Lab supply with switching pre-regulator

[tbj]

Hi,

I am an amateur electronics enthusiast currently involved in designing and building homebrew class D amplifiers. What this means is that my 'old faithful' 0-30v 3A lab supply has outgrown its usefulness and I need to build a dual rail supply for my experiments with half bridges.

I have a nice toroidal transformer that has 2x35 volt secondary windings - I can't remember the VA rating offhand but I think it's good for at least 6 amps per winding. With this I can make a dual rail 0-45v, 6 amp PSU. My first instinct was to build a simple linear supply, but then I realised that at a theoretical maximum output of 270 watts per channel, with two channels, the amount of heatsinking required would be ungodly.

It may therefore be more sensible to make a switching supply, something that I have never done before (of course I've played around with boost converters, but I've never made a high power switcher before).

It then occured to me that for a nice clean output I could implement the switcher as a 'tracking pre-regulator' and still have a linear regulator as the output element. The switcher could follow the set voltage plus (say) 2 volts, which would mean that the maximum dissipation across the pass element would be 2 * 6 = 12 watts.

Would this seem to be a more sensible approach? Has anyone got any schematics I could look at as an example of how this is usually done?

Or should I forget using an SMPS and just go linear instead?

[Bored@Work]

This has been done before. Even Dave's uSupply project (currently on hold) uses one.

However, your problem will be that SMPS' and audio don't work too well. You should be prepared to spend some time adding filtering and shielding.

[riktw]

The linear regulator will filter some noise away but especially high frequency noise will still appear on the output.
You still need to design a proper SMPS, good quality low ESR caps and a well designed PCB are still important.

[edavid]

There's no real reason to run an audio amplifier from a regulated supply.  Just build an unregulated supply and put a variac in front of it for experiments.

[Jebnor]

You might find this useful, "Switching Regulators for poets" by the infamous Jim Williams http://cds.linear.com/docs/en/application-note/an25fa.pdf

[Andreas]

The switcher could follow the set voltage plus (say) 2 volts, which would mean that the maximum dissipation across the pass element would be 2 * 6 = 12 watts.

Or should I forget using an SMPS and just go linear instead?

That´s a naive calculation for a real lab supply which has to survive any kind of loads.

Let's assume you have a load which switches between 10 Ohms and 0.1 Ohms
at a high switching frequency (above 100 Hz) and a setpoint voltage of 45 V.

During the 10 Ohms phase all is ok. The switcher regulates
its output capacitor to 47V giving a dissipation of 2V*4.5A = 9W.

Then we go to the 0.1 Ohms phase so current limiting to 6A of the linear supply will be done.
The capacitor is still charged to 47V. Output voltage is 0.6V due to current limiting.
Resulting power over the pass element is around 46V * 6A = 276W at least in the first moment.

So you have at least to calculate with the energy stored in the output capacitor of pre regulator times
the maximum possible switching frequency of your load as power dissipation of the pass element.

And: you have to consider the behaviour of the lab supply when switching back from 0.1 Ohms to 10 Ohms
again. In this case depending on strategy of the pre regulator it will take long time until the output reaches 45V again.

With best regards

Andreas

[kizzap]

What you are saying Andreas, is that the output of the Power Supply when dealing with switching loads, will have a response dictated by the size of the output filter capacitor connected to the input of the linear regulator?

If so, would it not be possible to lower the value of the filter capacitor to remove a lot of the stored up charge, thus improving the response time to the output load requirement? As far as I know, the output capacitor is used to simply filter out any ripple on the output, which in this case ideally would be filtered out by the linear regulator.

My primary question would then be, how would we calculate the thermal properties across the die on the linear regulator, given the higher loading, and short time constraint? I'm guessing that you would first calculate the time to drain the capacitor, but that is where I get lost. Surely the size of a heat sink and it's thermal mass would have a greater effect on the temperature of the pass element, then the conventional ^oC/W of the heat sink would given that the load isn't "continuous". I am of course referring to a single cycle of the load cycling, not a long term loading effect. I'd assume over the longer time frame the load dumped into the heatsink would be dictated by the percentage of time with the high load to the low load.

-kizzap