise somewhat but I do not know what kind of attenuation or behaviour to expect.

Possible problems I have learnt so far:

- There is one application note from TI: http://www.ti.com/lit/an/sloa133/sloa133.pdf which discusses some issues when driving power transformers.
 
- Also I am aware that DC-offset might be a problem causing saturation of the transformers. My first though is to use series capacitor with the windings in order to block the DC-component.

Problems that may bust the project:

- The switching noise from the Class D amplifier after transformer, rectifier and simple filtering will be too high (like tens of millivolts peak-to-peak in hundreds of KHz or more) and the linear voltage regulator cannot filter the components below 500 microvolts p-p or so.

Is this idea totally silly in order to make a low-noise linear power supply with pre-regulator? Any ideas or suggestions?

[Yansi]

Yes, it is a silly idea. I'll leave room for others to explain.

Why don't you use a simple and stupid five-leg stepdown converter like LM2596 and make it do a pre-regulator? Just needs adding a pair of NPN transistors (preferably a matched pair like BCV62) and three resistors. How? Add a current mirror to the FB pin from the output of the LINreg.

If you need lower noise, make the preregulator output filtering better, preferably using a higher switching frequency, than the poor LM2596 haz.

[Kalvin]

Yes, it is an silly idea. I'll leave room for others to explain.

Why don't you use a simple and stupid five-leg stepdown converter like LM2596 and make it do a pre-regulator? Just needs adding a pair of NPN transistors (preferably a matched pair like BCV62) and three resistors. How? Add a current mirror to the FB pin from the output of the LINreg.

I would like to avoid having a switching regulator at the secondary side in order to reduce the switching noise. That's why I am looking at the Class D amplifier as the switching will happen at the primary side of the transformer and the transformer should attenuate the the switching noise. But like I said, I have no idea how much the transformer will attenuate the switching noise. 

[Yansi]

Kalvin: I understand. You will use the 50Hz transformer as a lowpass filter.

Question is, why don't you just slap a filter after the stepdown converter itself? It will be almost exactly the same, but 99% of the problems you will have with the "heavy xformer" solution will be gone.

If the PWM switching noise of the preregulator is a problem, then don't modulate the stepdown with PWM, use sigma-delta instead and make the noise more "out of band" to be more easily filtered.

To be exact, when I have read the title first, I have thought you want to use this idea exactly - use a SD modulated class D directly as a DC voltage source - which of course can be done, just remove all AC coupling from the amplifier itself.

[Yansi]

The solution using classD amp to feed a "chunk of metal" xformer won't be much more efficient either 

Just saying you can throw the xformer completely away from the equation.

[Kalvin]

@Yansi:
My intention is to use the transformer as a low pass filter.
Why not just post-filter the smps pre-regulator? I have considered that. Linear has some nice very low noise switched mode regulators, like:
http://www.linear.com/product/LT1738
http://www.linear.com/product/LT1533
http://www.linear.com/product/LT1683
I just want to avoid designing the switching power stage and use the existing Class D amplifier for that. 

@Blueskull:
The regulation bandwidth doesn't need to be very high if the rectifier has sufficient capacitance at its output to provide the energy needed during the load transients. Good comments on different transformers and their ability to attenuate the switching noise. I wouldn't want to use 50Hz transformer, but like you said, the SMPS transformers designed for tens of kHz frequencies may not filter the switching noise sufficiently. Maybe something in between could be usable.

About the efficiency: The efficiency should be somewhere between a linear regulator and a linear-regulator+switching pre-regulator. Preferable near the pre-rregulator variant.

[Kleinstein]

To reduce the switching noise, there should be a filter between the class D Amplifier and the transformer. There should already some filtering in the amplifier, but usually just enough to meat EMI standards, if at all.
A transformer made for good efficiency at 20 kHz will not filter very much switching noise in the MHz range, at least in differential mode. The good thing with the high switching frequency is, that filters can be quite small in size. However layout may be more important than just the choice of components. For the high frequencies it usually needs several stages, but there are a few places where you might wand inductors anyway (e.g. between transformer and rectifier.

[Yansi]

I still don't get why do you want unnecessary complexity over the straightforward solution. If you aren't sure how to design and lay out simple buck converter, I'm afraid the bloated solution using class D and iron core transformer won't help you anywhere. It just won't happen.

[Mechatrommer]

looking at wiki definition, now i know what i'm building is sorta the said class D amplifier, except the feed is not a nice sine and triangle, but feedback from linear regulator output. i know there are smps switching chips out there i even have bunch of 34063 that i know i can utilize, but i decided to continue to make it discrete solution because i need to learn some aspects of smps. the commonly available smps chip are sort of class B (class B(D) maybe i'm not sure what the name is, because its switching class B push only type, real class B is linear mode) amp incl the 34063, but my discrete solution evolved from class B(D) to class D because pre-regulator's "drain" mode is necessary for overvoltage Vo-Vi LM317 protection, if its not because i'm spec'ing the PSU up to 50V and from 60V input, i will just go with classical class B(D) type, ie pushed (P-MOSFET) smps only. so here is my discrete solution, i'll tell you when i have life to tell about it...

[Zero999]

Yes, it is an silly idea. I'll leave room for others to explain.

Why don't you use a simple and stupid five-leg stepdown converter like LM2596 and make it do a pre-regulator? Just needs adding a pair of NPN transistors (preferably a matched pair like BCV62) and three resistors. How? Add a current mirror to the FB pin from the output of the LINreg.

I would like to avoid having a switching regulator at the secondary side in order to reduce the switching noise. That's why I am looking at the Class D amplifier as the switching will happen at the primary side of the transformer and the transformer should attenuate the the switching noise. But like I said, I have no idea how much the transformer will attenuate the switching noise.

What difference would it make whether the switching occurs on the primary or secondary side?

A class D amplifier is basically doing the same thing as a switched mode power supply so the noise level will be similar. If you need an isolated supply, you might as well drive a transformer from an oscillator. There are even transformer driver oscillator ICs available, depending on your power requirements.

http://cds.linear.com/docs/en/datasheet/3439fs.pdf
http://www.ti.com/lit/ds/sllsea0g/sllsea0g.pdf

[Yansi]

looking at wiki definition, now i know what i'm building is sorta the said class D amplifier, except the feed is not a nice sine and triangle, but feedback from linear regulator output. i know there are smps switching chips out there i even have bunch of 34063 that i know i can utilize, but i decided to continue to make it discrete solution because i need to learn some aspects of smps. the commonly available smps chip are sort of class B (class B(D) maybe i'm not sure what the name is, because its switching class B push only type, real class B is linear mode) amp incl the 34063, but my discrete solution evolved from class B(D) to class D because pre-regulator's "drain" mode is necessary for overvoltage Vo-Vi LM317 protection, if its not because i'm spec'ing the PSU up to 50V and from 60V input, i will just go with classical class B(D) type, ie pushed (P-MOSFET) smps only. so here is my discrete solution, i'll tell you when i have life to tell about it...

Seems your terminology might be a little off, as I have completely lost in what are you doing. The schematic also needs cleaning and polishing. For example the diodes over the mosfets has what purpose?  And why only "PChannel"? Learn how to make a discrete bootsrapped driver instead, it is easy and much more elegant!

[Mechatrommer]

looking at wiki definition...

Seems your terminology might be a little off, as I have completely lost in what are you doing.

ignore the + sign in every net... its to differentiate with -ve tracking regulator (not shown) which has - sign, but they are both symetrical...

Vpd = Vpi / 5;
Vrd = Vro / 5;
Vro = post (linear) regulator output;
Vpi = pre (switching) regulator output = post (linear) regulator input;
if (Vpd - Vrd) > 4V (Vpm) then turn ON Q28 NMOSFET to drain Vpi ; // LM317 protection (not shown) ie if Vpi - Vro > 20V
if Q28 is turned ON, ensure Q26 is turned OFF by using Q4 to avoid driver short

Vpc = Vpa + Vrd;
Vpa = 1 (default) pretrack volt offset so that Vpi = Vro + 5V;
if Vpd < Vpc then turn ON Q26 PMOSFET

that may sound funny, but this is what happened when a mechanical + software inclined guy try to build an analog hardware circuit in one single headskull...

The schematic also needs cleaning and polishing.

thank you for the advice...

For example the diodes over the mosfets has what purpose?

if Vpi goes above V60+ by external force, D13 will conduct. if it goes -ve by external force, then D14 will conduct. both to protect the mosfet driver...

And why only "PChannel"?

i have NMOSFET as drain there... Q28

Learn how to make a discrete bootsrapped driver instead, it is easy and much more elegant!

oh that! i like the simplicity of high side PMOSFET without a bootstrap, + i need to save board space as much as i can, i have tight enclosure, a slight decision change on a tiny function will result in few cm^2 board expansion requirement resulting in the board will not fitting in the enclosure, i already have 4 boards crammed together inside the enclosure, i originally hoped only 1 board for the project but this is the best i can get to, + i have symetrical -ve rail remember? so i'm not sure if there is a symetrical -ve rail NMOSFET bootstrapper chip, or discrete example in the net. discrete? it will double the above mentioned trouble (board expansion problem), so i dont give much thought on that idea... comments appreciated, fwiw...

[dannyf]

Is this idea totally silly in order to make a low-noise linear power supply with pre-regulator?

It is both a good and bad idea, for different reasons.

Conceptually, it is sound: a class D amp is literally a smps, vice versa. So if it is possible to use smps as tracking regulators, it is possible to use a class D amp as such. Minor issues, like the ability to amplify DC, can be tackled with.

Practically, you run into two issues:

1) economics: it is much cheaper to get a laptop regulator that delivers 6amp and 20v, for next to nothing. So why use an amp?

2) noise: a linear regulator will have next to 0 psrr at the kind of frequencies the class D amp runs.

Just so you know, TI published a paper in the late 1990s talking about a class D audio amp that is literally a smps: the audio input is applied to the smps so that it generates a voltage on the speaker that tracks the input audio signal. So the TI guys were as "crazy" as you are today.

[Kalvin]

@Kleinstein:
Good suggestions. Placing a lowpass filter before the primary winding should attenuate the high frequency components before they will enter the transformer. Also some filtering after rectifier will further reduce the remaining noise.

@Hero99:
The primary side switching noise should be attenuated by the transformer. However, I am not sure how much attenuation there will be as I haven't been playing with the transformers in switching applications. I have read the Jim Williams's app note "Some Thoughts on DC/DC Converters" http://cds.linear.com/docs/en/application-note/an29f.pdf Figure 4 which uses a low distortion sine wave and push-pull driver. Using a Class D amplifier to drive the transformer and using a sine wave signal is quite similar idea but with improved efficiency. However, the switching noise will be greater, of course. But I am trying to figure out how much the transformer itself will reduce the switching noise. Possibly I just need to build a prototype in order to be able to evaluate the performance and residual noise.

@Dannyf:
I have a cheap 240VAC / 24V@5A smps which will be used to drive the class D amplifiers. The smps have typically quite high leakage current, so using a trasfomer driven by the Class D amplifier would reduce the leakage current to minimum. Also, using transformers will provide galvanic isolation between the two outputs so that I can connect the in series or parallel without any issues. Yes, the noise is an unknown.

Probably I just need to build a simple prototype and see what kind of performance I can get and  what kind of noise problems I will encounter. If the design ends up to be a complete fiasco, I can build something like in Jim Williams's app note referenced above. Or, I may just use conventional approach and build the power supply using transformers with suitable secondary taps. A compromise between a low noise linear supply and a smps pre-regulator would be a hybrid: When the low noise is required (like powering precision analog circuits) the pre-regulator will be disabled and bypassed. When higher current is required, the pre-regulator will be enabled.

[dannyf]

I have a cheap 240VAC / 24V@5A smps which will be used to drive the class D amplifiers.

Wouldn't it be much easier to simply use that smps as the pre-regulator? You can use it as is, or to tap into its feedback circuitry and make it a tracking pre-regulator.

Either way beats using that smps to power a class D amp to pre-regulate.

[Kalvin]

I have a cheap 240VAC / 24V@5A smps which will be used to drive the class D amplifiers.

Wouldn't it be much easier to simply use that smps as the pre-regulator? You can use it as is, or to tap into its feedback circuitry and make it a tracking pre-regulator.

Either way beats using that smps to power a class D amp to pre-regulate.

The smps has too high leakage current between the mains and the secondary side, and I would like to have floating outputs in order to get both positive and negative outputs, two discrete outputs, or possibly more output current if both channels connected in parallel. That is why I would like to have transformers in the power path to provide galvanic isolation between the channels and to reduce the leakage.

[Kalvin]

There are quite a lot of benchtop power supplies based on ATX-style power supplies. Maybe this design could also be used with the ATX-designs to get higher output voltage (using  transformers with 1:2 ratio) and/or positive+negative output voltage for analog circuits.

[dannyf]

That is why I would like to have transformers in the power path to provide galvanic isolation between the channels and to reduce the leakage.

The ATX styled power supply (or a laptop power supply) is isolated.

If you are looking for a one-off design, look into some of those temperature controlled solder iron powered by laptop power supplies. they are essentially adjusting the laptop output.

If you are looking for a generic design, use the atx or laptop power supply as a (fixed) dc source to power a dc/dc converter whose output tracks with the linear regulator's output.

[Kalvin]

That is why I would like to have transformers in the power path to provide galvanic isolation between the channels and to reduce the leakage.

The ATX styled power supply (or a laptop power supply) is isolated.

There is Y-capacitor between the primary and the secondary windings which will cause a leakage current, typically hundreds of microamperes unless the smps is earthed. Earthing means that the output is not floating any more.

Here are two discussions about the problem with the floating smps:
- https://www.eevblog.com/forum/testgear/smps-leakage-currents-present-in-the-output-pins/
- https://www.eevblog.com/forum/testgear/feeltech-fy3224s-24mhz-2-channel-dds-aw-function-signal-generator/msg906746/#msg906746

[Yansi]

And whats the problem with the Y cap? Just leave it there, it primarily works as a shunt for the parasitic capacitance of the switching transformer, to suppress noise - which would be very likely present on an floating PSU output.
What problem would it cause to you to leave it leaking those safe microamps?

//The other method is to use a shield between primary and secondary windings of the switching transformer//