SMPS bandwidth!

Status
Not open for further replies.
Hi y'all!
I have a question for all you maniacs out there 😀
Say that I want to design an SMPS capable of delivering 2kW into rather low impedance loads (less than 5 ohms). This can be done without great difficulties, using for example a full-bridge forward converter (note that electrical isolation is a must!). If I then tell you that I want to be able to modulate the output voltage (-200 to +200 volts) at a maximum frequency of at least 10kHz (preferrably 30-40kHz). Can this be done using the same topology???
Do you think that I will have to use a fixed secondary voltage from my forward converter and hook up another full-bridge and PWM to get there?

Is there som simple relation between switching frequency and bandwidth? If not simple, do anyone know where I could find the maths involved?

I hope some of you have an answer as the thought of having to build me a prototype or at least spend some days simulatin seems boring...😉

Yours sincerely, Henrik
 
Hi Henrik

Did I get it right that you intend to build a class-D amplifier where the rectified mains voltage is switched directly and transformed down afterwards ?

What you definitely can't do is transforming a PWM signal like it is output by a PWM amp, at least not by using the usual sized SMPS transformers, simply because they can't handle low audio frequencies (and those transformer that can handle audio won't most probably handle the switching signal).

One clever idea is the one patented by Peavey. They switch the primary of a transformer with a constant duty cycle. On the secondary they have active rectifiers (using MOSFETs) that are controlled by a phase modulator (wich by itself is controlled by the audio signal) and thereby generating a PWM signal that only has to be lowpass-filtered to get the audio signal again.

The other one I once saw in a book (don't ask me about it's name, I don't own it I just saw it at a bookseller's, and that one isn't just around the corner) and which was called "Ampliverter". It is a SMPS where the primary side is switched according to the ABSOLUTE VALUE of the audio signal (which would in a normal SMPS give a "DC" output that is following the signal envelope). The secondary is then rectified (using active rectifiers again) according to the POLARITY of the audio input signal. While this might achieve the highest efficiency possible (only ONE switching power conversion process from mains to audio output !!!) I assume that it might have quite a lot of crossover distortion.

The question about the switching frequency isn't answered very easily (and by no means in one formula) since a lot of tradeoffs have to be made.

Regards

Charles
 
If you want to use a full bridge setup, have a look at how this chip does it's stuff. http://www-s.ti.com/sc/psheets/slos369a/slos369a.pdf
Notice that for zero output, both ends of the load are driven up and down in-phase (i.e. common mode) so there is nothing supplied to the load, but on a positive going signal for example, the left side pulse width widens symmetrically either side of it's centre point and the right side pulse narrows the same way. A really great topology.

For a class D using a conventional filter you would probably want to switch at 5-10 time the maximum audio frequency you want to reproduce.

2 kW can be done without great difficulty? Only in the glossy brochures of the controller chip manufacturers. 🙄 Either that or you have a better magic wand than I, sir. 😉

GP.
 
Get a Sorensen DCS Power Supply

If you can, get a copy of the instruction manual for the DCS series of power supplies made by Sorensen and you'll see how complex it is to keep it stable and all the care which is necessary to keep EMI from radiating back into the power lines. This is a programmable series of supplies is capable of 3kw (I think Sorensen is still part of Raytheon). They use synchronous rectification.

For the switching audio amplifiers look at the ap notes at Texas Instruments and International Rectifier.<p>http://www.irf.com/product-info/audio/fsaudio.html<p>http://focus.ti.com/docs/prod/productfolder.jhtml?genericPartNumber=TPA005D14&familyId=20<p>
 
jackinnj wrote:

They use synchronous rectification.

That was exactly the expression I was looking for when I wrote about "active rectifiers" but it didn't come to my mind anymore. 😱

The Texas IC proposed by Circlotron could definitely be used as a controller for class-D amps in the kilowatt range (thogh more than 0.1 % THD isn't quite state of the art nowadays). But it would still NOT be possible to transform it's output signal down by the use of a small transformer.

Regards

Charles
 
Well well...

Wel Hi there! Gee, lots of answers. Goodie!

Let me specify:
I want to be able to generate a high voltage, high current, relatively low ripple symmetrical power supply signal. THEN I intend to feed this to my 500W class A amplifier.

Hopefully the result will be great sound but low power dissipation (always only about 5-10 volts over my output stage). I think Bob Carver has done something similar, but not with Class A and not with fully switched powersupply.

Because of rather high PSRR in my class A stage, I can tolerate far more ripple on the mains than if I where to drive a speaker directly.

The basic idea is to design a standard SMPS with a full-bridged forward converter (h-bridge on rectified mains, through switch transformer, lowpassfiltered on the secondary side) and apply careful feedback so that I can control the output voltage at 20kHz or more.
Is this possible (I know it is as someon has already done it so I have to say 'easily') ??

/ Henrik
 
You can use the excess heat

from a 500w class A amp to fry french fries!

I've considered using the On-Semi (www.onsemi.com) MC33363A "Off Line" switcher chip for a different project, but this isn't "low noise". I think that if it's really high quality, low noise you want in the power supply you should consider one of the low noise, push pull gate switchers from Linear Technology.

From a safety perspective, I am a little reluctant to publish switcher circuits on my web-site. The last thing you want is some guy trying to build a switcher on protoboard.
 
If you need 2kW a full-bridge topology is required. The forward converter topology is another topology which will probably require a bigger transformer because of how it will traverse the B-H curve of the transformer. The best thing to do is to use an off-line voltage doubled current-fed, current-mode controlled full-bridge topology with synchronous rectification. Using the current fed topology will simplify things by eliminating output filter inductors on the secondary side, however, achieving the synchronous rectification will require a whole other control circuit scheme on the secondary side of the transformer. The switching frequency of the main transformer will have to be higher than 10kHz otherwise the core area of the transformer will be large. You should choose about 50-100kHz to drive the size of the transformer down somewhat. As far as the bandwidth goes, the high speed outer voltage control loop has a much wider bandwidth compared to the slower inner current control loop. The best way to get symmetrical rail voltages, especially for a class-A amplifier, would be to use two power supplies each one regulated to itself and rectified for positive and negative voltages respectively. Otherwise, the voltage control will be from one voltage rail only or between the two rails and the ground could be anywhere. By doing two supplies it will eliminate the addition of a magnetic amplifier postregulator for the second supply rail. If isolation is important to you then don't forget to isolate using gate drive transformers for the the full-bridge and optoisolate the feedback loops. I am not sure what voltage input you will be using off-line wherer you are from but for the full-bridge, make sure that the transistors are rated for at least 2x the peak voltage plus 30%. There will be very large leakage spikes the must be snubbed. For low noise supplies, you should try a Phase Shifted Zero-Voltage Switching power supply, but this is a thousand times harder to achieve and is an extension of a full-bridge master. Good Luck!!!

BeanZ
 
Gee!

Thnx!
That did it I think. As this is grounds for my master thesis I dont think I have to worry about 'hard'. It is supposed to be hard :bawling:

Ill search some on the different topologies and report back with my experience.

Thanks for all the help guys!

/ Henrik
 
Try looking up Abraham Pressman's "Switching Power Supply Design", Keith Billings "Switch Mode Power Supply Design Handbook", Col. Wm McLyman's "Transformer and Inductor Design Handbook", Irving M. Gottlieb's "Power Supplies, Switching Regulators, Inverters, and Converters", and you will have a good start. For magnetics design check out the design literature from Magnetics Inc. Good Luck!!! Feel free to email if you get stuck somewhere.

BeanZ
 
Status
Not open for further replies.