I have started to look at switch mode power supplies that track the audio, a la Bob Carver amps, and later Lab.Gruppen and others.
The use is a home theatre system with tri- and bi-amped speakers, so there's a dozen or so power supplies to build, hence a simple IC based solution is attractive.
What is the current state of the art in circuits, ICs and power semiconductors for this?
The trend in industrial SMPS seems to be towards Zero Volt Switched circuits.
My first impression is that it seems attractive for Audio too, but I see there are some tricks and traps with this technique.
Standard MOSFET or bipolar power semiconductors still look cost effective for < 600 V but perhaps there are new SiC or fancy semiconductors that I should consider.
Any comments appreciated, except that I should just use Class D amps - different discussion!
David
The use is a home theatre system with tri- and bi-amped speakers, so there's a dozen or so power supplies to build, hence a simple IC based solution is attractive.
What is the current state of the art in circuits, ICs and power semiconductors for this?
The trend in industrial SMPS seems to be towards Zero Volt Switched circuits.
My first impression is that it seems attractive for Audio too, but I see there are some tricks and traps with this technique.
Standard MOSFET or bipolar power semiconductors still look cost effective for < 600 V but perhaps there are new SiC or fancy semiconductors that I should consider.
Any comments appreciated, except that I should just use Class D amps - different discussion!
David
Last edited:
The Labgruppen and the similar Yamaha EEEngine don’t require special power supplies per se. They have simple self-oscillating buck converters inserted into each rail of each channel. They don’t require a ridiculous number of parts or anything hard to find and are all discrete. The actual power supply can be just a big toroid - or switcher if you’re up to building it.
The early Carver circuits were variations of class G and H, which require multiple tapped power supplies. Easy to build and get working, not quite as efficient as the tracking supplies but still beat regular class B by a huge margin. You can easily build the euquivalent of the Carver “cube” amps with a regular power supply instead of that tricky thyristor controlled supply and it would be just as efficient (and not eff up as easily). Later Carver amps adopted topologies similar to the Labgruppen to get even more efficiency.
The early Carver circuits were variations of class G and H, which require multiple tapped power supplies. Easy to build and get working, not quite as efficient as the tracking supplies but still beat regular class B by a huge margin. You can easily build the euquivalent of the Carver “cube” amps with a regular power supply instead of that tricky thyristor controlled supply and it would be just as efficient (and not eff up as easily). Later Carver amps adopted topologies similar to the Labgruppen to get even more efficiency.
Not sure about your reference to ZVS and < 600V devices. That sounds like off-line LLC converter stuff.
You should just use Class D amps....
Tongue in Cheek. Synchronous Buck driving analogue bridge amplifiers would appear to be the way to go. Synchronous Buck is, in effect, half bridge Class D with the bottom Mosfet switching to circuit ground.
You can get control ICs that will drive external Mosfets.
Buck Controller External Switch | Products | Step-Down Buck | TI.com
Buck Controller External Switch | Products | Step-Down Buck | TI.com
Other manufacturers are available.
I am interested if there are improvements to semiconductors that would be useful for this application.
I know there has been a lot of work on SiC but practically all for the industrial power market >600V, so not suitable here AFAIK.
There is also GaN but it is not really mature, also AFAIK.
There have been improvements in Si tech, super junction MOSFETs and even power JFETs.
So I wondered if anyone familiar with the latest state of the art could recommend the best choices at this time, save a search of manufacture's brochures.
Best wishes
David
I am not fully informed about all the history, especially of the Yamaha and various versions of the Carver.
The Labgruppens appear not to have a conventional power supply, just a switcher.
But maybe different models used different tech so it would be informative to learn more.
Best wishes
David
Labgruppens do have switching supplies - but so do a lot of amps, including regular old class AB. The only thing unconventional about the supplies is that they are made to take a pounding and have PFC. So do high end gaming computer supplies. The price reflects it - the amps are very expensive. The tracking downconverters are part of the amplifier, and operate independently from the “supply”. You could build the same amp and run it off a big 60 Hz toroid. It would be big and heavy but roughly the same efficiency, other than not being PFC. The tracking downconverters or some variant thereof could also be grafted onto your favorite class AB amplifier design if you have something you really like and want to keep the same sound.
That was not what I understood, it seems less efficient and more complex, perhaps this biased me to misunderstand, or perhaps this varied from model to model.
What model(s) does your comment cover, do you have a link or reference?
Best wishes
David
OK, the schematic I found does show this, hmm.
Does anyone make an amp where the SMPS tracks the audio?
Seems simpler and potentially more efficient.
Best wishes
David
A Class D amp is powered by a DC supply, a SMPS is powered from the mains.
Or that's how I use the terms in this context.
Obviously very similar but the different uses do lead to some difference in approach.
Like Power Factor Correction and isolation for the SMPS.
Also SMPS are normally optimized to respond to variations in load current demand whereas Class D track a variable "input" (required output).
Yes, this is discussed in an informative thread in "Solid State" by Damir Verson (DadoD).
That thread is on hold while he is out of action so I came to this forum to learn from a different perspective.
I'm a little disappointed in the lack of interest but at least it's educational to have to work it out from scratch.
Best wishes
David
Last edited:
It's a really unusual thing to want to do.
I have no experience of this.
I can try to think of useful questions, such as what switching f are you thinking of and what supply impedance do the amps require?
Your question was already useful, made me think more about why commercial manufacturers use SMPS to drive class-D rather than a one step solution audio frequency capable SMPS (that could be called an off-line class-D).
It seems kind of clumsy to have a SMPS that is nearly a class-D amp and then follow it with a whole class-D amp.
However it does have some benefits, divide and conquer - first handle isolation, line variation, PFC, etc, then efficient amplification.
Similarly the Labgruppen approach, first a SMPS, then a switched buck tracker than an amp.
But it seems potentially more efficient to have a tracker SMPS, only one set of switcher losses.
And if we have a tracker SMPS then we can clean up with a nice linear class A or B+, no need for a class-D.
So I need to combine the SMPS solutions to isolation, PFC, efficiency, with class-D accuracy and speed.
Not as simple as you may expect, there's a tonne of patents to study from Carver, Yamaha, Labgruppen and I haven't even looked yet at QSC, Crown and similar pro amps.
It's not that I want to work it out from scratch, just that I don't even know if anyone has actually done it, and power supply experts like Eva haven't commented yet.
So at the moment it's more about circuit architecture than details.
Just for back-of-envelope "is this feasible?" questions I assume a Fs around 400kHz, at least 200 kHz but less than 1 MHz seems reasonable with the semiconductors I know about.
Best wishes
David
Last edited:
ST used to have some chips which implemented amps based on this architecture - a classAB amp fed from a tracking supply. It's called 'BASH' but all the devices are obsoleted now. Datasheets are still available, for example : https://www.st.com/content/ccc/resource/technical/document/datasheet/94/66/e8/69/52/36/45/88/CD00002942.pdf/files/CD00002942.pdf/jcr:content/translations/en.CD00002942.pdf
Thank you for the link.
I understand that the BASH concept simplifies the integration of the power supply and the linear amp, with the power supply mostly treated as a black box.
At the moment the internals of the power supply are precisely what I need to work out, but BASH is on my study list for further down the track.
Best wishes
David
How much efficiency are you aiming for? I don't think you can reduce the wasted heat at maximum amplifier power by using a SMPS so the speakers would have to be able to cope with this much heat anyhow.
For a linear amp operated at low bias, like 20mA, is there a significant benefit to modulating the supply voltage? I can see if you are using class A amplifiers that modulation is going to significantly reduce the average heat output.
My speakers are built from JBL theater components, not home theater, AES rated at 600 W continuous for each woofer, subs a little less.
So maximum amplifier power should never be a concern thermally, it's only there for headroom on transients.
My primary concern therefore is quiescent dissipation, a linear amp capable of 600 W requires a fair bit of bias to keep distortion low.
Ideally it would be around 900 mA, multiply by +- 75 volt supplies and that's a lot of wasted power.
But a certain amount of baseline load can help to keep SMPS parameters more stable so this current draw may be actually be useful.
The quest is to see how little compromise I can make with the linear amps and still keep acceptably low power at idle.
I have noticed that the class-D amps don't really try too hard on efficiency, it's already way better than linear so just make it cheap.
Whereas the SMPS culture is very efficiency focused, it would be nice to use a State-Of-The-Art efficient circuit.
Luckily these also tend to have reduced EMI so that would be a nice bonus.
Best wishes
David
Last edited:
If you had a tracker that kept 15 volts across the outputs you’d have 27 watts idle dissipation per channel. No big deal, especially considering that you’d need a heat sink capable of more to handle the average current demand times those 15 volt rails plus the 27 watts idle. That’s a lot less heat than a regular amp biased that high. The fact that you need some heat sinking anyway is good because it prevents the thermal issues you have with cheap class D. Class D has very good efficiency and low switching losses at low power, but dissipation goes up like a hockey stick at high power. With no heat sink they get blazing hot in a few seconds at high power peaks, and the thermal cycling will eventually kill them. The dissiaption in a TD amp goes up gracefully as power output goes up, and the converters themselves are immune to cross conduction issues which plague class D at high power. Keeping efficiency high in a class D is a balancing act - the more power required, the smaller it needs to be electrically to keep switching losses down. But the smaller you make it physically, the higher the conduction losses and the more current density you need to run thru PCB tracks. And the harder it is to get heat out. With a class AB amplifier, it can be electrically large and not hurt anything, other than it being more difficult to get 1 ppm distortion.
The more think about this the more I understand the reasons.
My proposal makes sense for a unitary amp
But for a set of amps it does makes sense to centralize the line filter, Power Factor Correction, Soft Start, deal with line droop etc.
Then each individual amp has a simple and fast buck controller.
I have had the idea to use the buck converter to deal with problems such as DC offset at the speaker, short circuits or critical over temperature - we just stop the power.
This saves the speaker and the amp if there is a failure, plus no speaker relay to fail or distort either.
Make sense to anybody?
David
Last edited:
I haven't done the simulations yet but that's in the ballpark, I expected 10 V would be adequate maybe even a little less, IIRC Labgruppen use 7.5 V ?
So yes, less than 20 W is attractive and I really don't need to pack it into a 1U rack like a pro-sound amp.
Is this the famous "hockey stick" curve?
Well I don't care how much science you have, I refuse to believe it
And despite the extra power the temperature has not been proved to increase.
And the increased temperature is actually an improvement.
Best wishes
David
I think I read that class D is about 90% efficient and I would guess a SMPS might be as good as 95%. Is this about right?
I would be concerned about the voltage slew rate of a SMPS. They are designed to maintain constant voltage after all and have sizeable output capacitors. Would it be necessary to run tweeters from non-modulating supplies?
I imagine there is a lot of work needed to evaluate and modify a stock SMPS to work in this mode. Are you convinced that this is more fruitful than redesigning the linear amps to use much less bias current?
I would be concerned about the voltage slew rate of a SMPS. They are designed to maintain constant voltage after all and have sizeable output capacitors. Would it be necessary to run tweeters from non-modulating supplies?
I imagine there is a lot of work needed to evaluate and modify a stock SMPS to work in this mode. Are you convinced that this is more fruitful than redesigning the linear amps to use much less bias current?