UCD180 questions

[Jaka Racman]

Hi,

attached is a simulated response of UcD180 to 10kHz burst signal into proposed ESL load. UcD180 simulation model was provided by Analogspiceman here.

Traces are:
red=input voltage
green=UcD output voltage
blue=UcD filter inductor current
grey=13.4uF load capacitor voltage

It can be seen that classD amplifiers are actually capable of driving ESL. Problem with actual UcD circuit might be current limit since peak mosfet current in simulation exceeds 22A.

I also think that it would be possible to design classD amp without output filter using actual ESL speaker as a filter. It should be integrated with outpt transformer, or you would enter EMI hell. Also high voltage feedback might be a little problematic. Transformer losses from switching action of output stage might also be problematic, although my experiience shows that deltaB of less than 25mT at 80kHz is quite acceptable for standard iron cores.

Best regards,

Jaka Racman

[goudey]

Jaka, good work on the simulation of the ESL load.

I presume that you also included the resistor and inductor elements of the ESL model.

I have been wondering if the ucd amplifiers might be the "magic amplifiers" that work with ESLs. The ucd is strange in that its stability (or designed-in instability to ensure oscillation) depends upon oscillation where most loudspeakers present an inductive load (at > 400khz). Whether ESL or conventional, for such speakers the ucd output filter will dominate the open-loop gain at the oscillation frequency. The peak output current required is another issue.

I have thought of building a ucd-based ESL amplifier with the transformer installed inside the amplifier to form a simple second stage of RF filtering. Another approach, simpler to implement, is to move the ESL "tuning resistor" (the 1 ohm resistor I mentioned) into the amplifier case, and add an impedance compensated notch that would provide > 20 dB suppression between 400 and 500khz. The load seen by the amplifier would be about 4-6 ohms between 30khz and 1Mhz.

[analogspiceman]

Originally posted by goudey
I have been wondering if the ucd amplifiers might be the "magic amplifiers" that work with ESLs. [...] I have thought of building a ucd-based ESL amplifier with the transformer installed inside the amplifier to form a simple second stage of RF filtering.

This looks like a perfect application for the leapfrog method of switching amplifier design. I noticed that with critical damping, the ESL rolls off the upper treble starting at about 12kHz or so. This could probably be extended to >20kHz closed loop BW with leapfrog.

Actually, since the ESL is well damped, just a passive, open loop pre emphasize would probably do. -- a.s.

[goudey]

regarding ESLs and step-up transformers, I should mention:

The output voltage of the ESL step-up transformer can be lethal, or can lead to severe damage of vital organs. We haven't had many reports of death-by-electrocution arising from the use of electrostatic loudspeakers, but I am sure we would if the proper precautions are not taken to make accidental contact with the high voltages nearly impossible. For this reason I have discarded the idea of moving the step-up transformer into the amplifier case, though this may make sense if the amplifier case is then permanently installed and attached to the ESL loudspeaker.

regarding the leapfrog method:

I took a quick look at the writeup at the start of the leapfrog thread. It looks like the analysis of the design can be repeated using the ESL load model to determine the overall response. Such a design may offer good load independence (for ESL and conventional) and good filtering of the switching frequency.

Back to the ucd amps:

I am seriously considering acquiring some ucd400's to see how they work into an ESL. I think it will work, but I want the higher power ucd400 rather than the ucd180 simply because ESLs benefit from higher voltage and better current reserves.

[jmateus]

Let us know, Goudey, how the result was, it would be interesting to know.
My ESL were made from scratch and the transformers were
salvaged from an old tube amplifier, they are big ones for
40 watts, but I´m not sure if they are more than 50:1 turns
ratio even though they respond quite well.
What I´m trying to say is does this have any effect on the
impedance or otherwise parameters seen by the amplifier?

[goudey]

jmateus:

Quick answer: Nothing so far indicates a problem with your particular ESL setup, other than the somewhat low turns ratio and the resulting loss of peak sound output. The general question of compatibility of the ucd amplifiers with and ESL is unproven but plausible. Until proven any attempt should be considered an experiment and appropriate precautions should be taken, such as starting with a high value of the tuning resistor.

You can measure the impedance of the transformer/ESL panel yourself if you have the right equipment (oscillator, voltmeters, amplifier). If not, then we can only guess the parameters of the transformer. A 50:1 ratio may be correct for a 40 watt amplifier, though it could be more or less. For a 1200pF ESL panel this gives a 3 uF net capacitance as seen by the amplifier (multiplying by 50 squared).

You will need a tuning resistor in series with the transformer winding on the amplifier side, probably between 1 and 2 ohms (determined by experiment). The transformer resistance may be approximated by the sum of:
The DC resistance of the amplifier-side winding and
The DC resistance of the ESL side winding divided by the turns ratio squared.

The leakage inductance is unknown, but probably not of concern (can't be changed).


One option to ensure compatibility with most any amplifier is to use a 4 ohm tuning resistor and then fix the resulting high frequency rolloff by using electronic equalization before the amplifier. There is of course the issue complexity and additional peak voltage required, but high peak sound levels at such high frequences would cause hearing damage anyway, so maybe its not a problem.

[Arty123]

HI All,
With Quad ESL speakers (yes the old radiator-like monstrosities), the way to test if the amplifier is stable enough to drive the speaker used to be to have a 1KHz square wave input to the amp, 8Ohms load, and put a 2Uf capacitor in parallel to the load.
If there was no excesive "ringing" (overshoot on the leading and falling edges) on the output of the amp, it is deemed stable enough.
This has been borne out by numerous tests I did with almost any brand of amplifier. Some amps just burned if the output was over 10% of maximum power, and some, like Quad (of course) SAE, Crown and Krell did fine.

I haven't tested this with my modules, but that I think is the way to go....

Any comment is welcome, and I'l be trying this next week and get posted some real time results.

[goudey]

Arty123:

The test you are suggesting may be helpful for understanding general amplifier stability (with the help of a signal generator and an oscilloscope), but does not well represent an ESL load. The load you suggest presents an impedance of less than 0.2 ohms at 400khz, and an impedance of about 3.6 ohms at 20khz. The typical ESL impedance is very low at 20khz, and very high at 400khz. The ucd amps rely on self-oscillation at about 450khz, so the impedance < 0.2 ohms may be a problem for it.

jmateus:

When you destroyed you Sonic Impact did you have a resistor in series with the transformer winding on the amplifier side? Without one the ESL setup will very likely either cause a weak amplifier to shut down or blow up. A resistor in series is necessary.

[Bruno Putzeys]

I suppose I should chime in on the topic of UcD and electrostatics. I didn't exactly take the time to wade through the whole discussion so bear with me if I duplicate some answers.

A friend of mine has been using a pair of pre-hypex ucd's on his electrostatics for the better part of the last three years now, and before that the old SODA's. Intuitively one tends to worry about the high frequency behaviour of the combo. The load-independency of UcD makes this a non-issue. Even if you tie a pure 1uF capacitor straight to the output of the amplifier nothing noteworthy will happen. Besides, an electrostatic speaker is not a clean capacitance. Above 20kHz the leakage inductance of the stepup transformer takes over and the impedance becomes inductive.

No, the trouble lies at DC.

At DC, the speaker's impedance equals the resistance of the transformer primary, which is quite low. As long as you don't feed it DC, no problems. Unfortunately and surprisingly, AC coupling the input signal is not sufficient to prevent DC. As you may have seen on an oscilloscope, audio signals are asymmetrical. Any musical sound that has a second harmonic (=99.53% of all musical sounds) is asymmetrical. When you clip such a signal, you will inevitably chop off more on one side of the waveform than on the other. The result is that the chopped waveform has a DC component. This DC component produces a large current through the transformer primary and the short circuit protection cuts in.
This is not only a class D problem, any amplifier will run into trouble when clipped into a transformer load.

Another minor issue with half bridge amplifiers is that with a transformer load, not much DC offset at the output is tolerable before rail pumping occurs, tripping the overvoltage protection. Not normally a problem but not to be ignored either.

[patriz]

I saw the new ucd power supply and they have a DC protection , so looks like Hypex has choosed to put the realy on the rail . If so is it before or after the big power caps?
Giorgio