A statement George loves to repeat without practical data to back it up with. The switching frequency is several 100sKHz at the very least and the output filter does not filter it out entirely. Regardless the filter does not have the effect at audio frequencies that George suggests. The effect of the filter is to remove the switching frequency, which none of them do entirely. What is left is called 'the residual' and will be a sine wave at the switching frequency. It can't be heard, and because its a sine wave does not interfere with radio and the like. But it is energy, and should be unable to damage the tweeter. So it should be at a low level. The inductance of the speaker itself is often enough that the residual is of no consequence unless the switching frequency is stupid low (and there are no class D amps in production like that). |
If you think about it Ralph, linear solid state amps that are well made
but with global feed back and a bit of it, also have the same lifeless
sound that many say Class-D has. No argument there, but the kind of feedback Bruno is talking about is a different matter. Nelson Pass wrote a great article about this- you can find it on his DIY site. In it he speculates about 60 db, but points out how impractical it is to add more gain/circuitry to get there! The thing is, in class D making gain isn't a problem. An essential bit of the class D circuit is something called a comparator, which is a lot like an opamp run open loop. If you know your opamp theory, that means you have nearly infinite gain. I'm nutshelling this a bit, but the point is you can make a lot of gain with class D without adding extra circuitry to get there. So 60 db of feedback is very doable. |
Large amounts of global feedback used in linear amps is used in poor
designs, to get good specs, the challenge is to get good specs with only
a little local feedback. George, I'm a bit uncomfortable with this statement- a lot depends on how its interpreted. I'm of the opinion that the amp should have good linearity open loop, then add the feedback once that is achieved. This would make the latter portion of the statement true. But if the amp is to have good specs with only a little feedback, then this statement is likely false. Unless its a very small amount of feedback (under about 4 db) the application of feedback (even in a good design) will be detrimental. This is due to bifurcation of the input signal by the feedback itself (Bruno says that the harmonics 'show up out of the blue' but its easier to understand if you apply Chaos Theory). So the solution for a **good design** is to use a lot of feedback, hopefully on a design that already has good linearity to begin with before the application of feedback. Sorry to go on- I just felt this needed clarification. |
They heard that right. But had they been in a position to add 60dB, well
then, suddenly they would have been confronted with a sound that is
little short of magical." Bruno is spot on with this. The usual rule of thumb is that very low amounts (under 4 db or so) is not harmful, but more than that is a problem; after about 20 db or so things start to settle down. 60db hasn't been practical until the introduction of class D (since gain is developed in an entirely different way); so I have no argument with this; my prior comments should be limited to traditional amps where making the kind of gain is impractical. |
Feedback, while suppressing distortion, adds some of its own.
In addition, it is a destabilizing factor in any amplifier design. 'Stable' means that it is resistant to oscillation.
The distortion added by feedback contributes to brightness (a coloration) and hardness (which is unpleasant); both due to the ear's quality of converting all forms of distortion into tonality.
So far we've been running our class D zero feedback as well. Output impedance does not suffer, so driving a load really isn't the problem. Controlling distortion is, but if you are careful with the design that can be controlled too.
|
Naturally your not going to tell us how it’s done, I’m hope for you that
you can do it, but also sceptical at the same time, as minds more
attuned to solid state Class-D topology than yours have not succeeded,
save for Technics who are half the way there. I'll be happy to after the patent is approved. That EE turned out useful after all... |
We've already been listening to the amp. FWIW the squarewave response you linked does look pretty bad. Our tube amps can do better than that!
Obviously we're not letting this thing out without it doing what its supposed to do.
The thing is, you can have all sorts of issues when building a switching amplifier. The layout of the boards is critical- if not done right you can get stability problems. The little 'dimples' in the squarewaves you linked say to me that there is a suppressed oscillation in the amp somewhere. The squarewave should have been flat across the top, but its not! That's not a switching frequency thing as it is a stability thing.
|
I completely agree that switching frequencies have to be increased!
But, the problem is not switching speed! The problem is how long it takes for the device to turn on and off, which is different. Because that takes some time, the circuit has to wait until the device has changed state before the other device can do its thing. That waiting time is deadtime, which increases distortion.
**That** is why you go with faster and more expensive devices; its all about keeping deadtime to a minimum.
Put another way, with conventional class D circuits, a switching device that can do 10MHz can only be used at a few hundred KHz before deadtime becomes the big impediment.
Now the parts we work with are not that expensive, but we found a way to eliminate deadtime. This allow us to switch at much higher frequencies.
|
"if you want to have your product performing at the cutting edge it is
not possible with today’s known switching technologies. In order to
come close to the performance of the best linear design we would need
high-current semiconductors that provide switching frequencies of
several MHz or even GHz."
Several MHz isn't a problem. We're working with inexpensive devices that can easily do 5MHz and beyond. The problem is deadtime. That has so far proven to be the switching limit; that's why they don't switch much higher than maybe 2-5MHz tops. The deadtime increases distortion, and the faster you switch, the more you need it- so there is a minimum distortion that you can hit, sort of like Whack-A-Mole. I did state this in my previous post. When you are switching that fast though, the speaker itself is part of the filter; there's no problem getting a filter to work and it won't burn up. |
@merrillaudio +1 In the future when technology allows the switching can then be much
higher to allow the output filter to do it's job properly and cut out
all the switching frequency noise without effecting the audio band. This statement is false. The filter can do its job properly at current frequencies. There is an advantage to going to faster switching frequencies- lower distortion. But by switching faster, you either have to have faster and more expensive output devices and deadtime to allow the outputs to switch. Deadtime increases distortion. So there's a bit of a carrot that is being chased. @merrillaudio would you be interested in a circuit that bypasses the need for deadtime? |
