^Stator thickness doesn't really matter much.

Is there no reason not to coat/paint the stators for insulation, you should do that, right? Don't see that mentioned often.

davidsh said:

^Stator thickness doesn't really matter much.

Is there no reason not to coat/paint the stators for insulation, you should do that, right? Don't see that mentioned often.

Click to expand...

I'm going to paint mine, just to be safe.

I made a electrostatic headphone by own design. I used wire instead of perforated plates. You can find a description on http://esl.hifi.nl/project09.htm
Tekst is in dutch but google translate is your friend

dude_500 said:

 
I doubt contact cement will conduct at all. Also, if going for the full effect of isolating the two sides with high resistance connection to bias, it would be necessary to keep the coating away from the edges or charge current could flow across the two sides that way. I think that would be the hardest thing.

Click to expand...

 
Well, here is my proposition... 
 
The copper track on the spacers occupies only half internal of their surfaces. 
 
Remember, each spacer and the copper on the stators are separated by a small gap (say 2 mm).
 
The half external part of the spacers is used for gluing them to the diaphragm. Both spacers have a small "ear" on the central and lower part of each, where the copper track finishes. This will provide a very secure and safe connection both to the polarization and to the membrane. 
 
Any comment welcome.

If I don't have access to a CNC machine to cut PCB, what would you guys suggest that I do for spacer material? I'm thinking plastic would work well because it's an insulator, but I don't know how I would get it in the shape/size I need.

chinsettawong said:

A heat gun can't shrink the diaphragm enough to make it stable, in my experience.
 

Click to expand...

 
Ah, OK. I was resigned to rebuilding them a few times. I'm trying to finish the HT supply.
 
I went to the library today, to use the laser printer. My UV lightbox is playing up, I haven't figured out how to mill PCBs but I know I can do toner transfer with the regular paper in the library if I keep the tracks wide. Anyway I forgot to take a vernier. When I got home the printout was undersize. Grrr.
 
w

toddthemetalgod said:

If I don't have access to a CNC machine to cut PCB, what would you guys suggest that I do for spacer material? I'm thinking plastic would work well because it's an insulator, but I don't know how I would get it in the shape/size I need.

Click to expand...

 
Oval/circular shape might be hard, but you can make some nice rectangular drivers without any CNC tools. Here are a few pictures of my early drivers before I got CNC access. These parts are all made with nothing but a dremel and a tiny drill press.
 

 

 

dude_500 said:

 

jcx said:

  one sided coating, the thickness of the diaphragm hardly matters to the E field, the force, balance since it is a small fraction of the spacing and the e_r of the Mylar even reduces the small effect of the thickness of the diaphragm more  - do you really get the spacing/centering/flatness within few microns?

Click to expand...

 
Here is the paper that outlines the concept of isolated coatings on each side: http://douglas-self.com/ampins/wwarchive/wwarchive.htm#diel
 
I went through the math a while back and it seems to be a correct assessment. Intuitively it seems absurd though, so possibly some math trick going on I fell for.

Click to expand...

the “trick” with the WW analysis is that the 2 conductive coatings on the micron thin Mylar membrane is a huge capacitance and therefore the front and back coatings are AC short circuited regardless of an R between front and back
 
the principle that the charges can move in the conductive coating is fine though and the E field force is proportional to the density of charges – they just totally miss the real geometry effect
 
the problem with conductive membranes is that the charges can move around within the conductor under the influence of the E field which varies as the distance between the conducting, polarized surfaces move
so even if no new charge comes in from the polarizing supply because of the big resistor, if the membrane coating is low resistance the charges can move over the surface of the diaphragm in the low resistance coating to be closer to the oppositely polarized stator
 
I ran up these sims in the toy student version of QuickField, the 256 node limit on the mesh doesn't get you far so the dimensions aren't to scale
 
the model geometry is axisymmetric about the bottom x axis,
 
right, left vertical lines are the radial sections of circular planes, our “stators” +1000 V left, -1000 V right
 
curved line in the middle is a radial section of a displaced diaphragm
 
in the 1^st conductive membrane sim the diaphragm is a uniform potential over the whole surface: +1000 V polarizing V
 
in the second the curve is given a constant surface charge density to approximate a high resistance coating
 
(“diaphragm” curve is random hand entered – not from anywhere in particular)
 
it turns out that the E field energy density (plotted in color) difference across the diaphragm is proportional to the surface force on the membrane (expand the units and you get n/m^2)
 
the 1^st pic's conductive membrane is being pulled harder in the middle where the charges in the conducting coating have been free to run to the center – this piling on of the force in the middle over deflects it as the membrane gets closer to the stator giving distortion
 
I think you can see that the 2^nd, “constant charge” situation shows much more uniform force on the diaphragm
 
 

 
 
I believe the ideal “constant charge” ESL implementation would be to “glue” some charges uniformly distributed over the surface of the membrane without having any surface conductor at all
but electrits don't seem to be good mechanically so we use high resistance coatings on the mechanically good Mylar – we want the resistance so high that the charges we trickle in with the polarizing V supply don't want to move across the surface because the RC time constant is much longer than any audio signal
this high surface resistance is not just to prevent the charges from moving on/off the membrane through the polarizing supply resistor but also the time constant should be too long for the charges to even move from one part of the diaphragm surface to another
 
 
coating both surfaces really doesn't matter - its the deflection/E field concentration/charge movement that causes the force to be nonlinear with applied V when the membrane is highly conductive

toddthemetalgod said:

If I don't have access to a CNC machine to cut PCB, what would you guys suggest that I do for spacer material? I'm thinking plastic would work well because it's an insulator, but I don't know how I would get it in the shape/size I need.

Click to expand...

You can find spacer material everywhere in your home. For instance insertable TAB index deviders for office binders are available in the right thickness and can be shaped with scissors or a Stanley knive. This material offen is polypropylene, which is difficult to glue, but good sanding will help.
 
You can use massive cardboard as well. For low resistance (coated) diaphragma's fine, but if you use a high resistance diaphragma its better to paint the cardboard to avoid the risk of leak of charge as carbonboard has a very low conductivity..

poecillia said:

You can find spacer material everywhere in your home. For instance insertable TAB index deviders for office binders are available in the right thickness and can be shaped with scissors or a Stanley knive. This material offen is polypropylene, which is difficult to glue, but good sanding will help.

You can use massive cardboard as well. For low resistance (coated) diaphragma's fine, but if you use a high resistance diaphragma its better to paint the cardboard to avoid the risk of leak of charge as carbonboard has a very low conductivity..

Click to expand...

All right, thanks .

After some research and estimations, it seems that the distorsion remains very low because the film is very thin !... 
 
Thus, we don't have to worry about the coating : only one side, will do the job.

Yet you don't want the charge to move when excursion of the diaphragm happens, I suppose? That'll introduce distortion as jcx lines out?

jcx said:

  the “trick” with the WW analysis is that the 2 conductive coatings on the micron thin Mylar membrane is a huge capacitance and therefore the front and back coatings are AC short circuited regardless of an R between front and back
 
the principle that the charges can move in the conductive coating is fine though and the E field force is proportional to the density of charges – they just totally miss the real geometry effect
 
the problem with conductive membranes is that the charges can move around within the conductor under the influence of the E field which varies as the distance between the conducting, polarized surfaces move
so even if no new charge comes in from the polarizing supply because of the big resistor, if the membrane coating is low resistance the charges can move over the surface of the diaphragm in the low resistance coating to be closer to the oppositely polarized stator
 
(...)
 
I believe the ideal “constant charge” ESL implementation would be to “glue” some charges uniformly distributed over the surface of the membrane without having any surface conductor at all
 
(...)

Click to expand...

 

davidsh said:

Yet you don't want the charge to move when excursion of the diaphragm happens, I suppose? That'll introduce distortion as jcx lines out?

Click to expand...

 
What is the minimum coating resistance that allows "constant charge" to avoid migration within the diaphragm surface and what is the minimum coating conductance to charge the diaphragm?
 
How long it takes to achieve full surface charge of the diaphragm with such "high resistance coating"?
 
With uncoated membranes, what is the sensitivity? If I push them hard, can the stators achive the air voltage breakdown? 
 
Suppose one could laser or photo etch shields with several patterns* and spray gun the coating chemical solution at a certain pressure and during certain time so that the coating is made at certain regions at certain thickness according to a given precision (variables: shield accuracy, proportion accuracy of elements in the solution, pressure in the spray gun, time spraying…).
 

 
When the animation starts, in diaphragm to the right, green color regions are coated surfaces of the diaphragm and blue color then represents uncoated Mylar. There are traces that connect the concentric coated regions. Which length and coating thickness the traces that connect the concentric coated regions must have in order to its resistance reduce charge migration from circle to circle?
 
Suppose 24 hours or more (48 hours, i.e.) to fully space charge all the concentric circles at the rated voltage bias. Since the traces seem to add serial resistance, what is the charge gradient or voltage gradient from the inner circles to the outer circles and how they scale up from hour one to hour 48?
 
Urghh, this seems very hard to achieve, so now a different perspective.
 
Now the diaphragm depicted to the left. The outer coating in the front side of such diaphragm (green) could be connected to 580V pro bias and the inner coating in the back side (yellow) could be connected to +230V normal bias**. Blue color represents uncoated Mylar at each side of such “double bias” and “double side” coated diaphragm. Would that increase linearity?
 
I have drawn the pattern according to low frequency vibration modes that arnaud described. I presume low frequencies cause more diaphragm displacement and charge displacement than midrange. But vibrations modes also change as we go up in frequency (arnaud simulated the stators, is the same vibration mode for diaphragms?):
 

arnaud said:

(...)
Examples below with the first 2 stator resonances at about 750Hz and 1400Hz, left is the old stator, right is the new one with smaller perforated area. In both cases I ignored the copper trace / etching so this is assuming 1mm FR4 (also I used isotropic average properties for the material even though its mechanical properties are actually somewhat direction dependent):
Mode 1 (730Hz for both stators):
 

 
 
 
Mode 2 (1460Hz for old stator, 1350Hz for new stator):
 

(...)

Click to expand...

 
Maybe segregating the pattern in east and west portion with 230v bias within the surfaces in the peak of the 1400Hz vibration mode?
 

Yellow (front 580V pro bias); green (back 230V bias); blue (uncoated Mylar).
 
Urghh, this also seems very hard to achieve. Perhaps there is no optimum diaphragm coating pattern for all vibrations modes of the diaphragm in the audio spectrum.
 
I do not know much about the physics behind this transducers, but this is certainly the coolest thread***.
 
--------------------------------
* Shields only used in production process.
 
** Or a -230V negative normal bias or -80V unstandardized negative bias, changing the coating pattern, I honestly do not know…
 
*** if I had a lab, I would build all crazy or stupid configurations just to test if they work… Maybe with theory background to simulate their behavior, I would probably build considerably less “prototypes”. Knowledgeable people in ee please do not shoot me if I said something stupid.

You should probably use 1.6mm for the PCB thickness, this is the most commonly encountered.

I have thought about how to make all these parts, including spacers, from regular PCB thicknesses, because thin PCB is not easy to obtain in the UK. If concentric stepped rings are used in the supporting structure, this is comparatively easy to achieve with a CNC mill, even if skimming the PCB proves too difficult.

w

wakibaki said:

You should probably use 1.6mm for the PCB thickness, this is the most commonly encountered.

I have thought about how to make all these parts, including spacers, from regular PCB thicknesses, because thin PCB is not easy to obtain in the UK. If concentric stepped rings are used in the supporting structure, this is comparatively easy to achieve with a CNC mill, even if skimming the PCB proves too difficult.

w

Click to expand...

 
I feel like getting sufficient repeatability for channel matching by either making stepped rings or skimming a PCB would require vacuum holddown on a machine specifically designed for highly repeatable z-axis mounting, and it still might change day to day (but that might be irrelevant for DIY). The relevant distances we're talking about for channel matching are tremendously small to easily repeat over a wide x-y area.
 
I did recently see a video of someone's PCB mill that scanned the x-y area electrically detecting the copper surface while dropping z every half inch or so to create a mapping of the z non-perfections and then adjusted the g-code in realtime to correct for a flat surface. By doing this they were able to mill circuit boards taking off effectively no more than the copper whereas you have to go fairly deep if you don't have some elaborate mounting scheme or a z-axis mapping system.