Choice of converter topology (one for the power engineers)

[tomizett]

Hi All... This is just a question to satisfy my curiosity.

I've recently had cause to take apart a couple of high-power audio amplifiers (same manufacturer, two different models) and I was interested to note that both used a flyback converter topology for their power supply. Now I generally think of flyback converters being used for supplies into the high-tens of watts, but these are 8-10kW. Most other comparable products I've worked on use full-bridge or half-bridge (presumably resonant) forward converters. So why are these different?

I'm wondering: How common are flyback converters of this size?

And the bigger question: How do you choose converter topologies when designing PSUs? What are the relative merits of different designs?
The advantages I can think of for the flyback approach would be the elimination of a high-side switch and only incurring one set of Rds(on) losses. But there must be downsides too (higher peak current I suppose)?

Be interested to hear your thoughts!

 

[AnalogKid]

It could be a single-switch forward converter. Do you have a schematic or are you guessing from looking at the board.

Also, what is the make/model of the 10 kW *audio* amp?

ak

 

[schmitt trigger]

I too wonder about 10 Kw coming from a flyback.

The flyback is a wonderfully simple, general purpose topology. But one of its key drawbacks is the excessive peak currents on its switching components (transistor and diode).

I'm not saying it cannot be done, only that there are better topologies for that power level. A resonant full bridge would be my choice.

 

[spec]

Hi All... This is just a question to satisfy my curiosity.

I've recently had cause to take apart a couple of high-power audio amplifiers (same manufacturer, two different models) and I was interested to note that both used a flyback converter topology for their power supply. Now I generally think of flyback converters being used for supplies into the high-tens of watts, but these are 8-10kW. Most other comparable products I've worked on use full-bridge or half-bridge (presumably resonant) forward converters. So why are these different?

I'm wondering: How common are flyback converters of this size?

And the bigger question: How do you choose converter topologies when designing PSUs? What are the relative merits of different designs?
The advantages I can think of for the flyback approach would be the elimination of a high-side switch and only incurring one set of Rds(on) losses. But there must be downsides too (higher peak current I suppose)?

Be interested to hear your thoughts!

Hi tomizett,

Here is a link to a pdf document that gives a description of the various switch mode power supplies, including their advantages and disadvantages. I too am doubtful that a flyback converter could be a practical proposition for the kind of powers you describe- far more likely to be a feedforward type: **broken link removed**

spec

 

[AnalogKid]

Because a flyback transformer isn't a transformer (it is a coupled inductor), it would be very inefficient at these power levels because of the enormous peak currents mentioned in post #3. Size, weight and cost would be excessive relative to other topologies, much more than the differences in complexity.

ak

 

[MrAl]

Hi All... This is just a question to satisfy my curiosity.

I've recently had cause to take apart a couple of high-power audio amplifiers (same manufacturer, two different models) and I was interested to note that both used a flyback converter topology for their power supply. Now I generally think of flyback converters being used for supplies into the high-tens of watts, but these are 8-10kW. Most other comparable products I've worked on use full-bridge or half-bridge (presumably resonant) forward converters. So why are these different?

I'm wondering: How common are flyback converters of this size?

And the bigger question: How do you choose converter topologies when designing PSUs? What are the relative merits of different designs?
The advantages I can think of for the flyback approach would be the elimination of a high-side switch and only incurring one set of Rds(on) losses. But there must be downsides too (higher peak current I suppose)?

Be interested to hear your thoughts!

Hi,

As others, i also wonder about how they got away with this, if that's what it really is though.
One guess would be that it was possible due to relatively recent advances in most converters because of advances in parts over the years. Back in the 80's i dont think you could find a true zero recovery diode for example, and power MOSFETs (like the HexFET) were just coming into their own. Thus, efficiency went up and other ways to deal with input power problems became known. Ever see a CUK pre 1980? I dont think so, as using a capacitor in a DC converter in series with the main power flow would probably gain you the nickname of "Kook" as caps were not as good as they are today.

For the flyback advantages, in two words: cost and simplicity rule, and remember the all important isolation which is a requirement for many applications in power conversion. Ok that's three words: cost, simplicity, and isolation.

 

[tomizett]

Thanks for the input folks! I'm very interested in power converters, but certainly no expert in them.

It could be a single-switch forward converter.

That's interesting, never heard of such a thing. I presume that means roughly the same kind of circuit but the majority of the energy is transfered during the "on" period... will have to look it up.

Do you have a schematic or are you guessing from looking at the board.

Well... a bit of both. The amps are both made by Lab.Gruppen (Sweden, I think) and are a PLM10000 and a LA48a. Now, Lab don't release schematics for these (and I don't think I'd be allowed to share them if they did), but I have found a service manual for the FP3400 floating round the web - this is part of the same product series as the LA48a (which was re-badged for another manufacturer).
I've attached the relevant pages from that - you'll see that it's "described" as a flyback converter, but that could be mis-translation... or mis-information?
Looking at the boards (without doing an exhaustive trace) they look very much like what's pictured.

As to the power ratings , I'm always a little skeptical about these - in general it tends to be thermal capacity that limits the output of this kind of equipment to a <100% duty cycle over timescales in the minutes. Nevertheless, they're certainly capable of pulling quite a current.

Spec, thanks for the link. Can't prommise I'll read all 153 pages tonight, but I'll certainly have a look!

In general, people are coming back with reactions similar to my own - namely "that seems unlikely". As I say though, it's a topic I'm tying to learn about so I'm keen to discuss it.
If I can get some photos of the transfomers I'll post them for your perusal.

 

[schmitt trigger]

Well...................by the relationship of the primary vs secondary winding phase, and the lack of an external inductor.............it does appear to be a flyback.
It has to have three hefty IGBTs to perform the switching though.

 

[spec]

Well, it looks like a flyback switch mode power supply. Three IGBT in parallel certainly would put out some juice.

(our posts crossed st)

 

[tomizett]

I've had a look through that Phillips ap note - as you said, it's a nice succinct review of topologies, and it looks like there's a wealth of practical detail on the latter pages.

I've learnt two things from it so far... Firstly that I've been mis-using the term "forward converter", so apologies for any confusion I've caused there.
Secondly, I encountered one of these the other day. It had me foxed, but now I understand!

On the original topic I should have added that another give-away is that the transfomer in the LA48a is massively gapped. I think it's what you'd describe as a C-I core, with two bobbins. Both joins in the core sections are spaced to gaps that must be about 2mm. As I say, I'll get a photograph if I can.

 

[spec]

I've had a look through that Phillips ap note - as you said, it's a nice succinct review of topologies, and it looks like there's a wealth of practical detail on the latter pages.

I've learnt two things from it so far... Firstly that I've been mis-using the term "forward converter", so apologies for any confusion I've caused there.
Secondly, I encountered one of these the other day. It had me foxed, but now I understand!

Glad you found the ap note useful- it is only one chapter of many dealing with various semiconductor application areas.

On the original topic I should have added that another give-away is that the transfomer in the LA48a is massively gapped. I think it's what you'd describe as a C-I core, with two bobbins. Both joins in the core sections are spaced to gaps that must be about 2mm. As I say, I'll get a photograph if I can.

Putting a big gap in a core is an old dodge to stop the core from saturating and also allowing a higher switching frequency. Looks like a blood and thunder design, but obviously works. Have you any idea of the switching frequency- pretty high I would think to minimize the size of the transformer and keep the switching peak current within bounds.

 

[MrAl]

Hello again,

There's very little doubt ... that is a flyback ...unless they drew the circuit wrong.

 

[strantor]

Where is a 10kW audio amp used? Is that used for concerts? I can't imagine anyone having a 50A breaker in their house dedicated to their stereo.

 

[MrAl]

Where is a 10kW audio amp used? Is that used for concerts? I can't imagine anyone having a 50A breaker in their house dedicated to their stereo.

Hi,

Ha ha, yeah, rock concert

 

[tomizett]

Indeed, they're professional sound re-enforcement amps. That power is split across four channels in the PLM range of products - they also do a PLM20000, giving a claimed 5kW per channel.

Have you any idea of the switching frequency- pretty high I would think to minimize the size of the transformer and keep the switching peak current within bounds.

No idea at the moment, but if I get time later I'll see if I can get an idea... reckon if I wave a scope probe near it I should pick up enough stray field to get a measurement.

 

[tomizett]

For the curious, here are some pics of the power transfomers concerned.

I fired up the PLM amplifier this afternoon (that's the one with the RM type core, the first picture, and is the newer model), and saw thin pulses with a repeat rate of only 22kHz. I can only imagine that it increases the frequency under load, or preforms some kind of pulse skipping.

One other observation is that they avoid the use of high-side drivers like the IRF2010 and friends anywhere in this equipment (the power amplifier stage has a buck pre-regulator), instead prefering to use high-speed optocouplers. Perhaps wanting to avoid the high-side switch was the rationale behind this unusual design... but then why not use a floating supply and an opto?

 

[spec]

For the curious, here are some pics of the power transfomers concerned.

I fired up the PLM amplifier this afternoon (that's the one with the RM type core, the first picture, and is the newer model), and saw thin pulses with a repeat rate of only 22kHz. I can only imagine that it increases the frequency under load, or preforms some kind of pulse skipping.

One other observation is that they avoid the use of high-side drivers like the IRF2010 and friends anywhere in this equipment (the power amplifier stage has a buck pre-regulator), instead prefering to use high-speed optocouplers. Perhaps wanting to avoid the high-side switch was the rationale behind this unusual design... but then why not use a floating supply and an opto?

View attachment 97936 View attachment 97937 View attachment 97938

Hmm. the transformer is smaller than I expected and the switching frequency is lower. Presumably the amp is class AB so when it is not driving speakers at a high level it will take very little current from the supply lines. Be interesting to see what happens to the PSU signals at maximum PSU current drain.

spec

 

[tomizett]

That figure of 22kHz is suspiciously low... it was just a very quick look at the end of the day (didn't have a suitable HV probe to hand so just picked up some capacative coupling from the IGBT heatsink onto the probe).
The amplifiers are AB with a swiching (buck) pre-regulator for efficiency (I'd describe it as "switching class G"). From memory they draw about 140W quiescent, and get surprisingly warm on it!
If I get a chance I'd be quite interested in seeing what it does on load - I'll let you know my findings if I do.

 

[MrAl]

Hi,

20kHz was the norm several years back when all designs used bipolars. After MOSFETs came into the picture several of the controller chips did not really increase the allowable switching speed for some reason. For example, many PC power supplies still use 10kHz to 20kHz, probably because most of them still use bipolars and as you know the storage time limits the saturated switching time significantly. It could be that EMI constraints got more strict so a design that switches as 100kHz might have more trouble passing the test. There's been so many problems in that area that some designs even require a slowdown in the rise and fall times just to allow it to conform to EMI specs, and so some chips are made with that in mind, with slower rise and fall times. Such a shame really, MOSFETs can do 100kHz, but radio communication is so important too.
The core size area of the transformer would be proportional to the frequency and volume is A*D where D is the last dimension, so a transformer that runs at 100Hz and has area 100 square inches and 10 inches high could end up with area 0.5 square inches and 10 inches high at 20kHz. Quite a size reduction. We also see this in the relatively new DC 'regulated' wall warts that are made with a drop in single switch IC that drives a small transformer that weighs almost nothing.

Back in the 80's we used opto's for BOTH high side AND low side switching for a number of reasons. One, using the opto's for both means closer matching switching times. Two, using opto's meant a standardized switching circuit and therefore standardized PC board to do that switching. Three, to reduce interference along the line from the controller board to the individual 100 amp transistors. These were also paired with a multi winding transformer to provide power to each individual transistor drive circuit: four for a single phase design, six for a three phase design.
The opto's we used were made by HP and were fast.