Possible reasons for low sensitivity in a full range ESL

[geraldfryjr]

Here is a simulation I did of a bias supply that I did the other day.
This is assuming that the input voltage is 230Vrms.
The frequency input has know effect on the voltage results.

With 230v in you should have at least 660v across the first capacitor as you didn't state what your input voltage range is.

For every two diodes (one stage) there is an aprox. 230v X 2.88=662.4v increase.

I like to run as high of a bias I can providing the stator insulation doesn't breakdown.
In which it has happend to me.

I was typically running about 7kv normally and as high as 10kv on some tests and the sensitivity just kept going up and I confirmed an increase of nearly 6db ( or was it 3db I don't remember which one,Sorry My bad) by each doubling of the bias voltage and was why I was trying to withhold a 14kv bias on a 1.8mm D/S.
This proved very difficult but not impossible.

It was my high signal level that my panels had met their demise.
With the data I did mange to write down I got as high as 89db to 91db at 1 meter at 2000HZ with a 5.56Kv bias and a 5vpeak signal into to a 1:256 transformation ratio for a small panel of 3.25" X 9.75" using test tones and noise signals.

It sounds to me that your bias voltage may not be were it should be and if you get it up above 3kv and closer to 5kv you should see the results that you are seeking.

Even though you have much large panels than I increasing the transformation ratio would help a great deal as well, try doubling it with another transformer and see if that helps providing that your amplifier can handle the lower impedance load.

But first get your bias supply issue solved first otherwise the extra drive will be barely noticeable.



jer:

[Legis]

Originally Posted by Bazukaz

Hi,

Is the multiplier ladder fed with low frequency(50 Hz) current?
If so, the multiplier can be significantly loaded by the input impedance of your voltmeter - even when measured at first stage.
So you can see incorrect reading.
Simulation shows this should be a problem(with 0.01uf caps) if multimeter has an input impedance of 1 megaohm, but in case of 10 megaohms or more the effect is small.

Regards,
Lukas.

Hi, the caps are 0.01µF but I think the supply's ladder and the trafo are fed by LM4950 with 700-900Hz sine (low-passed square). The multimeter should have 10meg input impedance for DC according to the link I posted.

[beun]

Let me chime in here a bit. I am the designer and builder of these stats. Currently I am preparing for the Lone Star Audio Fest show so I have limited time to answer questions, but let me attach a schematic of the supply. The oscillator indeed runs at about 800Hz and the jumpers are set into 'high' mode.

[bolserst]

Originally Posted by Legis

This is interesting. The supply is supposed to be variable between 3,5kV - 7kV. I measured the DC voltage accross the first capacitor and it was 122V at the minimum setting and 288V at the maximum setting. The ladder has 10 steps/capacitors. So it seems that the actual voltage range is approx. 1,22kV - 2,88kV? If I did the sensitivity measurement with only 2kV voltage it would explain low sensitivity as the sensitivity should be ~80,28dB at 1cm away from the panel according to the mentioned ESL simulator...

Since the input impedance of your DVM is 10Mohm, your measurements and calculations should be correct.
Looks like you have isolated the major cause of the lower than expected sensitivity.

It hit me last night....large curved panels, metalized diaphragm....I had seen a link to a build thread for an ESL just like this.
Atom66 posted it here:
Building an Electrostatic Speaker

Scanning thru the pages, it appears the final design for the HV supply was documented here:
Building an Electrostatic Speaker

Is this is the HV power supply unit you have? If so, two things you might try.

1) change the high/low jumper setting and re-measure the voltage across the first capacitor to confirm the labeling is correct.

2) I see that the voltage is varied by changing the amplitude of the square wave fed to the low pass filter and amplifier driving the step-up transformer. But, the output voltage is also dependent on the level of +Vcc supplied by the external power adapter which may or may not be regulated. You might measure this DC voltage between +Vcc and ground, then check with the builder to see if it is what he had intended when designing it.

[Legis]

Originally Posted by bolserst

1) change the high/low jumper setting and re-measure the voltage across the first capacitor to confirm the labeling is correct.

2) I see that the voltage is varied by changing the amplitude of the square wave fed to the low pass filter and amplifier driving the step-up transformer. But, the output voltage is also dependent on the level of +Vcc supplied by the external power adapter which may or may not be regulated. You might measure this DC voltage between +Vcc and ground, then check with the builder to see if it is what he had intended when designing it.

Hi bolserst, the bias was set to high mode as beun said. The 15VDC supply (rated 1A) is actually regulated so I think it should not be the problem. The power draw of the bias supply + it's supply appears to be below 1 watt, my power consumption meter cannot detect any consumption at all (error should not be over 10% in any inductive or capacitive load), so the 1A current capability is likely very sufficient.

[bolserst]

Originally Posted by Legis

Hi bolserst, the bias was set to high mode as beun said. The 15VDC supply (rated 1A) is actually regulated so I think it should not be the problem. The power draw of the bias supply + it's supply appears to be below 1 watt, my power consumption meter cannot detect any consumption at all (error should not be over 10% in any inductive or capacitive load), so the 1A current capability is likely very sufficient.

Hello Legis,

I see that beun, the designer, has joined the conversation. He should be able to help you track down the discrepancy in the HV output.
If the input(pin 1 to 2) and output(pin 3 to 7) voltages on the driver chip are the same as beun measured during design, you might check the transformer secondaries. Perhaps one of the secondaries has shorted leaving you with 1/2 the voltage output you would expect. Measuring voltage from transformer pins(1 to 2) and (3 to 4) should be similar in magnitude. Low output from one coil would indicate a short.

But, it seems extremely unlikely that both HV supplies had a similar failure at a similar time.
I'll leave it to you guys to sort it out.

@beun,
How did you measure HV output?
Hopefully Legis can reproduce your measurement process.

[beun]

Bolserst,

I cannot directly measure the HV-voltage, but I can check with the scope what the peak-peak signal voltage is at the output of the transformer. That nicely matches up with the simulation and from that I can deduct the HV-output.

[bolserst]

Originally Posted by beun

I can check with the scope what the peak-peak signal voltage is at the output of the transformer. That nicely matches up with the simulation and from that I can deduct the HV-output.

Ok.
What is the peak-to-peak signal level you see at the output of the transformer with the control set to maximum?
Legis should be able to measure Vrms with his DVM and and compare after factoring up by [2 x sqrt(2)].
This will confirm if his HV supplies are operating as intended.
As you said, the final HV-ouput can then be calculated from this information.

[Legis]

It looks like this case is now closed. The reason for lower than expected sensitivity was/is because the bias supply does not output what it should. I measured the supply again today, now at the trafo's output insted of across the first capacitor in the ladder. The results were in accordance to the first measurements I made. Bias set to maximum the output seen at the trafo's secondary is 209.7Vrms. 209.7V x 1.414 x 10 = ~3kV. So I had measured the sensitivity with approx 2.15kV bias insted of 5kV like I though, when I had set the bias setting to halfway.