Hi folks,
would like toanswer Martins basic question.
1. A planar ESL will show an linear 3dB/octave level increase towards higher frequencies
2. Depending on panel width, level will drop 6dB/octave towards lower frequencies due to acoustical phase cancellation. E.g.a 30cm panel width will drop lower than 400Hz
3. At fundamental resonance of the tensioned membrane level will increase significant, showing a heavy peak, which can be +10dB higher than averaged panel level. This overshoot is more or less depending on ESL construction, means supporting the mebrane at fundamental resonance.
Capaciti
would like toanswer Martins basic question.
1. A planar ESL will show an linear 3dB/octave level increase towards higher frequencies
2. Depending on panel width, level will drop 6dB/octave towards lower frequencies due to acoustical phase cancellation. E.g.a 30cm panel width will drop lower than 400Hz
3. At fundamental resonance of the tensioned membrane level will increase significant, showing a heavy peak, which can be +10dB higher than averaged panel level. This overshoot is more or less depending on ESL construction, means supporting the mebrane at fundamental resonance.
Capaciti
Hi Martin Jan,
the 3dB increase is a result of better acoustical transformation at higher frequencies fpor dipole radiators.
A curved panel is equalizing this effect but only for frequencies higher than about 5 Khz and depending of curvature, will even drop slightly. As a result, the midrange is overshooting and needs reduction.
ML compensates for that by an overshooting filter at the low end and a "ringing" step-up at 10 Khz (like Calvin mentioned before). It sounds a little bit funny, but ML flatens the high frequency by using a technically worse transformer.
I would even make the statement, that using a "good" transformer will suffer from adequate high frequencie energy.
My biggest concern with some kind of curved panels is the fact, that curved panels only perform well if cross over frequency isn't to low. But cross-over frequency of about 250 Hz is exactly "where music lives". Even for the newer vantage or summit, i can recognize the transition from ESL to dynamic woofer easily. If people don't recognize they never played the right music to look for. E.G lets play a cello and you will recognize that the music stage will drop down to the woofer as the cello plays the lower octaves.
Capaciti
the 3dB increase is a result of better acoustical transformation at higher frequencies fpor dipole radiators.
A curved panel is equalizing this effect but only for frequencies higher than about 5 Khz and depending of curvature, will even drop slightly. As a result, the midrange is overshooting and needs reduction.
ML compensates for that by an overshooting filter at the low end and a "ringing" step-up at 10 Khz (like Calvin mentioned before). It sounds a little bit funny, but ML flatens the high frequency by using a technically worse transformer.
I would even make the statement, that using a "good" transformer will suffer from adequate high frequencie energy.
My biggest concern with some kind of curved panels is the fact, that curved panels only perform well if cross over frequency isn't to low. But cross-over frequency of about 250 Hz is exactly "where music lives". Even for the newer vantage or summit, i can recognize the transition from ESL to dynamic woofer easily. If people don't recognize they never played the right music to look for. E.G lets play a cello and you will recognize that the music stage will drop down to the woofer as the cello plays the lower octaves.
Capaciti
Hi Martin Jan,
a flat one having a width of 30cm will start beaming at about 800hz and beaming isn't the big issue, but interferences with the outer edges of a panel, which cause significant ripple in the frequency response.
ways to deal with: Curvature or segmentation, or a baffle size at least 2 times wider than the ESL-panel width.
Michael
a flat one having a width of 30cm will start beaming at about 800hz and beaming isn't the big issue, but interferences with the outer edges of a panel, which cause significant ripple in the frequency response.
ways to deal with: Curvature or segmentation, or a baffle size at least 2 times wider than the ESL-panel width.
Michael
Hi Michael,
Do you mean the interference caused by the baffle (-step)?
Maybe these could be adressed by rounding of the edge of the esl e.g. a tubular construction
I 'am asking since interference theoretically also is caused by the esl itself. The beaming of sound is a result of this interference.
I've still not decided whether to continue the curved project or to make another segmented. The thing that I not like about segmentation is it is a kind of RC-crossover but curved esls seem to have their own problems too
MartinJan
Do you mean the interference caused by the baffle (-step)?
Maybe these could be adressed by rounding of the edge of the esl e.g. a tubular construction
I 'am asking since interference theoretically also is caused by the esl itself. The beaming of sound is a result of this interference.
I've still not decided whether to continue the curved project or to make another segmented. The thing that I not like about segmentation is it is a kind of RC-crossover but curved esls seem to have their own problems too
MartinJan
Hi Martin Jan,
people might tell you a curved ESL is easy to build. Maybe it will work someway, but the first shot is certainly far away from what it could perform when optimized.
Take a look at Calvin's history and learning curve he underwent for years.
A flat one is much easier to build and will work well at a high percentage even for a prototype.
Don't bather with the R-C filtering needed for segmentation. I think you doubt about the possible negative influence to group delay, step response etc......, but those R-C filtered segmentation will provide best results if your final frequency response will be flat. if you can perfrom measurements, you will be astonished !
Finally i strong advice:
Noway if you build a curved or flat one, the mechanical symmetry of both stator halfs needs to be perfect, which is much more difficult for a curved one. E.g. a flat panel with 2mm stator-membrane distance
will loose at least 30% of its possibel performance (efficiency, maxpressure, distortions......) if the distance tolerance is larger than 0,2mm.
Capaciti
people might tell you a curved ESL is easy to build. Maybe it will work someway, but the first shot is certainly far away from what it could perform when optimized.
Take a look at Calvin's history and learning curve he underwent for years.
A flat one is much easier to build and will work well at a high percentage even for a prototype.
Don't bather with the R-C filtering needed for segmentation. I think you doubt about the possible negative influence to group delay, step response etc......, but those R-C filtered segmentation will provide best results if your final frequency response will be flat. if you can perfrom measurements, you will be astonished !
Finally i strong advice:
Noway if you build a curved or flat one, the mechanical symmetry of both stator halfs needs to be perfect, which is much more difficult for a curved one. E.g. a flat panel with 2mm stator-membrane distance
will loose at least 30% of its possibel performance (efficiency, maxpressure, distortions......) if the distance tolerance is larger than 0,2mm.
Capaciti
Hi Michael,
I've build several flat esls. While the first one wasn't the one I looked for, I made fast progress. Build unsegmented first, but added electrical segmentation later on. Despite my prejudgement of electrical segmentation (which has indeed to do with group delay, step response etc), the segmented esl sounded way better than the unsegmented esl.
The first curved wire esl prototype wasn't perfect. The symmetry was compromised and construction awkward. Yet, I think esls are quite forgiving at a relatively large spacings. I would give it another try, but it is rather expensive and time consuming, so I will only give it a try if it is theoretically possible achieve a flat response between 300 Hz and 20 kHz, without the 'dirty' tricks of ML. Unfortunately there is little literature available explaining the response of curved surfaces, so I fear I've to build and measure it first.....
I am thinking of a curved panel of 25 cm width, 100 cm long, 1,5 to 1,8 mm spacing and 1:50 step-up.
I've build several flat esls. While the first one wasn't the one I looked for, I made fast progress. Build unsegmented first, but added electrical segmentation later on. Despite my prejudgement of electrical segmentation (which has indeed to do with group delay, step response etc), the segmented esl sounded way better than the unsegmented esl.
The first curved wire esl prototype wasn't perfect. The symmetry was compromised and construction awkward. Yet, I think esls are quite forgiving at a relatively large spacings. I would give it another try, but it is rather expensive and time consuming, so I will only give it a try if it is theoretically possible achieve a flat response between 300 Hz and 20 kHz, without the 'dirty' tricks of ML. Unfortunately there is little literature available explaining the response of curved surfaces, so I fear I've to build and measure it first.....
I am thinking of a curved panel of 25 cm width, 100 cm long, 1,5 to 1,8 mm spacing and 1:50 step-up.
Hi Martin Jan,
if i would make a curved ESL it would be wire as well.
1. 25cm width is OK, but 1m height is a "playtool". Imaging will be limited and Music will not provide the "breath taking factor". So make it al least 140cm.
2. You will need a rigid and precise crossbar construction to avoid deviations of the curve and the eveness in height direction.
3. Your spacing need to be different, since the membrane will be closer to the backstator. So make about 1,5mm spacing for the backstator, stretch the membran as much as possibel in height and less as possibel in width and glue it to the backstator spacers.
4. Make the spacing for the front stator adjustable. Give it 0,5mm at the very start, assemble it without gluing.
5.Charge the ESL with HV and increase voltage until membrane collapses towards one of the stators.
6. As 0,5mm is much lower than 1,5 it should collapse to the front. Now increase the distance of the front stator by using self adhesive tape of 0,1mm and repeat collapse procedure.
7. There will be a spacing when the membrane change to collapse into the backstator. Remove one layer of tape (0,1mm) and you have found the best matching spacing relationship back to front.
8. This means some work to do, but believe me, the result will be the most mechanical stable, most efficient and important, lowest distorting curved ESL you can make !!!!!!
9. If you think you might be able to predict or pre-calculate the correct spacing you will end up building a lot of prototypes
Good luck,
Capaciti
if i would make a curved ESL it would be wire as well.
1. 25cm width is OK, but 1m height is a "playtool". Imaging will be limited and Music will not provide the "breath taking factor". So make it al least 140cm.
2. You will need a rigid and precise crossbar construction to avoid deviations of the curve and the eveness in height direction.
3. Your spacing need to be different, since the membrane will be closer to the backstator. So make about 1,5mm spacing for the backstator, stretch the membran as much as possibel in height and less as possibel in width and glue it to the backstator spacers.
4. Make the spacing for the front stator adjustable. Give it 0,5mm at the very start, assemble it without gluing.
5.Charge the ESL with HV and increase voltage until membrane collapses towards one of the stators.
6. As 0,5mm is much lower than 1,5 it should collapse to the front. Now increase the distance of the front stator by using self adhesive tape of 0,1mm and repeat collapse procedure.
7. There will be a spacing when the membrane change to collapse into the backstator. Remove one layer of tape (0,1mm) and you have found the best matching spacing relationship back to front.
8. This means some work to do, but believe me, the result will be the most mechanical stable, most efficient and important, lowest distorting curved ESL you can make !!!!!!
9. If you think you might be able to predict or pre-calculate the correct spacing you will end up building a lot of prototypes
Good luck,
Capaciti
Hi Few,
you are right, there is one with 80cm height. You need to consider that we offer ESL-panles for different customer expectations. This is meant for smaller hybrid systems where limited size and optics are priority
But if you ask me which panels are best, i answer the longest one you can achieve !
As Martin is doing the ESL for himself and, i guess he wants to get as much as possible performance, he should make no compromise.
BTW: there is another way to center the membrane between to curved stators, EVEN they have the same spacing:
1. You put the HV+ charge to the stators and the ground to the coated membrane. Now you are able to set different voltages to the both stators. Obviously you need higher for the front stator, so that the increased force will compensate for the curvature bending of the membrane.
2. easy trick to optimize: Sit in front of the panel and illuminate the mebrane with a light - you will see the bending of the membrane. Increase front voltage as long you will see that the mebrane is centered. Thats works pretty well.
3. Another advantage is that the differnet HV-voltages compensates for the differnet excursion by stator forces (higher at back, lower at front) and therefor improve K2 distortiions significant.
4. But one thing to considerr: i would not apply this method to isolated perforated matel sheet stators, since their capability to withstand the higher voltage potential is limited !!!
Capaciti
you are right, there is one with 80cm height. You need to consider that we offer ESL-panles for different customer expectations. This is meant for smaller hybrid systems where limited size and optics are priority
But if you ask me which panels are best, i answer the longest one you can achieve !
As Martin is doing the ESL for himself and, i guess he wants to get as much as possible performance, he should make no compromise.
BTW: there is another way to center the membrane between to curved stators, EVEN they have the same spacing:
1. You put the HV+ charge to the stators and the ground to the coated membrane. Now you are able to set different voltages to the both stators. Obviously you need higher for the front stator, so that the increased force will compensate for the curvature bending of the membrane.
2. easy trick to optimize: Sit in front of the panel and illuminate the mebrane with a light - you will see the bending of the membrane. Increase front voltage as long you will see that the mebrane is centered. Thats works pretty well.
3. Another advantage is that the differnet HV-voltages compensates for the differnet excursion by stator forces (higher at back, lower at front) and therefor improve K2 distortiions significant.
4. But one thing to considerr: i would not apply this method to isolated perforated matel sheet stators, since their capability to withstand the higher voltage potential is limited !!!
Capaciti
Sorry to interupt your thread but me and my best friend have been spending the last 4 month taking over the legacy of VRT transformers with regards to ESL x-formers. VRT was our tusted supplier from when we started our hobby some 3 years.
We are very courious with regards to the commercial x-formers from amplino and the ones in ML. Could any one of you please post a freq versus db plot for those x-formers unloaded and loaded with 680 pF wich equals an element size 100*30cm and spacer 2mm. If it's possible to do a primary impedanse measurement unloaded and with 680pF i would really love it
Ofcourse as stated in this thread the spread inductance is one of the limiting factors together with capacitance in the winding technique.
Anyway with the above mentioned element we have known accomplished a passive system with -3dB at 20 Khz and a 7 dB shelving at 200 Hz. The tranny is a 1:150. It have taken may hours put we are getting there.
Best regards
Svend Holby
We are very courious with regards to the commercial x-formers from amplino and the ones in ML. Could any one of you please post a freq versus db plot for those x-formers unloaded and loaded with 680 pF wich equals an element size 100*30cm and spacer 2mm. If it's possible to do a primary impedanse measurement unloaded and with 680pF i would really love it
Ofcourse as stated in this thread the spread inductance is one of the limiting factors together with capacitance in the winding technique.
Anyway with the above mentioned element we have known accomplished a passive system with -3dB at 20 Khz and a 7 dB shelving at 200 Hz. The tranny is a 1:150. It have taken may hours put we are getting there.
Best regards
Svend Holby
Hi,
in case You don´t know yet, but You can get data-sheets for the Amplimos. Unlike most other manufacturer who often just offer a price tag but no data at all, Amplimo lists the data so that You can evaluate with spice-based simulation tools already.
Your question isn´t very specific either, since Amplimo offers several models which additionally can be connected in more than one way.
Since the frequency plot depends very much on the connection and type of transformer and on the construction of the panel I assume You specify Your demands clearly.
I measured ML-Panels (Prodigy and Sequel) with the 1:75-Type St106PP and the resonance formed from strayinductance and capacitance was in both cases well above 20kHz. Still though frequency response was slightly tilted down since ML used a 12µm thick film in those panels which lead to a early rolloff. Tests with paralleled panels and loads up to 3nF still showed an resonance close to 20kHz.
If I may ask?
Why do You use a tranny with such a high transformation factor?
(Which is high even for an segmented stator and would be unpractically high for a unsegmented one)
Assuming usage of the panel from 200Hz and above why do You chose such a big d/s?
jauu
Calvin
in case You don´t know yet, but You can get data-sheets for the Amplimos. Unlike most other manufacturer who often just offer a price tag but no data at all, Amplimo lists the data so that You can evaluate with spice-based simulation tools already.
Your question isn´t very specific either, since Amplimo offers several models which additionally can be connected in more than one way.
Since the frequency plot depends very much on the connection and type of transformer and on the construction of the panel I assume You specify Your demands clearly.
I measured ML-Panels (Prodigy and Sequel) with the 1:75-Type St106PP and the resonance formed from strayinductance and capacitance was in both cases well above 20kHz. Still though frequency response was slightly tilted down since ML used a 12µm thick film in those panels which lead to a early rolloff. Tests with paralleled panels and loads up to 3nF still showed an resonance close to 20kHz.
If I may ask?
Why do You use a tranny with such a high transformation factor?
(Which is high even for an segmented stator and would be unpractically high for a unsegmented one)
Assuming usage of the panel from 200Hz and above why do You chose such a big d/s?
jauu
Calvin
Sorry for not beeing so informative.
The reason for using 1:150 is that we place the shelving on the panel side of the x-former. This reduces the load capacitance significantly seen from the transformer at the high end but of course in the midrange where the you have full blow through there you see the full capacitance but here it is of no concern.
Note:
The size of shelving is +7 dB at 250 Hz (here N=1:150) at 20 kHz "N" is 7 dB lover.
One thing that has pussled me is how one could make a passive shelving for the primary side.
My courisosity in amplino and others impedance / freq with load is just to compare with our results. Thanks for the model hint Capacity Ill look into it this weekend.
Our panel is run from 150 Hz with a 3order passive and with a high bias to get good sensivity. That is the reason for the 2mm D/S spacing.
The reason for using 1:150 is that we place the shelving on the panel side of the x-former. This reduces the load capacitance significantly seen from the transformer at the high end but of course in the midrange where the you have full blow through there you see the full capacitance but here it is of no concern.
Note:
The size of shelving is +7 dB at 250 Hz (here N=1:150) at 20 kHz "N" is 7 dB lover.
One thing that has pussled me is how one could make a passive shelving for the primary side.
My courisosity in amplino and others impedance / freq with load is just to compare with our results. Thanks for the model hint Capacity Ill look into it this weekend.
Our panel is run from 150 Hz with a 3order passive and with a high bias to get good sensivity. That is the reason for the 2mm D/S spacing.
Hi sholby,
i calculated a little bit:
If you want to go down to 150 Hz , your primary inductance need to be at least 4mH. Even a perfect stepup won't show better coupling between primary and secondary windings of about 0,002.
So 1: 150 and 680pf Panel capacitance results in high frequency response as follows:
response when serial DC-resistance is 2 Ohm:
i calculated a little bit:
If you want to go down to 150 Hz , your primary inductance need to be at least 4mH. Even a perfect stepup won't show better coupling between primary and secondary windings of about 0,002.
So 1: 150 and 680pf Panel capacitance results in high frequency response as follows:
response when serial DC-resistance is 2 Ohm:
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.