Interesting. I've only just become aware of the Soulution brand, through WBF, and had a quick look across at their website. The TAS review was very enlightening, and apart from the typical silly twaddle you read in audio magazines about how feedback works, and is a problem, it indicated that the Soulution 700/710 amps were built with the right technical attitude, and the designers have done their homework properly. They impressed the reviewer with the "cleanness", or lack of audible distortion, of the performance, so they seem to me to be an excellent reference point to aim at for myself in the design of the beast here ...
Another quick update. I'm sorry, Jack, this is slow(!) , no pretty pictures yet, just concepts swirling around in my head, and being played around with, inside a circuit simulator ...
Still haven't decided whether to go class AB or D: advantages and disadvantages with both, might end up doing totally realistic, virtual models of both, with both ending up having similar paper performance. If that's the case I could build both and see how they compare in the real world.
In the AB solution the big trick will be to get a highly linearised MOSFET output stage, I'll be playing around with Hawksford's feed-forward technique to improve the behaviour here, and to get around the fact that p-channel devices aren't really available to do the job will try making the circlotron topology work. This has yet to be to properly looked at ...
Mainly considering class D at the moment. Off the shelf units like Hypex are too expensive, and still don't perform well enough, but luckily the information is out there to allow you to do your own versions of same. The intrinsic distortion is still a little high but using active filtering techniques means it's possible to get the figures down to what I'm after. What's bugging me at the moment is that the UcD configuration delivers a lot of switching noise to the outputs, far too much for my liking, which means that muck is travelling down the speaker cables and then relies on the speakers to effectively filter it out: not my cup of tea. So right now I'm trying to see if I can get rid of that high frequency noise where it should be eliminated: in the amplifier itself.
The big advantage of class D, again, is that the RMS(!!!) power can be almost twice that of AB and the total amount of metal used in the making will be much less. A greener machine, in other words ...
Are you talking about doing the actual class-D amp design, or implementing an off-the-shelf module in your own box (e.g. your own power supply)? Or following the app notes from e.g. IR to roll your own based on a proven set of plans?
The last option, Don, the UcD concept and circuit is available from Philips, and someone has already done a Spice simulation with the key elements all working. This shows where the "problems" are, and in fact have been measured by JA on a real production item to match almost perfectly. So I'll be taking this as a starting point and seeing where I can go with it.
Agreed, in Spice you can see very nasty ultra high frequency spikes everywhere, reducing component to component, and trace inductances to absolute minimums is crucial for success ...
Seeing this is my thread, I'll say something nice and controversial now, as least as far as audiophiles are concerned. There is some talk going on now about valves, tubes in other threads, about the impact of rolling them, etc ...
As far as I'm concerned, as soon as you have a piece of equipment whose sound changes when you alter the active components, typically tubes, then you have a tone control mechanism as Tom would say. Correctly engineered, inaudible distortion equipment means that the active components can be altered and there is no change in sound characteristics. If you intend to have a system whose audible distortion characteristics alter every time you change a tube that's fine, but don't claim that it is a system without audible distortion ...
And so here's the point about my focus on a component with high levels of negative feedback. Properly implemented, this reduces the tonal characteristics if any still exist to the qualities of the actual components in the feedback path, typically a single resistor. The active components, bipolar, jfet, mosfet, tube become completely irrelevant, they are merely servants to the master of the circuit, the passive components of the feedback path.
And this is the intent of the amp here: it will have the sound qualities of a high quality resistor, for better or worse ...
Correct. Except that I wouldn't use the expression "behave" to describe what's being done. Active devices such as transistors, fets, tubes all intrinsically distort to a very high high degree as part of their essential characteristics, unless you examine only a very small area of their operating region. So the only way to correct that is to apply feedback in one form or another within the circuit makeup. All opamps use immensely high levels of feedback as part of their very nature of operation, they do not function unless used that way. Every digital recording has passed through circuits using extreme levels of feedback by the time it reaches your speakers, and essentially that is the case for the vast majority of analogue recordings.
So there is nothing intrinsically wrong with feedback, it just has to be used correctly. Which is where they went wrong to some degree in the 70's, and the audio world has been living in the shadow of that ever since ...
Pictures? Pictures of plans? Pictures of parts? Pictures of Frank drawing plans or shopping for parts? Any kind of verification that this isn't just another audio fantasy?
I was going through some of my equipment photos today from another site and came across this. It's the real deal at least 2.4kw of weldable power, enjoy.
Pictures? Pictures of plans? Pictures of parts? Pictures of Frank drawing plans or shopping for parts? Any kind of verification that this isn't just another audio fantasy?
Patience, please, Tim. I'm repeating myself, but but the concept has to be right for any chance of something real happening. Now's probably a good time to mention, as I have just elsewhere, that I'm focusing strongly right now on using class D topology, which has all sorts of advantages. This is using a proven concept by Philips (that name again!), which I could assemble tomorrow using standard parts. However, this won't provide sufficient power, only around the 200-400W mark.
To get the power I'm looking at, which if I go this way will allow for 1200W continuous (RMS!!) I need to scale up to handle higher voltages, and at the moment this causes instability problems in the simulation alone. So I have to understand exactly how to deal with this, not a trivial exercise. But fear not, I'm still thrashing with it ...
All I can offer as "proof" at the moment is a schematic, and I'm not too keen to do that!
Down the track, if class D works out, or DOESN'T work out, then I'll go with more standard AB architecture as another unit, to compare with the class D version's performance if it can fly, at the very least.
Thanks for that input, Roger, but if one wants real "proof" that you can do really high power with class D there are a number of units out there, capable of 10's of thousands of watts. The trick is to do it so it sounds "nice" ...
A very, very quick update: things are still on track, no pictures yet, sorry! The headache has been coming to grips with class D, I had instability at high power in the simulation, and it was a learning exercise to understand what was going on. Turns out it was a nasty parasitic behaviour, the sort of thing real devices have which makes them not perfect, but which you have to take into account to have any chance of creating a real device. I knew the issue was along these lines, the trick was finding where it was! In essence, the equivalent of class AB clipping, but it manifests in a different form ...
Things are looking better now, getting a handle on ensuring low DC output levels, and able to pump out 2400W into 6 ohms at 20,000Hz, very clean waveform with negligible switching noise, only a 1mV or so. This is still fairyland stuff because real devices in the output stage to match what the model assumes have be designed in, and perfect power supplies are being used up to now. Also, the true influence of parasitics everywhere in the connecting wires and traces has to be understood. Interesting times ahead ...
Would you get on a plane, of a new design, that was already starting to be assembled before the engineers had worked out what all the parts needed were, and what quality they needed to be?
There's a difference between an engineer and a hobbyist: the latter gets various bits and pieces and throws them together, and keeps fiddling with them using duct tape, rubber bands and spare pieces of fencing wire until it sort of works. The engineer, on the other hand, does it correctly from the starting point.
I was just trying to figure out where you were in the process. You are the master fiddler though aren't you? Going outside and flipping breakers back and forth and hoping to get a few minutes of music on song and all that.
I was just trying to figure out where you were in the process. You are the master fiddler though aren't you? Going outside and flipping breakers back and forth and hoping to get a few minutes of music on song and all that.
Yes, those were the bad old days! I gave up on good audio altogether back then for many, many years because it all became too frustrating, knowing you could get really top notch sound if absolutely everything was in alignment, but it was so hard to make that happen on a continuous basis. Much, much more knowledge now, makes all the difference ...
, no pretty pictures yet, just concepts swirling around in my head, and being played around with, inside a circuit simulator ...
