Is the UcD modulation scheme less than optimum?

[Lars Clausen]

Yes i did waste a lot of time discussing up against the commercial ad machine of a certain Class D factory back then.

But i'm not going to repeat that mistake

That's why

[classd4sure]

Good call.

[Lars Clausen]

Some nice people have encouraged me to repost my points,
so i will try to rephrase:

Originally posted by Bruno Putzeys:

If the output filter is in the loop, the load impedance also becomes a factor in this. At a certain load impedance (guess which value), the UcD modulator is astoundingly linear up to nearly full signal swing. At higher and lower impedances, odd order effects show up, with positive sign for lower impedances and negative sign for higher impedances. You can't do much more optimum than to minimize the effect near rated impedance.

Every amplifier that puts the output filter in the loop has this problem, quite regardless of whether they use a fixed carrier or not. Trying to find a simple trick to alleviate the effects of "ripple aliasing" (for lack of a better term) in amplifiers employing feedback around the output filter is a local hobby in Rotselaar, as much as it is in Copenhagen.

I agree with Bruno on these points of course.

The question is: why would you want to have the output filter inside the control loop? Thus adding a complex uncertainty (the load).

[classd4sure]

Hi,

But they didn't encourage you publicly?

I guess one answer would be because it at least let's you linearize the response for a given load, which would have to be better than ignoring it altogether?

Regards,
Chris

[Lars Clausen]

Hi Chris

I can agree with that.

But everytime a solution to something is introduced, it should be evaluated if the solution causes a cost on something else. And if so, is the cost higher than the benefit? (Normal cost/benefit analysis).

Otherwise it's like driving to your neighbor city to buy cheap gas for your car. It is undispuatbly true that you save 50c per gallon, and a total of 8 $. But you overlooked that you spent 10 $ driving there to get it. Mostly the latter part is overlooked by women though

In this case it can be determined that almost all loudspeakers are self linearizing because of the voice coil inductance is higher than the choke of the Class D amplifier. Look here for an example:

http://www.tymphany.com/datasheet/printview.php?id=58

This tweeter from Scanspeak has 80 uH voice coil inductance, much higher than any modern Class D amplifier filter. And so the resonse is ruler linear, even if you don't use post filter compensation.

So you are solving a problem, that is in almost any case not important, but at a cost. The cost is a complex load component in the control loop. The question now is: is it an overall benefit?

I am not saying one or the other type is b**ls**t.
In some few cases the compensation of post filter feedback can be a benefit.

What i am proposing is why not make pre or post filtering switchable, so the user can switch back and forth as he pleases,
to find the sound of preference? It's definitely possible on almost any Class D system.

[sovadk]

Nice simulation there analogspiceman. Is the AIM attachment from post #72 up to date (it's called aim1.zip)?

Regards
Kaspar

[analogspiceman]

Originally posted by sovadk
Nice simulation there analogspiceman. Is the AIM attachment from post #72 up to date (it's called aim1.zip)?

Yes, but be sure to read and make the changes mentioned in post #92.

By the way, here is the LTspice shematic that goes along with the frequency stabilized hysteresis plot from post #99.

Regards -- analogspiceman

[SSassen]

Lars?

I'm sorry, but I simply don't subscribe to your claims at all. Every class-D amplifier, but for UcD, that I have measured exhibits a change in the frequency response due to a loudspeaker being attached. This is purely due to the reactive load that a typical loudspeaker represents which is not compensated for by the amplifier.

I'm not just talking about minor, tenths of a dB, variations, but pretty extreme variations up to +/6dB, and usually worse at higher frequencies. UcD is the only topology that does not exhibit this behaviour and as a result has the same tonal character and balance with every loudspeaker I attach it to.

To me this is as close to the proverbial 'piece of wire with gain' one can get at the moment. As effectively the load invariant character of the UcD concept takes the amplifier out of the equation. There's no, or by comparision several magnitudes less, interaction between the (complex and often highly reactive) load the loudspeaker represents and the amplifier.

If you think, and your post expresses this, that a load variant character is an asset for an amplifier to have I suggest you rethink what we're trying to achieve here. Ideally we'd like every component in the chain from the recording to the soundwaves that finally reach the ear to have as little influence on the reproduction. Only then we can properly reproduce the recording how the artist orginally intended it. This obviously includes anything from error correction in the CD-player to room-gain.

An amplifier however is a big factor in this equation as it provides the highest amount of gain and thus its contribution to the sum of the parts is not insignificant. Therefore an amplifier should have very low THD and a flat FR curve (vitually) independent of frequency and load. Currently the only class-D topology that comes close to meeting this goal is UcD. And you still think pre-filter feedback is the way to go?

The moment you show me a class-D amplifier that uses pre-filter feedback that's able to match or surpass UcD's load invariant behaviour I'm willing to explore this further, up untill now however all attempts to design one (which I know of) have failed.

Best regards,

Sander Sassen
http://www.hardwareanalysis.com

[Lars Clausen]

Of course i don't think that:

a load variant character is an asset for an amplifier to have.

But you might say that i think the load variant character is of minor importance, and too expensive (in terms of sound quality) to get rid of.

After all you are placing the...

reactive load that a typical loudspeaker represents

... in the middle of the amplifier, instead (as i propose) on the output. The reactive load is in other words a (rather unpredictable) component in the amplifier's control loop. I don't think this can contribute to anything good.
But of course this is only my personal opinion.

Your website suggests you have someting to do with testing, measurement etc. Let me suggest you a test, ok?

Take your PC soundcard, and add a couple of resistors in the input to dampen the signal. Then connect the input of your soundcard to the output of your UcD modules, and
the output of the soundcard to the input of the UcD modules.
(or any other Class D module for that matter).
Make a frequency / THD sweep with loudspeakers connected as a load. (Not a resistor). Just sweep at say 1-2W or so.
(Can download free software here: www.ymec.com )

Then i think you will see, that post filter feedback will in fact not solve your reactive load problems, like you think.
But don't take my word for it, try it out yourself

Every class-D amplifier, but for UcD, that I have measured exhibits a change in the frequency response due to a loudspeaker being attached.

May i ask which Class D amplifiers you are referring to, i want to replicate your findings.

All the best

Lars

[SSassen]

Hi Lars,

I don't just think this is the case, I know so, as this is what I've been investigating over the past two months or so. I have already posted such measurements in the past, hence the reason I'm replying, here's a few examples of what I'm refering to.

hyst. osc. 8-ohm loudspeaker



sodfa 8-ohm loudspeaker



UcD 8-ohm loudspeaker



I think this properly illustrates what I'm trying to get across. As you rightfully pointed out these measurements can be duplicated by anyone with the right software (for example the freeware RightMark Audio Analyzer) and a PC equipped with a decent soundcard.

As you can see both the hyst. self. osc. and the sodfa topology that use pre-filter feedback see a rise in their FR at higher frequencies due to the reactive character of the attached loudspeaker. It is a simple two-way configuration, with no impedance correction network on the woofer or tweeter and hence has a rising impedance curve and due to the passive crossover a reactive component as well which is not compensated for.

Best regards,

Sander Sassen
http://www.hardwareanalysis.com