@Hans: I did a quick check on the test amp, looking at it's output response while loaded with the ESL63.
So the effect you mentioned (dropping drive voltage at high frequency) is there, but it is much less in magnitude than what I see at the secondary of the xformer. Interestingly, with load, the amp output is about 0.1dB higher in the lower band, 0.2dB at higher freqs. Weird.
Jan
So the effect you mentioned (dropping drive voltage at high frequency) is there, but it is much less in magnitude than what I see at the secondary of the xformer. Interestingly, with load, the amp output is about 0.1dB higher in the lower band, 0.2dB at higher freqs. Weird.
Jan
Attachments
Last edited:
Hi Steve,
Again a great contribution and of course a thing to add directly into the model.
Hans
Hi Jan,
Well, when it's not the amp, there must be another reason for the 19dB drop at 20Khz.
The 1Meg load places a significant load on your secondary as can be seen in your first and second recording below.
With 1M load, current drops 12dB from 2.8m@8Khz to 0.7m@20Khz, while without this load current drops 3.8dB from 2.8m@8Khz to 1.8m@20Khz.
So when you tried to regulate the secondary voltage to a constant level by measuring at your 1Meg/1K divider in the third image, you get wrong results of ca 12dB-3.8dB= 8.2dB@20Khz changing your current curve quite a bit.
When projecting the recording without 1Meg on top of the recording with a "constant" 250Volt, correcting the just calculated 8.2dB error will change the the currents at 20Khz to land exactly on top of each other.
Hans
.
Attachments
ref Baxandall, chapter 3 p 192, Loudspeaker and Headphone Handbook, editor Borewick.
IMHO a must read if you try to understand Quads electrostats. I seem to remember there was a pdf that could be downloaded, but I have the book...
/örjan
Hans, I understand what you are saying, but you assume that the secondary voltage remains the same with or without the 1M/1k divider. That I do not know, especially not at the high frequencies.
Also, I agree that the 20kHz currents become the same in magnitude, the curve is still quite different I think.
BTW I looked at electrostatic high voltage meters but the cheapest I found was € 5k ...
Jan
Last edited:
Here is the version when cross-coupling the inductors with K=0.3 for the adjacent and K=0.1 for those one place further.
I didn't have the time to further experiment with these two factors, but the model is attached for further experimenting.
The first image shows at the left resp. in blue the transformer current and in red the SPL output in front of the speaker without these inductor couplings, and at the right with the mentioned couplings.
It has definitely a positive effect on the FR between 5Khz & 8Khz.
Hans
.
I didn't have the time to further experiment with these two factors, but the model is attached for further experimenting.
The first image shows at the left resp. in blue the transformer current and in red the SPL output in front of the speaker without these inductor couplings, and at the right with the mentioned couplings.
It has definitely a positive effect on the FR between 5Khz & 8Khz.
Hans
.
Attachments
I tried another way to measure the high voltage. Connected a 1:100 probe, 100Mohm, to the AP input set to 100k (cannot be higher). Input is 20dBV.
At low frequencies it is very accurate at 60dB attenuation, but increases at higher frequencies. I will have to learn how to turn this into an EQ file and apply it to the measurements to correct them.
And so it goes ;-) .
Jan
At low frequencies it is very accurate at 60dB attenuation, but increases at higher frequencies. I will have to learn how to turn this into an EQ file and apply it to the measurements to correct them.
And so it goes ;-) .
Jan
Attachments
Jan, I assume you are going to direct drive the speaker from the point where the transformers connect to the delay line, do you know the approximate voltage required for full output and are you going to incorporate the ionization detection that Quad used to trigger the crowbar
Stuart