ULTRA-high frequency driver impedance vs. Class D amps?

You can easily measure how Class D amps have a wildly varying frequency response depending on the load impedance. This is because they inevitably have a built-in passive lowpass crossover to supposedly protect the speaker from the Class D switching frequency.

This makes me wonder, what happens if you just took out the filter?

And further made me wonder, what IS the speaker impedance at such high frequencies?

Does anyone have a setup to get valid/believable speaker impedance measurements at ultra high frequencies?
(Typically Class D amps switch at perhaps 300 kHz, though it can vary from 100 kHz to >1 MHz as far as I know).
 

CLS

Member
2005-06-17 6:58 am
Taiwan
In ultrasonic frequencies, mechanical loads of the VC could probably submerged in all the not-rigid-enough materials and not predictable.

Ignore those, just look at the inductance of the VC, the impedance can be calculated @ those frequencies, which are skyrocketing.
 
In ultrasonic frequencies, mechanical loads of the VC could probably submerged in all the not-rigid-enough materials and not predictable.

Ignore those, just look at the inductance of the VC, the impedance can be calculated @ those frequencies, which are skyrocketing.

Well, the coil does not have an "inductance" really, it has a rising impedance due mostly to eddy currents in the pole piece. More like a semi-inductance. But I don't know that anyone has checked those models at ultra-high frequencies.

That plus the "not predictable" is why I'm wondering if anyone has a rig they could trust to actually MEASURE the impedance up at say 300 kHz...
 
Measuring the impedance to a useful degree wouldn't require any more than a good non-inductive resistor and a scope. You could even use a signal diode to rectify the resistor drop into a small cap and find the current with a DVM across it. You don't need an amplifier to run tests, just a simple high power square wave oscillator. Watching supply current would tell you practically all you'd need to know about load impedance. Putting your hand on the magnet structures might be interesting. I'd start with a low supply voltage on the oscillator. ;) The loss will vary greatly depending on driver design. Some tweeters will probably bow out in a hurry. Pole caps and shorting rings will drop the iron heating but drive more current into remaining impedance, making various conductors hotter. If the inductance changes a lot with excursion, which is usually the case, you'll get all sorts of strange effects. Even if you could design a driver to be highly compatible with the switching signal you'll run into serious lead wire problems were the only good plan would be almost no length. Twisted pair would be very good especially for AD bridge amps. Ultimately though, best performance is likely to be had by putting a low loss high impedance in series with the switching waveform and then a low impedance to trap what gets through, your basic two pole filter. That way it wont matter what anything after the output terminals will do with the HF stuff.
 
Last edited:

CLS

Member
2005-06-17 6:58 am
Taiwan
Low power, yes I guess so.

The HF 'garbage' in the output of class D amp must go somewhere. Feeding it directly into the speaker seems not a good idea - it would transform into heat anyway. VC is already very 'busy'. Heat is no good.

So let the garbage (heat) dissipate in places other than VC is a good practice.
 
Of course the coil has inductance, lots of it. The eddy currents will be responsible for a falling impedance, and that's pretty much what will make the pole piece real hot if you try to drive a huge 300kHz square wave into the coil.

Well, I meant that more for those with less understanding than you obviously have, who think that because there is a coil of wire in there the whole thing acts like an inductor.

Don't you mean a RISING impedance?

Your suggestion of using a scope is good...but how do I know if a "non-inductive" resistor is still just resistive at 300 kHz? Actually even at the 100 kHz I use them for...
Mmmm, I guess I could look at the voltage vs. current phase...but don't you need a perfect resistor for that also... :(:confused:

As for the heat, a recent AES paper contended that with the speaker impedance real high, the additional heat contribution would not be much. I was looking to try and check that for myself.

The reason for all this would be to avoid the varying high frequency response and huge peaking, which don't seem to be a good thing. Whether directly audible or not, I get kinda bugged seeing a 17 dB peak at 30 kHz, and wonder what audible effects it can be causing (like clipping the amp, or intermodulation, or ???)
 
Last edited:
I guess in some vague way it's a bit like CD filters...they brickwall above 20 kHz to cut out folddown artifacts (there's a technical term escaping me-aliasing?)

But if you oversampled to 88k or 96k or whatever, couldn't you just dispense with that and say "eh, it won't be that big a deal, and I can get rid of all the brick wall problems"

That's where I was going with this-the physical class D filters cause these big response problems, so what if we ditch them? Will the coil really get heated that much?

On a number of class D amps, the residual read like tenths of millivolts without a passive filter. Doesn't seem like that would harm more speakers.