A custom transformer will typically be more expensive than buying something off-the-shelf. There are also many, many details to get right. It is not something I would try on a first build (or on a 20th build for that matter, since I've never been tempted to design my own transformer after building many different ESL prototypes).
As golfnut suggested, ER Audio is one source for ESL-specific transformers:
http://www.eraudio.com.au/Electronics/electronics.html
A cheaper option may be using multiple Hammond 1609 transformers. The last segmented full-range ESL I made used 3 of them, for a total step-up of 150. I don't have long-term reliability data on this setup, but the overall performance was good in testing. The speaker was roughly 16 by 48 inches.
https://www.hammfg.com/part/1609
Another option is to to buy an interface from a commercial speaker and use its transformers (ESL-63 or equivalent transformers would be suitable, and they're probably easier to get than most others since they have been made for many years).
Jazzman/CharlieM has done a few segmented builds in recent years. His electrostatic section isn't full-range, but the segmentation ideas are still comparable.
https://www.diyaudio.com/community/threads/new-speaker-project-underway.349664/
golfnut used circuit board stators on his
https://www.diyaudio.com/community/threads/another-segmented-esl.234975/
As golfnut suggested, ER Audio is one source for ESL-specific transformers:
http://www.eraudio.com.au/Electronics/electronics.html
A cheaper option may be using multiple Hammond 1609 transformers. The last segmented full-range ESL I made used 3 of them, for a total step-up of 150. I don't have long-term reliability data on this setup, but the overall performance was good in testing. The speaker was roughly 16 by 48 inches.
https://www.hammfg.com/part/1609
Another option is to to buy an interface from a commercial speaker and use its transformers (ESL-63 or equivalent transformers would be suitable, and they're probably easier to get than most others since they have been made for many years).
Jazzman/CharlieM has done a few segmented builds in recent years. His electrostatic section isn't full-range, but the segmentation ideas are still comparable.
https://www.diyaudio.com/community/threads/new-speaker-project-underway.349664/
golfnut used circuit board stators on his
https://www.diyaudio.com/community/threads/another-segmented-esl.234975/
With Bolserst's help, I've been working on the transformer matching problem for ESLs for a couple of years (very much on the back burner though).
As a rough approximation, the transformer should be designed so that
transformer leakage inductance/ transformer winding capacitance = reflected (amplifier + lead inductance)/ ESL capacitance.
The reflected amplifier + lead inductance is the (output inductance of the amplifier + the inductance of the speaker leads)x step up ratio squared.
Consider an amplifier with 1 uH output inductance and 3m of speaker lead = 5 uH total. When reflected to the secondary (HV) side of a 1:100 transformer, this looks like 50mH of inductance. 50 mH in series with 2 nF capacitance has a cutoff frequency of 16 kHz. With a 1:200 step up ratio, the cutoff is 4 kHz. Any additional transformer inductance will pull this down further. You can see the importance of short speaker leads (or low inductance leads) and low amplifier inductance when driving ESLs with a large capacitance. Also a very low resistance in the primary is important (and an amplifier with current limit 🙂 )
Fortunately, transformers can be designed to trade leakage inductance and winding capacitance. There are three cases:
Plane source ESLs: Very large unsegmented ESLs have a very large capacitance > 1 nF. The only way to drive these is with transformers of very low leakage inductance, which requires large toroidal cores and/or interleaved windings. The transformers have large capacitance, similar to the ESL. These ESLs have a flat frequency response when driven by normal voltage amplifier and listened to in the near field where the wavefront is flat - so they are an ideal plane wave source. Step-up ratios of 1:60 are close to maximum unless you accept a cutoff below 20 kHz.
Summary: transformer with high capacitance, low leakage inductance.
Point-source ESLs: The Quad 63 family of ESLs are segmented and connected as an LC transmission line, a.k.a. delay line. See wikipedia entry on telegraphers equations. The LC transmission line has a pure resistive impedance - so the ESL appears to have zero capacitance! That means the transformer should have small capacitance and large leakage inductance. That means short, deep windings on small cores and relatively large step up ratios. The QUAD ESL transformers have an enormous leakage inductance of several henrys and a capacitance of about 30 pF. The delay line behaviour combined with the circular segmentation causes the quad ESL to mimic a point source - it has a flat frequency response in the far field. Because the ESL is resistive, the ESL does not alter the resonant frequency of the transformer, so it can have a resonant frequency as low as 20 kHz. Step up ratios of 1:200 easy. Carefully optimised transformers matched to amp and ESl could probably do 1:250.
Summary: transformer with low capacitance and high leakage inductance.
Line source ESL: The segmented ESLs with resistors connecting the different segments form an RC transmission line. This case is between the other two extremes (above). The RC transmission line has an impedance partly resistive and partly capacitive with the capacitance about 1/20th of an unsegmented ESL of the same area and spacing. That means the transformer should have modest leakage inductance and modest capacitance. The much reduced ESL capacitance means it shifts the resonant frequency only a small amount, so the unloaded resonance can be as low as 24 kHz, for 20 kHz bandwidth. The RC transmission line gives a frequency response that is perfect for a line source (cylindrical wavefront). So the ESL should be floor-to-ceiling and use the floor and ceiling as mirrors to make it look like an infinite line source. It operates in the far field with respect to its width, and the near field with respect to its height. Step up ratios of 1:150 very practical, could probably be pushed to 1:200 with careful design. A largish UI core with a winding on each side works well.
Summary: transformer with medium leakage inductance and medium winding capacitance.
Hope this helps
best wishes for the new year
As a rough approximation, the transformer should be designed so that
transformer leakage inductance/ transformer winding capacitance = reflected (amplifier + lead inductance)/ ESL capacitance.
The reflected amplifier + lead inductance is the (output inductance of the amplifier + the inductance of the speaker leads)x step up ratio squared.
Consider an amplifier with 1 uH output inductance and 3m of speaker lead = 5 uH total. When reflected to the secondary (HV) side of a 1:100 transformer, this looks like 50mH of inductance. 50 mH in series with 2 nF capacitance has a cutoff frequency of 16 kHz. With a 1:200 step up ratio, the cutoff is 4 kHz. Any additional transformer inductance will pull this down further. You can see the importance of short speaker leads (or low inductance leads) and low amplifier inductance when driving ESLs with a large capacitance. Also a very low resistance in the primary is important (and an amplifier with current limit 🙂 )
Fortunately, transformers can be designed to trade leakage inductance and winding capacitance. There are three cases:
Plane source ESLs: Very large unsegmented ESLs have a very large capacitance > 1 nF. The only way to drive these is with transformers of very low leakage inductance, which requires large toroidal cores and/or interleaved windings. The transformers have large capacitance, similar to the ESL. These ESLs have a flat frequency response when driven by normal voltage amplifier and listened to in the near field where the wavefront is flat - so they are an ideal plane wave source. Step-up ratios of 1:60 are close to maximum unless you accept a cutoff below 20 kHz.
Summary: transformer with high capacitance, low leakage inductance.
Point-source ESLs: The Quad 63 family of ESLs are segmented and connected as an LC transmission line, a.k.a. delay line. See wikipedia entry on telegraphers equations. The LC transmission line has a pure resistive impedance - so the ESL appears to have zero capacitance! That means the transformer should have small capacitance and large leakage inductance. That means short, deep windings on small cores and relatively large step up ratios. The QUAD ESL transformers have an enormous leakage inductance of several henrys and a capacitance of about 30 pF. The delay line behaviour combined with the circular segmentation causes the quad ESL to mimic a point source - it has a flat frequency response in the far field. Because the ESL is resistive, the ESL does not alter the resonant frequency of the transformer, so it can have a resonant frequency as low as 20 kHz. Step up ratios of 1:200 easy. Carefully optimised transformers matched to amp and ESl could probably do 1:250.
Summary: transformer with low capacitance and high leakage inductance.
Line source ESL: The segmented ESLs with resistors connecting the different segments form an RC transmission line. This case is between the other two extremes (above). The RC transmission line has an impedance partly resistive and partly capacitive with the capacitance about 1/20th of an unsegmented ESL of the same area and spacing. That means the transformer should have modest leakage inductance and modest capacitance. The much reduced ESL capacitance means it shifts the resonant frequency only a small amount, so the unloaded resonance can be as low as 24 kHz, for 20 kHz bandwidth. The RC transmission line gives a frequency response that is perfect for a line source (cylindrical wavefront). So the ESL should be floor-to-ceiling and use the floor and ceiling as mirrors to make it look like an infinite line source. It operates in the far field with respect to its width, and the near field with respect to its height. Step up ratios of 1:150 very practical, could probably be pushed to 1:200 with careful design. A largish UI core with a winding on each side works well.
Summary: transformer with medium leakage inductance and medium winding capacitance.
Hope this helps
best wishes for the new year
Last edited:
sorry
I made a mistake in the pasting language.
I will just write what I wrote the other day.
I will write the rest separately.
Hello maudio, mattstat, golfnut,
Thank you for letting us know the specific goals of the transformer.
I don't know if it can be realized, but I will try to talk with the transformer manufacturer about the largest possible core and winding ratio. 8kVpp output at 50Hz. It can be expensive and difficult to order.
My idea was to first make a speaker system like AudioStatic-DCI and then improve it depending on the results.
From what you've said, it sounds like segmentation is a must.
Please give me a good example for reference.
Just in case, tell me about segmentation.
1: Isn't it a way to avoid ESL beams by using resistors generated using esl_seg_ui.exe?
2: You mean a speaker system that uses an ESL panel and a transformer for each band, separated by a dividing network? Like ESL-63.
I made a mistake in the pasting language.
I will just write what I wrote the other day.
I will write the rest separately.
Hello maudio, mattstat, golfnut,
Thank you for letting us know the specific goals of the transformer.
I don't know if it can be realized, but I will try to talk with the transformer manufacturer about the largest possible core and winding ratio. 8kVpp output at 50Hz. It can be expensive and difficult to order.
My idea was to first make a speaker system like AudioStatic-DCI and then improve it depending on the results.
From what you've said, it sounds like segmentation is a must.
Please give me a good example for reference.
Just in case, tell me about segmentation.
1: Isn't it a way to avoid ESL beams by using resistors generated using esl_seg_ui.exe?
2: You mean a speaker system that uses an ESL panel and a transformer for each band, separated by a dividing network? Like ESL-63.
@golfnut: excellent write up. I fully agree. All the folks that try to improve a Quad by installing "better" transformers should read this.
The effect of the output inductor of the amp on the hf response is a good point as well, very often overlooked. It once fooled me too, got very dissapointing measurements from one of the transformers I built. Took me a while to find out the amp coil was ruining my measurements...
The effect of the output inductor of the amp on the hf response is a good point as well, very often overlooked. It once fooled me too, got very dissapointing measurements from one of the transformers I built. Took me a while to find out the amp coil was ruining my measurements...
@ Chase: google Audiostatic interface for a reference. Audiostatic uses 44 wires, the middle 8 being driven directly the other wires (18 on each side) through a resistor of around 180k. Capacitance is around 320pf for the entire unsegmented panel for the smaller ES200 model. This is the most basic way, only one strip isolated so two sections. However, for good off axis response it is better to use three sections or more.
The esl_seg_ui.exe tool is excellent for modelling this.
The esl_seg_ui.exe tool is excellent for modelling this.
I was looking at and considering various opinions.
Segmentation that requires multiple panels and a network seems to be difficult for me. I was impressed that many people were challenged.
I saw a picture of golfnut's ESL, but it is beautiful with an aluminum frame. cool!
I thought in my heart that I could aim for a higher level by spending more money on the step-up transformer.
Golfnut's experience and commentary will be a learning experience. wonderful.
Either change the policy to the recommended segment ESL, reduce the panel size, and use the full range of ER Audio for the transformer,
We need to rethink from the ground up.
The system that maudio wrote in #27 might be realistic for me. I felt like learning about this.
Thank you everyone.
Segmentation that requires multiple panels and a network seems to be difficult for me. I was impressed that many people were challenged.
I saw a picture of golfnut's ESL, but it is beautiful with an aluminum frame. cool!
I thought in my heart that I could aim for a higher level by spending more money on the step-up transformer.
Golfnut's experience and commentary will be a learning experience. wonderful.
Either change the policy to the recommended segment ESL, reduce the panel size, and use the full range of ER Audio for the transformer,
We need to rethink from the ground up.
The system that maudio wrote in #27 might be realistic for me. I felt like learning about this.
Thank you everyone.
What is your intention? To build your own esl panels and transformers?
If you provide us with some more details about your plans, we can probably provide more accurate advice here
If you provide us with some more details about your plans, we can probably provide more accurate advice here
I know the last time the posting was over a year ago. But I thought I would add a few extra bits of information. For those that build multiple cells and arrange them in a curved pattern, then Dayton-Wright had the solution. Look for someone selling their transformers. I had seen them as cheap as $150 each, but the freight will add to that.
These transformers are designed with a 20kV voltage rating, and have those porcelain insulators at the output. Even with constant high voltage 10kV or more I have heard no failures, some of those speakers are now 50 years old. There are several advantages to these. They have four primaries each rated at 8 ohms and the output is specified at 80 000 ohms. If the inputs are wired in series, you get a 100:1 turns ratio. So two in parallel and then the pairs in series gives you 50:1. At the highest ratio, their response with the DW panels extends to about 20kHz.
Now the reason I mention this is, that it allows you to adjust the transformer inductance that way. The DW panels themselves were designed with utmost care to have almost no parasitic capacitance and an array of 10 cells which covered almost 1 sq meter had only 620 pF capacitance. Now there is one advantage to high transformer inductance, that with the appropriate series capacitor, in this case around 800uF, you can create a bass boost to lower the low frequency cutoff of the panels. In the case of DW, this was around 30-40Hz.
These transformers are designed with a 20kV voltage rating, and have those porcelain insulators at the output. Even with constant high voltage 10kV or more I have heard no failures, some of those speakers are now 50 years old. There are several advantages to these. They have four primaries each rated at 8 ohms and the output is specified at 80 000 ohms. If the inputs are wired in series, you get a 100:1 turns ratio. So two in parallel and then the pairs in series gives you 50:1. At the highest ratio, their response with the DW panels extends to about 20kHz.
Now the reason I mention this is, that it allows you to adjust the transformer inductance that way. The DW panels themselves were designed with utmost care to have almost no parasitic capacitance and an array of 10 cells which covered almost 1 sq meter had only 620 pF capacitance. Now there is one advantage to high transformer inductance, that with the appropriate series capacitor, in this case around 800uF, you can create a bass boost to lower the low frequency cutoff of the panels. In the case of DW, this was around 30-40Hz.
Dayton-Wright's ESL design was meant to be full-range, esp when doubled as pairs (which were made to dove-tail lock together). Very large stator spacing. His transformers, 1:100, weigh 32 lbs each - a lot of copper.
Just for informational purposes:
https://www.canuckaudiomart.com/details/650016343-transformer-for-diy-electrostatic-speakers/
Just for informational purposes:
https://www.canuckaudiomart.com/details/650016343-transformer-for-diy-electrostatic-speakers/
Last edited:
@bentoronto: You have the cells with the plastic stators, acprevious generation to mine. If you have ever mesured the stator to stator capacitance, I would be interested to know what the value is. The transformers were not changed, so the capacitance should not be much different