[tib8]

Don't get me wrong, I agree with what you said, it's just that for DIY it's very difficult to make light low resistance membrane. (as you mentioned the lighter you could find is 12um, so it's at least 4-5 times (including the gold) heavier than what we use here. High resistance stators shouldn't be so difficult to create - coating on a non-copper FR4.
Points 5-8 were for my experience with aluminium foil and low resistance stators - which is a terrible combination in my opinion. Capacitance is bigger than normal electrostatic headphones, because now the membrane is in the C equation (high resistance membrane is transparent for the capacitance stator-stator). Humidity was a big problem for me, because of good isolation - when there was a high humidity my ears were shocked with electricity, despite the megaohms resistors I had there.
By the way as I mentioned few times here, I use a different approach with my ES phones - pretty similar to these "inverted" ones.

Thank you for this reply - it has triggered many toughts in me.
I was looking at many of your previous post to find out what was your "inverted" construction exactly. During that I realized that you are (or you were) using a very similar stretching method, which I have discovered lately myself after several trials, and you also use the free air resonance to check membrane tensioning, even the resonance freq is pretty similar, so it is good to see some confirmation that I am on the right track.

Last weekend I have received high voltage cables for my oscilloscope, so I was able to measure my DIY amplifier finally. I was surprised how big harmoic distorsion components appeared at high volume when the input freq passed over 8-10kHz and they increased rapidly with the frequency.
I was surprised because the THD in the spice simulation was very low - BUT that simulation runs on 1kHz! So I repeated the THD simulation on 6kHz, 10kHz and 20kHz, and the simulation was in good correlation with the reality. Then I simulated several popular electrostatic amplifier designs from Kevin Gilmore and from STAX, and all of them showed the same behavior, even with smaller signals.
So lucky this is not my design fault, but it is due to the limitations of HW components. I am still thinking how to minimize this effect, but it has to be accepted that some distorsions are present in every amplifiers.

So back to the main topic: all these classic esl amplifers are symmetric, not single ended. And due to this they tend to cancel out the second harmonic distorsion of the signal and the third harmonic is stronger than the second. On the other hand Nelson Pass (an iconic amp designer) said that he made several experiments what kind distorsions the most people like (or accept more than others), and they find out that the second harmonic should be the biggest component and all higher harmonics should be lower than the previous one. (This is not meaning that you intentionally increase distorsion, but it means that you are trying to archtecture the amp in a way that the reamining small distorsions should follow the rule above.) From this the conclusion is that single ended amplifiers should be used.

So If you don't mind, I will likely try to replicate your completely inverted construction mainly because it can be driven with a single ended amplifier, which I would like to try and compare if it is really better...

 

[ts8051]

I am interested to hear if anyone tried Neil Pollock's headphone amps from Wireless world in 1979.

http://www.amateuraudio.fr/images/ww5.pdf

I am sure these are not a patch on Mr Gilmore's designs but I would have thought they would work better
than some re-purposed mains transformers.

is there a reason why no-one here seems to like this design?

-Steve

Hello,
I contructed the Neil Pollock's amplifier before many years and it has a mistake in circuit related to bias high voltage section. It has a potensiometer which after some hours of playing burned because of circuit error. The rest of the circuit plays very well. Note that the type of used transistors are obsolete and cannot be found any more. But as i constructed my headphones with success, i followed the stax prototype for 0.5mm spacing and 580v bias, i used a regular stax energizer.

 

[lkos98]

Thank you for this reply - it has triggered many toughts in me.
I was looking at many of your previous post to find out what was your "inverted" construction exactly. During that I realized that you are (or you were) using a very similar stretching method, which I have discovered lately myself after several trials, and you also use the free air resonance to check membrane tensioning, even the resonance freq is pretty similar, so it is good to see some confirmation that I am on the right track.

Last weekend I have received high voltage cables for my oscilloscope, so I was able to measure my DIY amplifier finally. I was surprised how big harmoic distorsion components appeared at high volume when the input freq passed over 8-10kHz and they increased rapidly with the frequency.
I was surprised because the THD in the spice simulation was very low - BUT that simulation runs on 1kHz! So I repeated the THD simulation on 6kHz, 10kHz and 20kHz, and the simulation was in good correlation with the reality. Then I simulated several popular electrostatic amplifier designs from Kevin Gilmore and from STAX, and all of them showed the same behavior, even with smaller signals.
So lucky this is not my design fault, but it is due to the limitations of HW components. I am still thinking how to minimize this effect, but it has to be accepted that some distorsions are present in every amplifiers.

So back to the main topic: all these classic esl amplifers are symmetric, not single ended. And due to this they tend to cancel out the second harmonic distorsion of the signal and the third harmonic is stronger than the second. On the other hand Nelson Pass (an iconic amp designer) said that he made several experiments what kind distorsions the most people like (or accept more than others), and they find out that the second harmonic should be the biggest component and all higher harmonics should be lower than the previous one. (This is not meaning that you intentionally increase distorsion, but it means that you are trying to archtecture the amp in a way that the reamining small distorsions should follow the rule above.) From this the conclusion is that single ended amplifiers should be used.

So If you don't mind, I will likely try to replicate your completely inverted construction mainly because it can be driven with a single ended amplifier, which I would like to try and compare if it is really better...

Thanks for the efforts to find and read my posts I would be also glad if someone else tries the fully inverted ES and to share/compare how they sound.
Well, I cannot tell my design is better or worse as sound quality. Since most of amps use negative feedback (except some class A tube amps) the distortion is mainly related to the depth (and type of the components used) of the negative feedback. With ESL amps it can't be too deep, because you have to amplify a small signal to a very large one, but since it's usually with a passive (resistors) components, the distortion is low.
So in this context differential output should be a bit better, because it needs twice lower amplification (twice deeper neg. feedback). It also cancels the common noise much better. Still I'm surprised you got such high distortion in high frequencies, I did not see this with my DIY amps, but that's maybe because I can't measure them very good (my setup can't measure below 0.4 % THD).
So I think a fully inverted ESH must have a bit worse distortions compared to the classic design and that's not because of the design, but because of the amplifier. Still the distortion of the both amp types can be quite low, so it is insignificant or lets say human ear can't distinguish it. What I like in my design I think I mentioned already, but I will repeat it here risking to bore all of the people here.
1. It can be done without sweat protection - membranes are coated only inside the "sealed" membrane-stator-membrane enclosure.
2. I bit more stable membranes - they are constantly sucked to each other, on negative side - they tend to touch the stator easier and they are bended - hence maybe some small non-linearity when moving.
3. Very low capacitance - theoretically 0, because there is just one stator and the membranes are of a very high resistance. Still the main capacitance of the cable remains, but in fact the amp can see as low as 60-80 pf opposed to the 100-120 of the classic design.
4. I think they have a better bass, but it's just my guess that this is because of the 2 membranes moving in unison.
5. Membranes can be glued to a thick and hard frames, without sacrificing the spacing to the stator.
6. The construction of the phones is relatively simple and can be assembled/disassembled easy.
On the other end you need 2 grills - for both sides, although I don't put one at the ear side - I just make sure the distance is big enough and I handle the phones quite careful. One of my pairs doesn't have even a foam in between, membrane is completely "naked" at my ear - you can imagine how careful you have to be not to touch or rip it.
But yeah, for sure there are some more pros and cons, but I like the design and I like how they sound, and that's most important for me. Currently I'm designing a new ones with a bit different approach - 2 stators in the middle, made with thin 0.5mm FR4 facing copper to each other (saving the isolation coating this way - I hate to do this!) and bend shape like this - )( , so they can be hard enough as an assembly. This allows me also to make very small holes (1mm or even smaller) diameter. The bend shape would also allow the already bended membranes to move further without touching the stator. The problem here is how to glue the 2 stators only in the middle and I will also have to increase a lot the bias voltages - to compensate for the bigger spacing.

 

[lkos98]

@tib8
I just noticed I forgot to ask about your stretching jig. I thought mine was good, (at least it's not bas as an idea), but lately I'm not using it, because of the following issues:
1. The wood is not slippery enough and when stretching the membrane gets wrinkles at sides and is unevenly tensioned.
2. The resonant frequency of the my jig is completely unrelated to the resonant frequency of the glued to the frame membrane. There simply is no correlation at all - sometimes my jig shows 30 Hz and one would expect lets say 120 for the glued membrane, but it goes to 150. Then with 35 Hz it should go to 175 or so, but it is 100.
3. I think depending on how much glue is applied, how long you wait for it to dry, in what sequence the frames are glued and pressed, and how hard they are pressed, and what is the ambient temperature/humidity - the tension changes quite a bit.
4. Maybe due to the uneven tensioning in the jig (or maybe because of the reasons mentioned in p3), the 4 membranes come out with slightly different resonance frequency +-10hz and those with lowest frequency become unstable.
Please share your design and your experience/results.
Thanks

 

[tib8]

Thanks for the efforts to find and read my posts I would be also glad if someone else tries the fully inverted ES and to share/compare how they sound.
Well, I cannot tell my design is better or worse as sound quality. Since most of amps use negative feedback (except some class A tube amps) the distortion is mainly related to the depth (and type of the components used) of the negative feedback. With ESL amps it can't be too deep, because you have to amplify a small signal to a very large one, but since it's usually with a passive (resistors) components, the distortion is low.
So in this context differential output should be a bit better, because it needs twice lower amplification (twice deeper neg. feedback). It also cancels the common noise much better. Still I'm surprised you got such high distortion in high frequencies, I did not see this with my DIY amps, but that's maybe because I can't measure them very good (my setup can't measure below 0.4 % THD).
So I think a fully inverted ESH must have a bit worse distortions compared to the classic design and that's not because of the design, but because of the amplifier. Still the distortion of the both amp types can be quite low, so it is insignificant or lets say human ear can't distinguish it. What I like in my design I think I mentioned already, but I will repeat it here risking to bore all of the people here.
1. It can be done without sweat protection - membranes are coated only inside the "sealed" membrane-stator-membrane enclosure.
2. I bit more stable membranes - they are constantly sucked to each other, on negative side - they tend to touch the stator easier and they are bended - hence maybe some small non-linearity when moving.
3. Very low capacitance - theoretically 0, because there is just one stator and the membranes are of a very high resistance. Still the main capacitance of the cable remains, but in fact the amp can see as low as 60-80 pf opposed to the 100-120 of the classic design.
4. I think they have a better bass, but it's just my guess that this is because of the 2 membranes moving in unison.
5. Membranes can be glued to a thick and hard frames, without sacrificing the spacing to the stator.
6. The construction of the phones is relatively simple and can be assembled/disassembled easy.
On the other end you need 2 grills - for both sides, although I don't put one at the ear side - I just make sure the distance is big enough and I handle the phones quite careful. One of my pairs doesn't have even a foam in between, membrane is completely "naked" at my ear - you can imagine how careful you have to be not to touch or rip it.
But yeah, for sure there are some more pros and cons, but I like the design and I like how they sound, and that's most important for me. Currently I'm designing a new ones with a bit different approach - 2 stators in the middle, made with thin 0.5mm FR4 facing copper to each other (saving the isolation coating this way - I hate to do this!) and bend shape like this - )( , so they can be hard enough as an assembly. This allows me also to make very small holes (1mm or even smaller) diameter. The bend shape would also allow the already bended membranes to move further without touching the stator. The problem here is how to glue the 2 stators only in the middle and I will also have to increase a lot the bias voltages - to compensate for the bigger spacing.

I do not have a CNC for FR4 milling, so I simply order nowdays PCBs from China. They are dirty cheap (with the cheapest shipping metod) you can get 5-10 piece 1/2/4 sided PCBs around 10$. They have so good quality that I do not feel the need for a CNC. The only drawback is that you have to wait a month for them to arrive.

Is this ")(" shape an important part of your design idea? If not, you could try to order 0.6mm 4 layer PCB with copper only on the middle layers (avoiding the grill holes a bit to get rid of the galvanized holes) and use the outer layers only at the signal connector. This would mean only approx 0.6/3 = 0.2mm FR4.
Or 0.6mm 2 layer PCB with only the normal green coating as isolator is likely also good: copper will be even closer to the membrane, but the isolation is weaker.
OR you could combine these two to create a virtually bent shape: the center of the stator could use the middle 2 layers, and the outer lets say 1/3 ring of the stator could use the outer two layers...

About glueing in the center: soldering could be also an option...

 

[tib8]

@tib8
I just noticed I forgot to ask about your stretching jig. I thought mine was good, (at least it's not bas as an idea), but lately I'm not using it, because of the following issues:
1. The wood is not slippery enough and when stretching the membrane gets wrinkles at sides and is unevenly tensioned.
2. The resonant frequency of the my jig is completely unrelated to the resonant frequency of the glued to the frame membrane. There simply is no correlation at all - sometimes my jig shows 30 Hz and one would expect lets say 120 for the glued membrane, but it goes to 150. Then with 35 Hz it should go to 175 or so, but it is 100.
3. I think depending on how much glue is applied, how long you wait for it to dry, in what sequence the frames are glued and pressed, and how hard they are pressed, and what is the ambient temperature/humidity - the tension changes quite a bit.
4. Maybe due to the uneven tensioning in the jig (or maybe because of the reasons mentioned in p3), the 4 membranes come out with slightly different resonance frequency +-10hz and those with lowest frequency become unstable.
Please share your design and your experience/results.
Thanks

Well, first I tried the usual bike tier around a wooden frame method, with "only a little", "a little more" and "a lot more" stretching (there is just enough space on the frame for 2 membranes - so better matching was not needed). Then I was using the results for a week. The "lot more" streched version was stable finally: it was not sucked all times to the stators, but the deep were completely missing. The others were relatively good after increasing the spacers from 0.6 to 1.2mm. (D=90mm active area, 580V bias)
Then I wanted to try 1um mylar, but at the first try they were broken on the strecher - so then I tried avoid stretching completely: I cleaned a flat surface, I put the mylar on it and smooth it out as much as possible, and drop the frame with the almost dry contact glue on it.
Of course these were immediately sucked to the stators, so I heated them to shrink a bit and used 1.2mm spacers. This way one of them become useabe. So I made a third one from 2um mylar with the same method - except I simply streched the mylar a bit on the table with my fingers to be smooth just before a helping hand dropped the glued frame on it (then pressed against the table of course) - this time heating was not needed. I reduced also the bias to 325V and finally they got stable.
I use this "asymmetric" pair since then (around two weaks), they go down to 30Hz easily and even with the 1.2mm spacing and the low 325V bias they are loud enough for my normal listening levels.
As you can see this is far from the wished exact, repetable method.
( I measured the resonance frequency of the ready membranes only beause of curiosity by blowing air on the membrane and running an FFT app on my phone and looking for the first peak. Unfortunately I do not have measurements to this latest pair, but I remember that the better ones were around 140Hz. )
Do you have any advice for improvement?

 

[tib8]

I am interested to hear if anyone tried Neil Pollock's headphone amps from Wireless world in 1979.

http://www.amateuraudio.fr/images/ww5.pdf

I am sure these are not a patch on Mr Gilmore's designs but I would have thought they would work better
than some re-purposed mains transformers.

is there a reason why no-one here seems to like this design?

-Steve

Well, I remembered a design, which is pretty much the same with modern parts:
https://nectarsound.net/wp-content/uploads/2020/09/Nectar1p0_HeadphoneAmp_Rev2_SCHEMATIC.pdf

There is no problem with these (except that the output transistor works agains only a simple resistor so likely it will not effectively use the full range of supply voltage), they can be a good start for the DIY amp hobby. Just be careful with the high voltages!

 

[ts8051]

Well, I remembered a design, which is pretty much the same with modern parts:
https://nectarsound.net/wp-content/uploads/2020/09/Nectar1p0_HeadphoneAmp_Rev2_SCHEMATIC.pdf

There is no problem with these (except that the output transistor works agains only a simple resistor so likely it will not effectively use the full range of supply voltage), they can be a good start for the DIY amp hobby. Just be careful with the high voltages!

do you have the parts list and a pcb ?

 

[chinsettawong]

If you want to make a good amplifier for your headphones, I highly recommend building one of Dr. Gilmore’s amps - KGST, KGSSHV Carbon, KG Ground-grid, for example. They aren’t that much more difficult to build.

 

[ts8051]

If you want to make a good amplifier for your headphones, I highly recommend building one of Dr. Gilmore’s amps - KGST, KGSSHV Carbon, KG Ground-grid, for example. They aren’t that much more difficult to build.

ok, I checked ebay and the seller for 2 micron membrane does not sell any more. Do you know other seller?

 

[lkos98]

@tib8
The )( shape is important for me for 2 reasons - 1st to make the stator sturdy, so it does not vibrate on electrostatic forces. (for my normal designs I use double sided 1mm FR4. 2-nd to "follow" the shape of the membranes (when +-360V is applied to them they bend against each other. With a flat stator the center of the membranes is closer to the stator. By the way, this also applies to the classic 2 stator design, because membrane is always attracted more to one of the stators. The option of ordering PCB is good, I have considered it many times, but this is going to be a pretty complicated board. And before I do that, I want to check if it's worth it - for CNC this is pretty simple job. I made already 4 FR4 0.6 mm stators for about 1 hour including the about 2000 holes.
Soldering in the middle is a bit of a challenge, I need to solder inside a 1mm hole a thin layer (35um) of copper.
Unfortunately I can't give you advise on improvement of the stretching, because I use the same wet "glass table" method. I stretch the membrane with a microfiber cloth to remove any wrinkles. Then as you do I just drop the frames and press for about half an hour with something flat and heavy. Your results are quite similar to what I measure with mine - about 140-160 Hz, +-360V and 0.6mm spacing. If I go higher they become unstable. Still I got pretty good bass measured down to 10-15 Hz. Once I got 5Hz, but with a DC amp (no decoupling capacitors) and membranes were unstable. By the way I measured many times the FR response and I think tension does not affect the bass so much as the sealing of the earpads.
Since I do this stretching once in a year or when I make a new model, I don't have much data to share. I believe tire jig is the answer, I just have to make one and do some experiments with different tensions and measurements. But the frames I have are only the ones I use in the phones and I'm too lazy to cut many of them to compare.
The sound level is also good for my taste, but it's far from a dynamic headphones. So if I want to listen very loud, the amp starts clipping.

 

[chinsettawong]

To my ears, I like 3 microns Mylar the best. It's also a lot easier to work with. Try it, and you'll like it.

 

[tib8]

@tib8
The )( shape is important for me for 2 reasons - 1st to make the stator sturdy, so it does not vibrate on electrostatic forces. (for my normal designs I use double sided 1mm FR4. 2-nd to "follow" the shape of the membranes (when +-360V is applied to them they bend against each other. With a flat stator the center of the membranes is closer to the stator. By the way, this also applies to the classic 2 stator design, because membrane is always attracted more to one of the stators. The option of ordering PCB is good, I have considered it many times, but this is going to be a pretty complicated board. And before I do that, I want to check if it's worth it - for CNC this is pretty simple job. I made already 4 FR4 0.6 mm stators for about 1 hour including the about 2000 holes.
Soldering in the middle is a bit of a challenge, I need to solder inside a 1mm hole a thin layer (35um) of copper.
Unfortunately I can't give you advise on improvement of the stretching, because I use the same wet "glass table" method. I stretch the membrane with a microfiber cloth to remove any wrinkles. Then as you do I just drop the frames and press for about half an hour with something flat and heavy. Your results are quite similar to what I measure with mine - about 140-160 Hz, +-360V and 0.6mm spacing. If I go higher they become unstable. Still I got pretty good bass measured down to 10-15 Hz. Once I got 5Hz, but with a DC amp (no decoupling capacitors) and membranes were unstable. By the way I measured many times the FR response and I think tension does not affect the bass so much as the sealing of the earpads.
Since I do this stretching once in a year or when I make a new model, I don't have much data to share. I believe tire jig is the answer, I just have to make one and do some experiments with different tensions and measurements. But the frames I have are only the ones I use in the phones and I'm too lazy to cut many of them to compare.
The sound level is also good for my taste, but it's far from a dynamic headphones. So if I want to listen very loud, the amp starts clipping.

Yes, soldering is can be difficult... I would apply some lead on the middle 10mm spot of the two not drilled panel and press them together with some (?) heated tool.
I would drill them together afterwards, and only after this would insert the (two half) spacer ring to get the )( shape.

 

[lkos98]

Yes, soldering is can be difficult... I would apply some lead on the middle 10mm spot of the two not drilled panel and press them together with some (?) heated tool.
I would drill them together afterwards, and only after this would insert the (two half) spacer ring to get the )( shape.

Yes, thanks, I thought of that, but I already drilled them . I noticed also that this 0.6mm FR4 I have, when overheated, the copper tends to peel off quite easy. My spacer ring is not two half, but a whole 1 piece, because I want to get the best possible seal between the )(. It could be done of course with some glue or silicon. Anyway I'll try few options to see how it goes.
@chinsettawong
I saw that the ebay seller is not there anymore, so sadly no mylar at all. Now we have to find another provider.

 

[chinsettawong]

the seller on Ebay is still there. https://www.ebay.com/itm/Electrosta...550352&hash=item1cfdb6be93:g:XBoAAOSwU9xUQPIj