electrostats lacking punch
I would just like to chime in this discussion and mention that most comercial ESL are designed with no compensation for dipole phase cancelation. This has the effect of rolling off the output exponentially as frequency decreases. The rolloff begins at the frequency of wavelength 1/4 the min of the smallest dimension of the panel. This is mentioned in Roger Sanders book The Electrostatic Loudspeaker Design Cookbook. Without equilization or an enclosure an ESL is going to sound like a big tweeter. This is why people find ESL to lack punch and sound thin. The reason that no companies EQ their ESL is because their is no way to do it passively (in the crossover) without a loss of approx. 8dB. This is an unacceptable loss in output. The way to correct this problem is be employing active crossovers before the power amp and biamp the hybrid ESL. This is costly and makes ESL less compatable with conventional amplification. Also the correction with EQ will reduce the max output of the ESL due to the increased excursion induced by the EQ. ESL do have the ability to rock. They just have to be EQ. Or correct phase cancelation with enclosure (not as economical or wife accepted).
Just thought you should know that the industry is not making them sound as good as they should.
-Jason
I would just like to chime in this discussion and mention that most comercial ESL are designed with no compensation for dipole phase cancelation. This has the effect of rolling off the output exponentially as frequency decreases. The rolloff begins at the frequency of wavelength 1/4 the min of the smallest dimension of the panel. This is mentioned in Roger Sanders book The Electrostatic Loudspeaker Design Cookbook. Without equilization or an enclosure an ESL is going to sound like a big tweeter. This is why people find ESL to lack punch and sound thin. The reason that no companies EQ their ESL is because their is no way to do it passively (in the crossover) without a loss of approx. 8dB. This is an unacceptable loss in output. The way to correct this problem is be employing active crossovers before the power amp and biamp the hybrid ESL. This is costly and makes ESL less compatable with conventional amplification. Also the correction with EQ will reduce the max output of the ESL due to the increased excursion induced by the EQ. ESL do have the ability to rock. They just have to be EQ. Or correct phase cancelation with enclosure (not as economical or wife accepted).
Just thought you should know that the industry is not making them sound as good as they should.
-Jason
Quad reduces driver area at high
frequencies, flattening the response, and broadening the dispersion. This is one purpose of the multiple concentric rings in the stators of the ESL-63 and newer models. They also add two driver panels (ESL-63 and 988) with highs rolled off passively (resistor in series) to raise the bass a little (the 989 uses 4 extra bass only panels).
Correcting phase cancelation at low frequencies is possible with equalization to a point (limited SPL), but is ultimately limited by the diaphragm excursion of the ESL, just as equalization of conventional bass drivers in boxes works only for limited SPL, due to limits in linear excursion of the speaker cone.
Equalization can be accomplished by rolling off highs, or boosting bass, or some combination of the two. The two latter cases require active equalization, the first can be passive.
You can increase the excursion limits of the diaphragm but you will simultaneously sacrifice sensitivity, so in the end you don't gain much.
I_F
frequencies, flattening the response, and broadening the dispersion. This is one purpose of the multiple concentric rings in the stators of the ESL-63 and newer models. They also add two driver panels (ESL-63 and 988) with highs rolled off passively (resistor in series) to raise the bass a little (the 989 uses 4 extra bass only panels).
Correcting phase cancelation at low frequencies is possible with equalization to a point (limited SPL), but is ultimately limited by the diaphragm excursion of the ESL, just as equalization of conventional bass drivers in boxes works only for limited SPL, due to limits in linear excursion of the speaker cone.
Equalization can be accomplished by rolling off highs, or boosting bass, or some combination of the two. The two latter cases require active equalization, the first can be passive.
You can increase the excursion limits of the diaphragm but you will simultaneously sacrifice sensitivity, so in the end you don't gain much.
I_F
Re: electrostats lacking punch
In light of this, please explain the rather flat response of the Acoustat speakers? No compensation per se that I am aware of...
They are full range cells, btw...
They don't lack punch... or bass
Sensitivity is ~mid and up 80s... depending on model.
--- regarding using "direct drive tubes" to run ESL panels? You need to equalize that. The reason is that even though the tubes you choose may be high impedance and tend to match the impedance of the panels, they still only match best at *one* frequency corresponding to the reactance of the cells! This means that the output varies exactly the same way that it will when matched by a transformer of a specific ouput impedance.
In otherwords, you can trade bandwidth for sensitivity/output with respect to "center frequency" , just as you will with a single transformer matching an ESL panel...
The usual requirement for wide band operation with a tube amp is therefore equalization to obtain flat response...
_-_-bear
tacit_tactix said:
In light of this, please explain the rather flat response of the Acoustat speakers? No compensation per se that I am aware of...
They are full range cells, btw...
They don't lack punch... or bass
Sensitivity is ~mid and up 80s... depending on model.
--- regarding using "direct drive tubes" to run ESL panels? You need to equalize that. The reason is that even though the tubes you choose may be high impedance and tend to match the impedance of the panels, they still only match best at *one* frequency corresponding to the reactance of the cells! This means that the output varies exactly the same way that it will when matched by a transformer of a specific ouput impedance.
In otherwords, you can trade bandwidth for sensitivity/output with respect to "center frequency" , just as you will with a single transformer matching an ESL panel...
The usual requirement for wide band operation with a tube amp is therefore equalization to obtain flat response...
_-_-bear
Take a look at the interface schematic. Ignoring all the Rs and Cs, you've got two transformers, one for the bass and one for the treble, each with different turns ratios but with their primaries connected in parallel.
I actually measured the drive voltage to a pair of 1+1s as a function of frequency. It looks (qualitatively) a lot like an RIAA curve.
I actually measured the drive voltage to a pair of 1+1s as a function of frequency. It looks (qualitatively) a lot like an RIAA curve.
SY said:
Well, I wouldn't call it an RIAA curve...
There are two transformers.
They both drive the same panel(s)
However, if I understand the theory correctly, they have a different turns ratio in order to match the cells at different frequencies. This, per Peter Walker.
You can run the cells with just the "HF" or the "LF" xmfr. What you find is not exactly what you might expect. They overlap some many octaves. It is not a "hand off" like you would expect in an xover...
(iirc, the "HF" xfmr is still working fine <200 Hz... and the "LF xfmr is still contributing up to ~2khz... So if anything you would expect on the surface to find a nicely bumped up midrange?)
So, given the width of the model 4 vs. the 1+1 one can see by inspection that there is nothing particularly variable in the interface's frequency response with respect to the total width of the cells being used. The only thing that actually changes is the very LF "peak" which only is effected ~ 30-40Hz. via a secondary tap change on the LF xfmr.
I doubt that the 8" nominal cell width and total speaker width of the 1+1 happens to correspond to the 40 Hz. frequency, nor is this the sort of equalization referred to in the earlier post that we are discussing... ?
The *much wider* model 4 uses the identical interface, but with a different tap selected (turns ratio) to make the LF response around 30 - 40 Hz flatter.
Again, per Peter Walker, it is my understanding that the turns ratio of a drive xfmr can be selected for output or bandwidth, but not both... so in his Quad 57 we find xfmrs of different turns ratio for the HF and LF sections respectively... Strickland's patent teaches one way to effectively get the same effect but using only one ESL cell rather than two. Drive one ESL cell with two xfmrs...
different ratios match the reactance at different points, yielding flat response...
As far as I can see the series resistors and the series capacitors provide the necessary "xover" action, presumably as a first order network - HP for the "HF" tranny and LP for the "LF" tranny.
_-_-bear
Re: electrostats lacking punch
Above is the original quote...
the question at hand is/was the rolloff due to the width of the panel.
In the case of the Acoustat, the drive system clearly is not concerned with this particularly - otherwise the same transformers could not produce a flat response from a panel that is ~8" wide and one that is ~36" wide!
The matching in the Acoustat/Strickland method is concerned with matching the same panel's reactance at two different ranges - not with providing a voltage or amplitude equalization correction, and seemingly not with the "width" of the speaker at all.
While it is true that the result of matching the reactances turns out to create a "curve", this is essentially a consequence of the matching of the reactance...
Interestingly, I happen to have a Mk-121a interface on the bench today for repair, so I will shoot a curve and see what it looks like for yuks...
_-_-bear
tacit_tactix said:
SY said:
Above is the original quote...
the question at hand is/was the rolloff due to the width of the panel.
In the case of the Acoustat, the drive system clearly is not concerned with this particularly - otherwise the same transformers could not produce a flat response from a panel that is ~8" wide and one that is ~36" wide!
The matching in the Acoustat/Strickland method is concerned with matching the same panel's reactance at two different ranges - not with providing a voltage or amplitude equalization correction, and seemingly not with the "width" of the speaker at all.
While it is true that the result of matching the reactances turns out to create a "curve", this is essentially a consequence of the matching of the reactance...
Interestingly, I happen to have a Mk-121a interface on the bench today for repair, so I will shoot a curve and see what it looks like for yuks...
_-_-bear
I think that the baffle compensation is the reason for the multiple taps. I haven't measured with anything other than the 1+1 tap, but I'll bet that the first part of the curve, the dipole-cancellation part above the fundamental resonance, is what gets raised and lowered. Of course, the load is changing too as more panels are added, which interacts with the transformer ratio....
Do try to get a couple of panels as the load- I substituted some HV ceramic caps as the load and got very different results as I approached the fundamental resonant frequency of the panels (roughly 100Hz).
Do try to get a couple of panels as the load- I substituted some HV ceramic caps as the load and got very different results as I approached the fundamental resonant frequency of the panels (roughly 100Hz).
I have panels, but they were not connected to the unit.
Here are the results:
Acoustat 121a interface; no load; 0 db ref 1kHz. (measured ground/CT to one leg of the secondary, both xfmrs & coupling components used)
20Hz. +7
40Hz. +7
100Hz +7
200Hz. +5
500Hz. +2
1kHz. 0
5kHz. -1
10kHz. -1.5
20kHz. -1.5
The HF pot was randomly set by the owner of the unit (it's here for repair - hv section) so that accounts for the minus figures above 1kHz.
The unit was swept, I recorded approximate frequencies of interest... HP ACVTVM recorded the dB levels.
I also dunked in a 0.001ufd and a 0.01ufd as a load to see... no significant change was noted.
The tap was on the Model 4 setting - least "boost" via the LF xfmrs *primary* tap - I got that wrong in the earlier posts.
What's a 1/4 wave at 8"? Seems higher than 150 hz., seat of the pants...
_-_-bear
Here are the results:
Acoustat 121a interface; no load; 0 db ref 1kHz. (measured ground/CT to one leg of the secondary, both xfmrs & coupling components used)
20Hz. +7
40Hz. +7
100Hz +7
200Hz. +5
500Hz. +2
1kHz. 0
5kHz. -1
10kHz. -1.5
20kHz. -1.5
The HF pot was randomly set by the owner of the unit (it's here for repair - hv section) so that accounts for the minus figures above 1kHz.
The unit was swept, I recorded approximate frequencies of interest... HP ACVTVM recorded the dB levels.
I also dunked in a 0.001ufd and a 0.01ufd as a load to see... no significant change was noted.
The tap was on the Model 4 setting - least "boost" via the LF xfmrs *primary* tap - I got that wrong in the earlier posts.
What's a 1/4 wave at 8"? Seems higher than 150 hz., seat of the pants...
_-_-bear
this is an old thread, but I stumbled on it again,
You cannot EQ an electrostat panel to correct the low frequency losses due to difrraction and phase cancelation.
The low-mid-bass drop in SPL in electrostats is due to destructive interference between the pressure from the front and back of the diaphram. These are what is reffered to as phase minimum. In other words, increased output will still cancel itself out, because the pressure waves are equal and opposite, they cancel out, and no matter how strong those pressure waves are, they still are out of phase and will always cancel eachother out for this reason. The only way to stop them from canceling eachother is to stop them from interacting with eachother. You can do this by adding more width to your pannel, or building an enclosure, but you cannot equalize phase minimum destructive interference.
You cannot EQ an electrostat panel to correct the low frequency losses due to difrraction and phase cancelation.
The low-mid-bass drop in SPL in electrostats is due to destructive interference between the pressure from the front and back of the diaphram. These are what is reffered to as phase minimum. In other words, increased output will still cancel itself out, because the pressure waves are equal and opposite, they cancel out, and no matter how strong those pressure waves are, they still are out of phase and will always cancel eachother out for this reason. The only way to stop them from canceling eachother is to stop them from interacting with eachother. You can do this by adding more width to your pannel, or building an enclosure, but you cannot equalize phase minimum destructive interference.
Hi All
The explanation for the bass suck-out often heard in ESLs has a more complicated explanation than given by Tacit Tactix, and has nothing to do with the leakage of air from front to back.
The problem stems from the fact that ESLs have three different operating modes, with three different frequency responses.
Point source: If you are in the far field of the ESL, the SPL falls in proportion to the listener distance. In this mode the ESL produces an expanding spherical wavefront.
Plane source: If you are in the near field, very close to the ESL, the speaker produces a planar wave front that does not expand with distance and the SPL is constant with distance.
Line source: There is an intermediate case for a very tall narrow ESL, where it produces a cylindrical wavefront for which the SPL falls in proportion to the square root of the listener distance.
The Bass suckout occurs in the typical rectangular DIY ESLs because they exhibit all three modes, depending how close you are. That is, the shape of the frequency response changes with distance, and there is
no equalization scheme that will give a flat frequency response for all listening distances. If the equalization is adjusted to give a flat response at 1m say, then at greater distances the bass falls away faster than any other frequency.
The solution is to design the ESL to operate only in one mode – there are three solutions.
Point Source: for example, the quad ESL-63 generates a spherical wavefront using the annular segmentation and inductances between the segments (LC transmission line)
Planar source: make the speaker occupy the whole wall - the reflections from the walls make the ESL look infinitely large – no good for stereo, only speakers operating in this mode are near-field monitors and headphones. Otherwise useless.
Line-source: make a narrow floor-to-ceiling ESL - the reflections from the floor and ceiling make the ESL look infinitely long. Segmentation and resistors between each segment (RC transmission line) give it a flat frequency response. see Another segmented ESL for example.
Regards
The explanation for the bass suck-out often heard in ESLs has a more complicated explanation than given by Tacit Tactix, and has nothing to do with the leakage of air from front to back.
The problem stems from the fact that ESLs have three different operating modes, with three different frequency responses.
Point source: If you are in the far field of the ESL, the SPL falls in proportion to the listener distance. In this mode the ESL produces an expanding spherical wavefront.
Plane source: If you are in the near field, very close to the ESL, the speaker produces a planar wave front that does not expand with distance and the SPL is constant with distance.
Line source: There is an intermediate case for a very tall narrow ESL, where it produces a cylindrical wavefront for which the SPL falls in proportion to the square root of the listener distance.
The Bass suckout occurs in the typical rectangular DIY ESLs because they exhibit all three modes, depending how close you are. That is, the shape of the frequency response changes with distance, and there is
no equalization scheme that will give a flat frequency response for all listening distances. If the equalization is adjusted to give a flat response at 1m say, then at greater distances the bass falls away faster than any other frequency.
The solution is to design the ESL to operate only in one mode – there are three solutions.
Point Source: for example, the quad ESL-63 generates a spherical wavefront using the annular segmentation and inductances between the segments (LC transmission line)
Planar source: make the speaker occupy the whole wall - the reflections from the walls make the ESL look infinitely large – no good for stereo, only speakers operating in this mode are near-field monitors and headphones. Otherwise useless.
Line-source: make a narrow floor-to-ceiling ESL - the reflections from the floor and ceiling make the ESL look infinitely long. Segmentation and resistors between each segment (RC transmission line) give it a flat frequency response. see Another segmented ESL for example.
Regards
The width of the panel, and the crossover are chosen to minimize operating in this region, then the low frequency loss is minimized.
Basic dipole radiation pattern shows zero sound pressure at 90 degree, because the waves from the front and the back meet here and cancel out completely.
Basic dipole radiation pattern shows zero sound pressure at 90 degree, because the waves from the front and the back meet here and cancel out completely.
Last edited: