Jan,
Suppose you put AC of +/- X Volt on top of the Bias Voltage and both stators at equal potential.
At Bias minus X you have equal potentials between diaphragm and stators resulting in equal forces.
With Bias plus X you have different but still equal potentials between diaphragm and stators and so different but nonetheless equal forces.
So, the diaphragm won’t move when alternating but equal forces are pulling which is the case when stators are having the same potential.
When you have a different view, please explain your vision.
Hans
Suppose you put AC of +/- X Volt on top of the Bias Voltage and both stators at equal potential.
At Bias minus X you have equal potentials between diaphragm and stators resulting in equal forces.
With Bias plus X you have different but still equal potentials between diaphragm and stators and so different but nonetheless equal forces.
So, the diaphragm won’t move when alternating but equal forces are pulling which is the case when stators are having the same potential.
When you have a different view, please explain your vision.
Hans
Refer back to Post#585
Note again the force equation has 2 terms, one due to the applied signal, and the other due to negative tension.
Even with no applied signal voltage, the negative tension force still exists and is proportional to the square of the bias voltage.
Since the diaphragm always deflections toward one stator when charged, modulation of the negative tension force term by the bias voltage does results in hum/noise.
Note again the force equation has 2 terms, one due to the applied signal, and the other due to negative tension.
Even with no applied signal voltage, the negative tension force still exists and is proportional to the square of the bias voltage.
Since the diaphragm always deflections toward one stator when charged, modulation of the negative tension force term by the bias voltage does results in hum/noise.
Hi Steve, as always you gave useful background information.
Using the formulas 3.12 and 3.13, when the diaphragm is in the middle, they both result F=0 or no force on the diaphragm and no hum being transferred by bias ripple.
But when because of mechanical imperfections the diaphragm is pulled out of it's middle position, the compliance force comes into the equation and will consequently be modulated by ripple on the Bias voltage.
I have two questions in that respect.
1) have you any idea of the magnitude of these imperfections, could this, for sake of easier calculation, be addressed to one defined area and it's dislocation distance.
If so, math could help us trying to calculate the effect of modulation by the ripple.
2) since our ESL's are working in constant charge mode, what is the time to change the charge because of the bias ripple. When this takes quite some time, the charge does not change with the same amount as the ripple's magnitude, therefore most likely diminishing the effect of modulation.
Hans
Using the formulas 3.12 and 3.13, when the diaphragm is in the middle, they both result F=0 or no force on the diaphragm and no hum being transferred by bias ripple.
But when because of mechanical imperfections the diaphragm is pulled out of it's middle position, the compliance force comes into the equation and will consequently be modulated by ripple on the Bias voltage.
I have two questions in that respect.
1) have you any idea of the magnitude of these imperfections, could this, for sake of easier calculation, be addressed to one defined area and it's dislocation distance.
If so, math could help us trying to calculate the effect of modulation by the ripple.
2) since our ESL's are working in constant charge mode, what is the time to change the charge because of the bias ripple. When this takes quite some time, the charge does not change with the same amount as the ripple's magnitude, therefore most likely diminishing the effect of modulation.
Hans
Marcel,
Yes imperfection in symmetry has been mentioned several times also by me.
I simulated the effect of a 2nF cap on the bias ripple magnitude.
This only reduced the ripple on the bias generator side by a few dB’s.
But I never measured with the 2nF behind the existing 10MOhm resistor that’s in series with the neon bulb.
And that makes a huge difference.
With 10nF caps in the Bias generator, I see a 0.4V pk-pk 50Hz triangle wave after the 10MOhm.
With 20nF caps it becomes even 0.2V pk-pk.
I never realized that this 10MOhm is meant as a ripple filter, but now it’s completely obvious.
The triangle wave shape has a frequency spectrum that’s also much more rapidly decaying towards higher frequencies as on the primary side of the 10Meg resistor
As far as I can see, there are no such resistors included in the ESL 57, making bias ripple much more of an issue than in the ESL 63.
So maybe simple resistors instead of a complex 50KHz bias generator could have solved their hum problem as well.
Now I think we can safely assume that in the ESL 63 Bias ripple after being filtered to such low magnitude can safely be disregarded as causing hum, as is the case in my ESL's.
Hans
Yes imperfection in symmetry has been mentioned several times also by me.
I simulated the effect of a 2nF cap on the bias ripple magnitude.
This only reduced the ripple on the bias generator side by a few dB’s.
But I never measured with the 2nF behind the existing 10MOhm resistor that’s in series with the neon bulb.
And that makes a huge difference.
With 10nF caps in the Bias generator, I see a 0.4V pk-pk 50Hz triangle wave after the 10MOhm.
With 20nF caps it becomes even 0.2V pk-pk.
I never realized that this 10MOhm is meant as a ripple filter, but now it’s completely obvious.
The triangle wave shape has a frequency spectrum that’s also much more rapidly decaying towards higher frequencies as on the primary side of the 10Meg resistor
As far as I can see, there are no such resistors included in the ESL 57, making bias ripple much more of an issue than in the ESL 63.
So maybe simple resistors instead of a complex 50KHz bias generator could have solved their hum problem as well.
Now I think we can safely assume that in the ESL 63 Bias ripple after being filtered to such low magnitude can safely be disregarded as causing hum, as is the case in my ESL's.
Hans
Hi Hans,
Unless I'm missing something, you still seem to ignore the resistance of the diaphragm itself. It's distributed, which makes it more complicated to include in a lumped model such as the ones Spice uses than a lumped resistor. Instead of treating the diaphragm as a lumped capacitor, you will have to approximate it as a matrix of lumped resistors with capacitors at the nodes.
I think someone mentioned a ballpark figure for the sheet resistance somewhere in this thread, but I don't remember the value, only that it was high.
Best regards,
Marcel
Unless I'm missing something, you still seem to ignore the resistance of the diaphragm itself. It's distributed, which makes it more complicated to include in a lumped model such as the ones Spice uses than a lumped resistor. Instead of treating the diaphragm as a lumped capacitor, you will have to approximate it as a matrix of lumped resistors with capacitors at the nodes.
I think someone mentioned a ballpark figure for the sheet resistance somewhere in this thread, but I don't remember the value, only that it was high.
Best regards,
Marcel
Hi Marcel
Maybe you mean something different, but the resistance that I calculted, based on the Neon’s 20sec relaxation time in parallel to a 47nF cap fed from 5.25KV was ca 40GOhm.
So the 10MOhm feeding the diaphragms is so much lower that the 40GOhm has no influence on the ripple filtering, but the 2nF that I used has.
The 2nF seems o.k., but splitting that in a series of resistors with capacitors at it’s nodes seems a bit odd to me, because in the simulation’s signal flow, half of this capacity is really present over the whole audio spectrum, doubling as viewed from the diaphragm.
As far as I understand only keeping the charge on the diaphragm has to deal with these high resistance values.
But I may be completely wrong.
If so, could you please try to explain why.
Hans
Maybe you mean something different, but the resistance that I calculted, based on the Neon’s 20sec relaxation time in parallel to a 47nF cap fed from 5.25KV was ca 40GOhm.
So the 10MOhm feeding the diaphragms is so much lower that the 40GOhm has no influence on the ripple filtering, but the 2nF that I used has.
The 2nF seems o.k., but splitting that in a series of resistors with capacitors at it’s nodes seems a bit odd to me, because in the simulation’s signal flow, half of this capacity is really present over the whole audio spectrum, doubling as viewed from the diaphragm.
As far as I understand only keeping the charge on the diaphragm has to deal with these high resistance values.
But I may be completely wrong.
If so, could you please try to explain why.
Hans
A classic Quad 63 problem has been hum from the common ground of the housing. Some amps don't care. Some hum a lot, possibly oscillation in the amp manifesting as hum. I remember those complaints from when they were introduced. Virtually no speakers had power cords then. Another possibility could be AC leakage from the power transformer into the speaker signal or return lines. Has anyone measured that potential leakage? Even though there may be a ripple on the bias supply it makes sense to check everything else.
Hi Hans,
By design, a diaphragm of an ESL 63 panel does not move back and forth as one rigid whole. In particular, the annular stator rings in the mid-high panels (which panels actually also play back bass) are driven with delays to get patterns of diaphragm motion that simulate a point source 30 cm behind the loudspeaker. The bass panels have no segmented stators, but even then, their diaphragms don't exactly move as one rigid whole.
To prevent distortion, the charge on the diaphragm has to stay at its place when the diaphragm moves back and forth ("constant charge mode"). There are some slow second-order effects like the minor dynamic range expansion that was discussed earlier in this thread, but within one audio signal cycle, the charge shouldn't move much.
To achieve this, the diaphragm is made very high-ohmic, so its resistance keeps the charge from moving much within one audio signal cycle. Or rather, the distributed filtering caused by the diaphragm resistance and its capacitance to the stators does so.
When you try to modulate the charge on the diaphragm by putting a 50 Hz ripple on it (and thereby to amplitude-modulate the loudspeaker's sound when it plays music, which causes sidebands 50 Hz above and below each musical spectral peak, and possibly to just reproduce 50 Hz if there is some asymmetry somewhere), the very same diaphragm resistance and capacitance will filter the 50 Hz ripple. The piece of diaphragm right next to the point where you connect the 5.25 kV (or whatever it may be) supply may see nearly the full ripple, but the further from the contact, the less ripple you see.
In the schematic, node S+ is a small piece of a positive stator segment, S- a piece of negative stator, the resistor network a lumped approximation of the distributed resistance of a piece of diaphragm and the capacitors a lumped approximation of the distributed capacitance from S+ to the piece of diaphragm and from the piece of diaphragm to S-. As the capacitors are always pairwise in series, the diaphragm resistance doesn't affect the capacitance between S+ and S-.
Best regards,
Marcel
Last edited:
Hi Marcel,
Thx for explaining, but we are IMO talking about two different things.
What you describe, concerning the change in charge, is exactly my point 2 in #666 and also mentioned in #672. So when the time to change the charge is much longer as the 50Hz and thereby gradually moving over the surface, this will hardly lead to signal modulation.
But the point I mentioned is the capacity seen by the bias generator, which to my opinion should be ca 2nF.
So in effect we have the filter formed by 10Meg and 2nF giving its output to the diaphragm, where it’s effect is even further reduced by the slow change of charge.
Hans
Thx for explaining, but we are IMO talking about two different things.
What you describe, concerning the change in charge, is exactly my point 2 in #666 and also mentioned in #672. So when the time to change the charge is much longer as the 50Hz and thereby gradually moving over the surface, this will hardly lead to signal modulation.
But the point I mentioned is the capacity seen by the bias generator, which to my opinion should be ca 2nF.
So in effect we have the filter formed by 10Meg and 2nF giving its output to the diaphragm, where it’s effect is even further reduced by the slow change of charge.
Hans
If you are interested in the ripple at the point where the high-voltage supply is connected to the diaphragm, you have to take into account that that 2 nF capacitor has an enormous effective series resistance. Judging by the couple of seconds it takes an ESL 63 to reach full volume when you turn it on while there is a signal applied - which you should never try at high volumes! - the resistance must be of the order of 1 Gohm. That's nothing but a very coarse estimate of the low-frequency asymptote of the resistance, I don't know what the ESR or the effective series capacitance at 50 Hz will be.
When you are interested in the ripple voltage at all locations of the diaphragm, the huge sheet resistance of the diaphragm is bound to have more influence than that 10 Mohm at all points that are not very close to where the high-voltage supply is connected to the diaphragm.
When you are interested in the ripple voltage at all locations of the diaphragm, the huge sheet resistance of the diaphragm is bound to have more influence than that 10 Mohm at all points that are not very close to where the high-voltage supply is connected to the diaphragm.
Sometimes it helps to think about extreme cases.
Suppose the diaphragm were a good conductor, you would then essentially just have a lumped 2 nF loading your high-voltage supply.
Suppose the diaphragm would not conduct at all. You would then have no capacitive load, in fact nothing at all, loading your high-voltage supply.
Reality is somewhere in between.
Suppose the diaphragm were a good conductor, you would then essentially just have a lumped 2 nF loading your high-voltage supply.
Suppose the diaphragm would not conduct at all. You would then have no capacitive load, in fact nothing at all, loading your high-voltage supply.
Reality is somewhere in between.
Made some acoustical measurements today on a reworked Quad ESL63. Today is a average calm sunday in the city, in neighbourhood to the railway station. Dry weather conditions, neon recharge blinking every 57 seconds. Close miking such as in post 583, mic preamp of the soundcard (RME Babyface) set to max. gain, sampling set to 48kHz, FFT Filter set to 131072. Laptop/Soundcard on battery operation, being disconnected from the mains. All graphs are averaged 20 times.
First graph with input sine 0.283Veff. All audio gear on, as if I was listening to a reasonable audio signal input:
Funny, what is there at 200Hz? Close-up measurements from now on: 20dB min-max and F 20Hz ... 300Hz.
Next measurement, as before, but without any test signal from the souncard. Therefore, measuring ambient noise and some acoustical litter from the now "silenced" audio gear (Amp, CD-Player, DAC, ESL63):
No 50Hz peak, but still this 200 Hz artefact. Now switching all audio gear off except of the DUT:
Now disconnecting the Quad63 from the mains:
200Hz artefact is gone ... ??? !!!
Does anyone have any reasonable ideas about this 200Hz peak?
First graph with input sine 0.283Veff. All audio gear on, as if I was listening to a reasonable audio signal input:
Funny, what is there at 200Hz? Close-up measurements from now on: 20dB min-max and F 20Hz ... 300Hz.
Next measurement, as before, but without any test signal from the souncard. Therefore, measuring ambient noise and some acoustical litter from the now "silenced" audio gear (Amp, CD-Player, DAC, ESL63):
No 50Hz peak, but still this 200 Hz artefact. Now switching all audio gear off except of the DUT:
Now disconnecting the Quad63 from the mains:
200Hz artefact is gone ... ??? !!!
Does anyone have any reasonable ideas about this 200Hz peak?
Last edited:
O.K. that was the point where I was looking for additional info.
So you think the same 40Gig resistance that's loading the charge on the diaphragm, is also in series with the 2nF caps and the transformer impedance, that seems to make sense, although the transformer with it's high impedance is somewhat reducing the effectiveness of the 2nF caps in suppressing the ripple on the bias.
So looking at this, why bother at all about 50Hz hum and 50Hz AM modulation from the ripple on the bias voltage.
Open questions to me still are:
What could be the function of a 10Meg resistor in series with (distributed) 40Gig.
Why did Quad reduce the bias ripple by substituting 10nF caps to 20nF caps.
And why could Burkhard have hum problems with shorted audio inputs that were solved when generating Bias at 50kHz.
With filtering seemingly having it's -3dB point way below the audio spectrum, asymmetry in the stators could hardly be the cause for hum when the charge on the diaphragm remains practically constant.
Hans
Isn't 40 Gohm an estimate for the leakage? I'm assuming the leakage to be negligible, otherwise you also need to add leakage resistors to the network. My guesstimated 1 Gohm relates to the resistance of the diaphragm.
Reducing the electric shock you get when you touch the high-voltage supply output, reducing ripple on the piece of diaphragm closest to the connection to the supply, reducing distortion caused by that piece of diaphragm. That 1 Gohm or whatever it may be is distributed, after all.
No idea.
O.k. when a distributed 1Gig is more close to reality, let’s keep that value.
When I’m correct all panels are connected top and bottom over the whole length to the EHV, so with four ESL 63 panels we have in fact 8 half panels to be charged.
Demian mentioned the use of individual 100Meg resistors per panel, maybe that’s not such a bad idea after all and could help in reducing the edge sensitivities to bias ripple.
Hans