ESL questions
bbr...@my-deja.com
I am starting to build some ESLs for a self-guided project class
here at Caltech. Let me tell you where I am and then ask some
questions:
I'm using perforated sheet steel stators, 63% open area, 40" by
8". I'm insulating them with appliance spray paint. The spacers
are 3/32" plexiglass, spaced in .25" on each edge. This makes
the effective driven area approx 39.5" x 7.5". The diaphragm
is the apparently much-revered 6 micron C-grade mylar.
I have a stretching device that I found that is usually used for
high tension silk screens (the kind you use to put graphics on
t-shirts and such). It is very fast to operate and I can put
lots of tension on the mylar. The biggest disadvantage is that
I can't get a very good quantitative measurement of how much
tension I'm putting, but this is offset by the fact that my
machine is 60" x 30", so I can easily stretch, graphite, and glue
two panels out of the same-tensioned piece of mylar.
I used regular old graphite rubbed in and I maintain at least
100kOhm at one inch intervals across the surface.
My power supply is built from the schematic provided by Mark
Rehorst on his web site. It uses an Emco DC HV supply, which
takes something like 0-20VDC and steps it up to 0-4kV. This
makes for convenient bias voltage tweaking.
I purchased two "matching" tube output transformers off EBay.
They have an 8kohm secondary and variable (2,4,8,16) primary
impedances, and a nominal turns ratio of 40. They are audio
rated 20-20kHz.
I have been very excited with the results from my first stereo
pair, but I have many questions. I'll just write a bunch of them
here and if any of you can help or even suggest a means to
answering just one of them, I'll be delighted.
1. My panels seem to be up at least 10dB at high frequencies.
(Probably ~10kHz.) I haven't made a measurement but I trust my
ears. It feels like it rolls up between about 1k and 10k.
Response seems reasonable flat between probably 200-300Hz and 1k.
Obviously, it rolls off below, as with any estat. My question
is, why the boosted highs? Is this typical? Must I resort to
an EQ network to solve this? Is it even possible to hazard a
guess without seeing everything? (My instinct is something to
do with the transformers, but I'm really not sure.)
2. The books and articles I've read address the directionality
of planar drivers and the necessity for having them aimed
perfectly and the like, but nothing I've found seems to talk about
the disadvantages of high frequency sound cancellation when the
wavelenghths are small relative to the horizontal dimension of
the speaker. (And vertical, but I'll assume my head doesn't
move much that way.) How do I avoid cancellation and "venetian
blind" effects at very high frequencies, or can I? It seems like
resorting to a very narrow tweeter panel is the only real
solution, but crossovers bother me.
3. What is a typical drive voltage coming off the transformers?
I can bias my panels up to about 3.2kV, and my quick measurements
using my little meter (no HV probe for the scope yet) show my
drive V peaking around maybe 500V. I've read, however, that I
should be looking at voltages practically an order of magnitude
higher. Am I just missing the peaks or? [The panels are
plenty loud.]
4. What direction would you recommend I proceed in next? I'd
love to continue tweaking the performance of these panels, but
ultimate I want as "full-range" of an ESL as I can get given
the amplifier constraints. Would it be reasonable to increase
D/S spacing by 1 or 2/32", or should I try larger panels? What
about parallelizing several panels?
Thanks for any help you might provide, and I apologize if my
questions are just "obvious" -- they're certainly not to me
(yet!).
--Ben
Sent via Deja.com http://www.deja.com/
Before you buy.
Barry Waldron
bbr...@my-deja.com wrote:
> I am starting to build some ESLs for a self-guided project class
> here at Caltech. Let me tell you where I am and then ask some
> questions:
Welcome, Ben, to the DIY ESL hobby.
> I'm using perforated sheet steel stators, 63% open area, 40" by
> 8". I'm insulating them with appliance spray paint. The spacers
> are 3/32" plexiglass, spaced in .25" on each edge. This makes
> the effective driven area approx 39.5" x 7.5". The diaphragm
> is the apparently much-revered 6 micron C-grade mylar.
> I have a stretching device that I found that is usually used for
> high tension silk screens (the kind you use to put graphics on
> t-shirts and such). It is very fast to operate and I can put
> lots of tension on the mylar. The biggest disadvantage is that
> I can't get a very good quantitative measurement of how much
> tension I'm putting, but this is offset by the fact that my
> machine is 60" x 30", so I can easily stretch, graphite, and glue
> two panels out of the same-tensioned piece of mylar.
Unless you striving to make a low resonance diaphragm,
it is not an absolute necessity to measure the tension.
Over the years people have successfully made speakers
by simply stretching the film to the brink of tearing.
Counting screw turns, etc. as the stretcher expands will
permit adequate panel to panel consistency as well as a
form of distributed resonance.
A simple method to accurately measure the tension was
described by Ronald Wagner in his DIY ESL book. He
recommends the use of four fish scales. Used in pairs,
two are used horizontally and two are used vertically.
The scales are secured near the corners of the top and
one side. The chains are layed across the jug and
attached to the opposite corners. If your jig is rectangular,
you will be able to ensure symetry and consistency.
Stretching was used regularly until ten years ago.
Most people heat shrink the diaphragm. Higer tensions
are possible and thus greater sensitivity. Type C
film is not recommended for this application.
> I used regular old graphite rubbed in and I maintain at least
> 100kOhm at one inch intervals across the surface.
It sounds like you have developed a good "touch".
> I purchased two "matching" tube output transformers off EBay.
> They have an 8kohm secondary and variable (2,4,8,16) primary
> impedances, and a nominal turns ratio of 40. They are audio
> rated 20-20kHz.
What is the make and model?
> I have been very excited with the results from my first stereo
> pair, but I have many questions.
Once bitten, you will probably not be satisfied
with conventional speakers again.
> 1. My panels seem to be up at least 10dB at high frequencies.
> (Probably ~10kHz.) I haven't made a measurement but I trust my
> ears. It feels like it rolls up between about 1k and 10k.
> Response seems reasonable flat between probably 200-300Hz and 1k.
> Obviously, it rolls off below, as with any estat. My question
> is, why the boosted highs? Is this typical?
The perceived high end rise is the result of two things:
T#1: Beaming, or the concentration of energy from the
group of frequencies whose wavelengths are shorter
than the minimum dimension of the panel/easel, and
thus do not disperse.
T#2: The electrical characteristics of the matching transformer
and its behavior when connected to a capacitive load.
The resonant frequency, reactance, and other perameters
all interact. Few audio output transformers are ideally
suited to the task.
The high end rise is typical because it is related to
wavelengh and speaker dimension. It is characteristic
of both planar and curved panels. What is unusual is
the point at which it seems to begin with your design.
Typically, the rise begins well above 8K Hz., the actual
point predicated upon the panel width and transformer
perameters. A 10dB rise is a bit excessive.
>Must I resort to an EQ network to solve this?
Yes. Every dipole, regardless of driver type(s) uses
some form of contouring to correct phase cancellation,
and in the case of ribbons and ESLs, beaming. The
usual resonse curve is flat in the center with a rising
high end and a falling low end. It is very easy to
mirror this response with a shelving equalizer. You can
find three versions of an adjustable network on my web site.
Compensation is minimal and the circuits are simple. They
do far more good than harm and are necessary if flat response
is to be achieved.
>Is it even possible to hazard a guess without seeing everything?
>(My instinct is something to do with the transformers, but I'm
>really not sure.)
Years ago, Roger Sanders evaluated several dozen output transformers.
He found that the Triad S-142a was the only suitable one. Sometime
later I found that the UTC LS-6x series was also acceptable.
"Acceptable" and "suitable" are not "best". I now have a transformer
custom made for this application. It has been engineered specifically
to meet the needs of a capacitive load.
> 2. The books and articles I've read address the directionality
> of planar drivers and the necessity for having them aimed
> perfectly and the like, but nothing I've found seems to talk about
> the disadvantages of high frequency sound cancellation when the
> wavelenghths are small relative to the horizontal dimension of
> the speaker. (And vertical, but I'll assume my head doesn't
> move much that way.) How do I avoid cancellation and "venetian
> blind" effects at very high frequencies, or can I? It seems like
> resorting to a very narrow tweeter panel is the only real
> solution, but crossovers bother me.
The Venetian Blind" effect occurs when two or mor individual
panels are placed side-by-side. The width of the dead zone
created by the width of the two adjacent frames is the cause.
The cure is to minimize the width of the dead zone(s).
Soundlab solved this problem by having the frames for their
arced arays custom cut. The width between active areas is
on the order of 3/16", if memory serves.
The solution for a planar speaker is to make one wide panel
instead of several narrow panels. A side benefit is that
this is much less work. (I know!)
> 3. What is a typical drive voltage coming off the transformers?
It is the product of the turns ratio times the voltage swing
of the amplifier. For example: A transformer with a 50:1
turns ratio will produce 4KV if the amplifier swings 80 volts,
i.e., +/- 40 volts.
> 4. What direction would you recommend I proceed in next? I'd
> love to continue tweaking the performance of these panels, but
> ultimate I want as "full-range" of an ESL as I can get given
> the amplifier constraints. Would it be reasonable to increase
> D/S spacing by 1 or 2/32", or should I try larger panels? What
> about parallelizing several panels?
True full range ESLs are impractical and very difficult to
assmble. If your desire is to have an Fs of around 30 Hz.,
then I urge you to reconsider.
* Achieving low resonances by stretching is possible
albeit tricky. Using the aforementioned jig and
scales is the best approach.
* Low frequency operation requires having a diaphragm
resistance considerably higher than most people can
achieve using graphite, if low distortion is to be
realized.
* It is much more difficult to correct for phase
cancellation the lower the ESL operates.
* To achieve low frequency operation you must trade
output. Full range ESLs do not play as loudly as
hybrid speakers.
* Bass impact and authority are not as great as with
a well designed conventional woofer system.
Diaphragm/stator (gap) spacing determines the low end cutoff
of the panel. Speakers expected to reproduce 30 Hz. should
have a gap of 1/8". The problem with this is that voltage
demands are increased. This often limits maximum SPLs, first
by reducing the field density (volts per mil of spacing),
narrowing the field of suitable amplifiers; second, by
compounding the transformer(s) interface requirements in
order to achieve high frequency response and high bass output.
Increasing the total radiating area is an easy way to increase
output. A 3dB increase can be had when the panel size is doubled.
Eight square feet is about minimum for good LF operation.
On the other hand, eight, ten, or twelve square feet is just
great for maximizing the SPLs, of any amplifier.
Electrostats designed for midrange and high frequencies need
only have a 1/16" gap and a crossover point of 400 Hz. The
gap can certainly be smaller; however, construction complexity
is radically increased. For example: the 3/32" gap used in
your speaker will support a maximum spacing of 9" for the
diaphragm support spacers. Because your panel is only 8" wide,
this is a non issue. A panel with a 1/16" gap will permit a
maximum spacer distance of about 6". Speakers built around
the Hermeyer/Sanders/Waldron architecture use this as a good
compromise for efficiency and ease of construction. Panels
with a 1/32" gap require spacers every 3" and this becomes
down right obnoxious to construct and somewhat wasteful of
the overall available radiating area. It is a compromise
few of us are willing to make simply to maximize output for
the available amplifier swing.
I do not know what direction you should take other than to
realize that there is no substitute for acreage, as a farmer
once told me. Please feel free to E-mail me and tell me what
you have in mind for a project. I'll be happy to answer any
questions and offer some guidence.
Barry
ESL Information eXchange
http://www.jps.net/eslinfo
.
Barry Waldron
Barry Waldron <e...@information4u.com> wrote:
> In article <88kfal$kd5$1...@nnrp1.deja.com>,
> bbr...@my-deja.com wrote:
(The rest of the previous post was cut off.
r_mac...@my-deja.com
Gudday Ben,
Hopefully I can help with some of your queries.
>
> I'm using perforated sheet steel stators, 63% open area, 40" by
> 8". I'm insulating them with appliance spray paint. The spacers
> are 3/32" plexiglass, spaced in .25" on each edge. This makes
> the effective driven area approx 39.5" x 7.5". The diaphragm
> is the apparently much-revered 6 micron C-grade mylar.
>
Whilst 63% O/A will work ok, you may find that you will get better
performance from a more "closed" stator. Around 40% to 50% O/A will
give greater sensitivity (depending on hole diameter and spacing) and
better damping at low frequencies.
The air gap of 2.4mm should be ok for this size panel.
> I have a stretching device that I found that is usually used for
> high tension silk screens (the kind you use to put graphics on
> t-shirts and such). It is very fast to operate and I can put
> lots of tension on the mylar. The biggest disadvantage is that
> I can't get a very good quantitative measurement of how much
> tension I'm putting, but this is offset by the fact that my
> machine is 60" x 30", so I can easily stretch, graphite, and glue
> two panels out of the same-tensioned piece of mylar.
>
You can get a durometer designed to test the hardness of soft foam
rubber. This doesn't give a definitive reading, just an arbitrary
figure on a scale 1 to 99, but is great for repeatability once you've
determined the best tension.
Try for a panel resonance of around 35 to 40Hz.
> I used regular old graphite rubbed in and I maintain at least
> 100kOhm at one inch intervals across the surface.
>
100k is too low for good low frequency reproduction due to charge
migration.
Surface resistivity is probably best at 1000 to 3000 megohms / square.
(I'm sure not everyone will agree with this, but this figure has been
determined partly by calculation and mostly empirically over the last
10 years or so.
> My power supply is built from the schematic provided by Mark
> Rehorst on his web site. It uses an Emco DC HV supply, which
> takes something like 0-20VDC and steps it up to 0-4kV. This
> makes for convenient bias voltage tweaking.
>
> I purchased two "matching" tube output transformers off EBay.
> They have an 8kohm secondary and variable (2,4,8,16) primary
> impedances, and a nominal turns ratio of 40. They are audio
> rated 20-20kHz.
>
> I have been very excited with the results from my first stereo
> pair, but I have many questions. I'll just write a bunch of them
> here and if any of you can help or even suggest a means to
> answering just one of them, I'll be delighted.
>
> 1. My panels seem to be up at least 10dB at high frequencies.
> (Probably ~10kHz.) I haven't made a measurement but I trust my
> ears. It feels like it rolls up between about 1k and 10k.
> Response seems reasonable flat between probably 200-300Hz and 1k.
> Obviously, it rolls off below, as with any estat. My question
> is, why the boosted highs? Is this typical? Must I resort to
> an EQ network to solve this? Is it even possible to hazard a
> guess without seeing everything? (My instinct is something to
> do with the transformers, but I'm really not sure.)
>
You don't mention whether the panel sits within a wider baffle to
reduce cancellation, is this the case??
There are several things you need to check for this. My first
recommendation is to feed the input with say a 1 volt signal at say
100Hz. You should get close to 50 volts on the secondary.(depends on
the efficiency of the transformer)
Increase the frequency to say 1 kHz measure the voltage on the output.
Should still be close to the original voltage measured. (if the Xformer
is flat)
Increase frequency again to say 2.5 kHz, measure secondary voltage,
again should be similar voltage.
Repeat this to 20 kHz noting voltage obtained.
Ensure the input voltage has not varied.
If you don't have a signal generator, beg, borrow or steal a test disc
with test tones on it.
There are a couple of things you need to look out for.
Standard (cheap) DVM's don't have good frequency response above about 1
kHz so if you don't have a high quality DVM use an analogue meter.
If you use the analogue meter, it probably has a dB scale. This will
make it easier to look at the response curve in the manner that is
normally displayed.
If the Xformer seems to be flat across the audio spectrum, the
subjective high frequency emphasis could be partly due to cancellation
effects causing the overall sound to appear overly bright.
This will be compounded by the charge migration mentioned earlier.
If the transformer has a rising frequency response or a peak at around
5 to 20 kHz, you can try taming this with a low value resistor in
series on the input of the transformer.
Use something like 1 or 2 ohms (5 watt) initially and check the result.
Depending on the result try a different value, higher or lower, see
what it does to the response. You may find that this quells the rise in
frequency by suppressing a reactive resonance in the transformer.
> 2. The books and articles I've read address the directionality
> of planar drivers and the necessity for having them aimed
> perfectly and the like, but nothing I've found seems to talk about
> the disadvantages of high frequency sound cancellation when the
> wavelenghths are small relative to the horizontal dimension of
> the speaker. (And vertical, but I'll assume my head doesn't
> move much that way.) How do I avoid cancellation and "venetian
> blind" effects at very high frequencies, or can I? It seems like
> resorting to a very narrow tweeter panel is the only real
> solution, but crossovers bother me.
>
As you are using a single stator which you are unable to electrically
segment, the only way to improve H/F dispersion (horizontal) is to
provide a "treble strip".
This strip is formed by applying an additional air gap spacer down the
side of the stator about 30 to 40mm from one of the edge spacers.
Whilst this section of stator is fed with full range signal the
diaphragm across this section is too narrow to reproduce the bass
signal. As long as the spacer that separates the treble strip from the
bass section is around 15 to 17mm wide, the treble strip will operate
independently of the remainder of the panel, giving a wider treble
sweet spot. This does introduce a comb effect which is noticeable
whilst walking past the speaker in the near field but is insignificant
at the listening position. A secondary effect of this technique, is to
provide a resonant peak at around 150 to 200 Hz which partly
compensates for the usual ESL suckout at this frequency.
> 3. What is a typical drive voltage coming off the transformers?
> I can bias my panels up to about 3.2kV, and my quick measurements
> using my little meter (no HV probe for the scope yet) show my
> drive V peaking around maybe 500V. I've read, however, that I
> should be looking at voltages practically an order of magnitude
> higher. Am I just missing the peaks or? [The panels are
> plenty loud.]
>
The subjective loudness here is typical of the test measurents I have
conducted with many of my panels.
Several of my designs have similar dimensions to your panels with
similar loudness results. Typically you won't need to exceed 20 volts
input for very high SPL's
> 4. What direction would you recommend I proceed in next? I'd
> love to continue tweaking the performance of these panels, but
> ultimate I want as "full-range" of an ESL as I can get given
> the amplifier constraints. Would it be reasonable to increase
> D/S spacing by 1 or 2/32", or should I try larger panels? What
> about parallelizing several panels?
>
As the electrostatic force between the 2 stators is bound by an inverse
square law, increasing the air gap, even by only small amounts, will
dramatically reduce the sensitivity of the panel.
As your panel width and air gap fall within the normal accepted ratio,
I don't think you would gain much by increasing the gap.
Larger panels or multiple panels will increase output considerably.
Try putting the panel into a baffle at least as wide again. You should
notice a big improvement in the lower midrange.
> Thanks for any help you might provide, and I apologize if my
> questions are just "obvious" -- they're certainly not to me
> (yet!).
>
Ben, I may be able to help you further with conductive coatings etc.
My Email address is rnm...@iinet.net.au
Hope this helps.
Rob Mackinlay
r_mac...@my-deja.com
Correction Here: should be volts. Sorry about that!!
Also forgot to mention that the panel should be connected to the
transformer when checking the response output.
Sorry about that too
Rob
r_mac...@my-deja.com
before I proof read!!
This should read 40 Volts.
Apologies to all.
bope...@my-deja.com
> 3. What is a typical drive voltage coming off the transformers?
> I can bias my panels up to about 3.2kV, and my quick measurements
> using my little meter (no HV probe for the scope yet) show my
> drive V peaking around maybe 500V. I've read, however, that I
> should be looking at voltages practically an order of magnitude
> higher. Am I just missing the peaks or? [The panels are
> plenty loud.]
Typical diaphragm voltage depends on spacing and insulating material
used, humidity, smoke/smog. I found it is best to make it so that you
can very it with ease. You can then customize it for each speaker and
make sure that you can get the max amount of bias voltage. Also if you
are using powder graghite only it WILL burn off over time under high
voltage.
> 4. What direction would you recommend I proceed in next? I'd
> love to continue tweaking the performance of these panels, but
> ultimate I want as "full-range" of an ESL as I can get given
> the amplifier constraints. Would it be reasonable to increase
> D/S spacing by 1 or 2/32", or should I try larger panels? What
> about parallelizing several panels?
The lower the speaker you get, the wider the gap between the front
and back grids and the diaphragm. To equilize the high freq. rise, try
splitting up the highs and lows with two seperate transformers, better
suited for the freq. that they are stepping up. Try using a toroid, of
course there is only one company that I know of that uses a Toroid
transformer in the audio side and that is Sound Lab. Just about
everyone uses them for the power supply. But Sound Lab uses one for the
high freq. call them and ask them how they do it? And why no one else
does?
www.soundlab-speakers.com
Ben (not the one asking the question different one)