Acoustat Full Range Electrostatic Speakers
A group of us have owned Acoustat ESLs since the early 80’s. We believe them to be vastly under-rated speakers that can compete on favourable terms with today’s finest. In common with many ESLs competing in the reality of the marketplace compromises were made, however the parts that matter most were of very high quality namely the transformers and panels themselves.
My panels started off as Model 2s but after a few years, these were dismantled and new frames (very crude) were made in 1+1 configuration. Richard started of with Model 4s and then Model 3s. He then found two more panels and re-framed them as 2+2s. Others in our group have 1+1s and 2+2s most having started off in some other form.
For a description of what Acoustat electrostatics are all about, see James Strickland paper on The Technology of Full Range Element Electrostatic Loudspeakers.
Frame
We believe the frame compromises a lot of commercial ESLs. For some reason, manufacturers build sleek stylish frames that are very flimsy. This must be for marketing/aesthetic reasons because most do not offer the rigidity required to do the job properly. An ESL may not have a large excursion but a full range ESL does have a large area and couples very well to the air in the room. It is our belief that the frame of a full range ESL must be rigid to resist movement.
The frames I’ve been using have never been totally successful but have been in use for 15 years. Time now to build some proper ones.
We have long thought that the vast majority of ESLs fail to reach their potential due to the limitations inherent in the frame design. This is especially true when the ESL is full range and is required to deliver substantial bass information.
What has concerned us is it seems the frame is usually only given a last thought and in most cases does not support the panel adequately. But what is adequate?
Adequate means to me supported evenly and rigidly over the whole length of the speaker panel without twist or bending in any plane. In a full range ESL, that means the panel is usually large. In the case of Acoustats that panel length is about 8 ft. As there is little or no information on this subject with respect to ESLs, I chose to look at how conventional speakers address this problem of rigidity and then try to apply it to an ESL.
Design aims:
- Rigid and Square – to be maintained over time
- Mass loading option
- Bolt to frame or clamp to frame option
- ability to slide around and spike (not at the same time)
- bolted/screwed together to keep design flexible and to ensure squareness when assembling
- able to experiment with compounds twixt panel and frame
- able to experiment with edge treatment/wings etc
- minimal stuff at rear and nothing (or little) in the way of the radiating surfaces reflecting energy back at the panel
A First Go
So what do they sound like? First I have to say I am even more impressed by Jim Strickland’s panel design. I believe the plastic grid construction making the stators very rigid contributes to their quality. Why anyone would go to all the trouble of building an ESL (unless that’s your thing of course) when a pair of 2’s or 1+1s (or 2+2s) could be picked up for peanuts relative to today’s prices. They have to be the best ESL speaker materials kit available.
After full view of speaker, rear view of top half, rear spike detail, attachment of copper to wood frame and joint detail.
For a summary of the 1 + 1s, check Hot Rod 1+1s below
The thing that finally got me to replace the old frames was hearing Jack Bouska’s awesome system. The broadband power response that this system has is illuminating and thought provoking but my Acoustats are still the ones for me.
Update – April 9th 2002
Rocket Lauchers
The frames are now nearly a year old and have performed superbly. One day, I hung half a bag of sand over a rear stay to gauge tilting effects. I had assumed I would hear a tighter bass but no – the reduction in HF distortion was very noticeable as was the increase in focus. Bass was hardly altered. This indicated that maybe the frames were “buzzing” on the floor? The design always had the option to add mass due to it’s tubular construction but I didn’t want to add mass to the tube frame itself.
So back to the plumber supplier for some big tube – 4″ / 100mm plastic waste tube seemed ideal so two 6’6″ / 2m lengths were capped off (screw type), sand filled and clipped to the rear frame tube. Each tube weighs about 56lb / 25kg.
The previous cleaning up of “noise” now extended down into the lowest bass. Up to that point, I was not aware just how noisy the frames had been and these way quieter previously than any other ESL frame I’ve heard.
I removed the wooden frame completely and using the pipe clips, coupled the panels to the tube frame directly. This confirmed a number of things; the panels sound like the material they are attached to and that they need to be attached to something massive and well damped. The non-wooden frame speaker was making itself heard in comparison with more of the music camped on the speaker. Room for further development.
After full view of speaker, side view (looks like a Rocket Launcher to me) and detail of attachment to frame
Update – October 14th 2002
After spending some time with the heavier frames, I find the life has gone from the music. Sure I thought things had been “cleaned up” but it would seem “cleaned up” is not how I like things. The mass in the frame seems to have caused a storage and release mechanism of the frame’s energy thereby messing with the musical message. This is not entirely obvious on first listen but became apparent over time. The vast open window that I enjoyed so much with the light weight frames disappeared, sanitised somewhat and so boring.
As an aside to the above, I’m revisiting some of the component recommendations – specifically the use of replacement resistors instead of the HF control, the HF capacitors (LV side), multiplier parts and a few others.
Update – December 10th 2002
Some of my assumptions in making the frames above – the requirement for static stiffness and mass to counter the energy in the frame may well have been misplaced. Experimenting with the frames and reading more about vibration measurement and control led me to think a light and rigid approach may well be better if applied to the right area.
Adequate support for static loads is way different to providing support at audio frequencies. If the main mode of resonance is below that being produced by the speaker, then it will stay still. The frames may not be as statically stiff as the old ones but they are compared to most panels so seem to provide good support as the bass is very clean and goes deep while allowing other instruments to stand alone and not be swamped. This aspect alone leads me to think this approach may yield better long term results.
Accidently, the wooden frame is relatively light and quite rigid. The tripod support does not now provide stiffness; only support for the frame. In the future I may be able to make the frame stiffer and lighter as the two requirements of support and holding the panels are seperate.
Although it’s early days yet, this is by far the best frame design I’ve had. The mass approach has benefits but this method is far more musical. I’m sure there are improvements I could make but it will be a while yet before I can get a fix on what. What is a sure thing – the panels sound more like the material they are suported by than anything else and so it will be this aspect that deserves development in the future.
Inspiration for further development may be based on the parallels in TT design. Both the speaker and cartridge are tranducers that convert movement to electrical energy or vica versa. The panel is like the cartridge, the headshell like the frame (support for the transducer), the tripod frame like the arm tube (lightweight, stiff support) and the TT base like some form of mass sink (currently the floor but could be added to) for vibration. How all this translates into a design will be the fun part.
From left – left speaker, right speaker, rear view bottom joint, rear view top joint and lastly, very important rear brace
Interface
The modifications described have been in use for approximately 15-20 years in a number of systems. We can therefore vouch for the reliability of the components we have used however due the the nature of the voltages in the interface we cannot comment on the use of other types. We welcome any input from the experience of others.
Interface Circuit
The interface contributes massively to the sonics of an Acoustat ESL – obvious I suppose. As the elements and transformers are excellent quality, we have concentrated our efforts on improving the interface components and its “interface” to the speaker elements, amplifier and wall socket. Although our mods haven’t changed the electrical characteristics of the interface, they now seem to be easier to drive i.e. they seem to go louder for a given input. However even though they are easier to drive, they still represent a difficult load – see graph below.
Audio Section
Replacement Wiring
An early mod was to upgrade the internal wiring to Monster Cable. However this cable suffers from internal corrosion (green discoloration) so should be replaced. I use Kimber Cable but any cable deemed to be ‘better’ can be used. We also bypass the phenolic tagstrip and push clip used for bass boost (transformer tapping dependant on the number of panels) by joining the two wires together. The tagstrip holding the 1 ohm R4 was moved to a better position to shorten wiring to the bass transformer – see pic below. The hole and bolt for mounting the tagstrip was previously used for p-clips holding the monster cable.
The input terminals of the interface have come in many guises. We prefer any heavy duty binding post as the originals were very poor.
There is also an input fuse. We took this out of circuit many years ago – do so at your own risk noting failure modes of the power amp and bearing in mind the transformers are no longer available! If the amplifier is solid state, we would not do this.
High Frequency Capacitor – LT side
The HF capacitor (C1, C3, C3) may have already been upgraded as part of an earlier mod. The value of this capacitor will depend on whether the interface has the medallion modification.
This capacitor will determine the voice of the speaker and is really system dependant. I have tried a number of capacitors in this position and found they all give a different presentation. Solen, North and Ansar all make metallised film capacitors that are suitable. To my ears, they all exhibit the “metallised” sound which is not my preference – the Norths the least of the group by far. I believe paralleled capacitors in some systems will provide the “zing” to get the system to sound alive but ultimately I find this to be tiring long term and compromises tonal purity. Some of us have used Black Gate types in this position. I did use them for a bit but lately haven’t tried them again in revisiting all the recommendations on this site. So the North Zen that doesn’t fit very well is in there at the moment. Check out the size difference below between the Black Gate option and metallised PP film. Some time soon (early 2003) I hope to get back to the BG and see how he goes.
High Frequency Capacitors – HT side
These looked like paper caps in the original (they’re covered in paper but may be mylar type) but as I don’t have them to hand I can’t recall accurately. Later these were yellow MKP types. We replaced these (C4 and C5) with 0.01uF WIMA FKP1s rated at 2000V DC. We measured the AC voltage when playing the speaker loud and although these are of lower rating than the originals or medallion replacements, they seem to survive. WIMA do have a 6000V rated 0.01uF FKP1 and they are probably a better bet. Very powerful solid state amplifiers can cause the 2000V capacitors to fail but a series parallel combo seems to fix this. I now use 0.01uF 5kV rated film/foil PP capactors from SEC.
Low Frequency Resistors
The original resistors (R2 and R3) were ceramic cored wirewounds. We originally replaced these with series parallel combination of Resista MK4s but they couldn’t handle the power and died. We then tried Dale 47K 50W resistors (the ones that look like a gold heatsink) and they are what we used until recently. Further experiments have shown that Caddock power film resistors to be quite grainy in comparison. We are now using 2 off 25K 10W Ohmite resistors of the same type as the Norths.
We have experimented with removing the 1 ohm resistor in series with the bass transformer (R4). In the past we have thought removing it to be of benefit but lately we have realised this is a mistake. It seems to roll off the bass transformer. Without R4 the treble is splashy and unrefined and the midrange congested. This can make the speaker sound more exciting but not accurate. The quality of this resistor in very important and the winner in this position at the moment are from North Creek (not in business anymore).
Removing R4 does not improve bass performance. From Andy Szabo previously of Acoustat …. “The resistor (R4, 1 ohm) is there for a very good reason, and it is not for reducing the output of the low frequency transformer. A transformer has very low DC resistance, in this case only a fraction of an ohm. Some amplifiers are unstable into such a load, so the 1-ohm resistor was added to guarantee that the DC resistance of the speaker was never less than 1 ohm. Additionally, if your amplifier has any DC offset on it’s output, then this very low resistance can cause significant current flow, possibly causing instability and/or overheating.”
High Frequency Resistor
This was originally a rheostat (R1) used to adjust the HF balance. The size of these resistors will depend on whether the interface has the Medallion modification. For non-Medallion interfaces I used 2 ohm and 4 ohm (2 ohm to speaker terminal) but have used 1.5 and 4.5 ohms in the past. For Medallion interfaces we used 1 and 15. We favour the North Creek resistors in this position too. We have used Caddock MP type power film resistors but they are grainy and closed in compared to the Norths. The down side is losing the flexibility of the HF control which can be a problem if the system or room changes. I’m revisiting these recommendations so living with the rheostat for a while and then swapping back again – this being early 2003.
Please note: We have chosen the 10W North/Ohmites for their sonics first. It is quite possible with very high horsepower solid state amplifiers that the resistors we’ve chosen may get hot. Although we listen loud at times, others may listen louder in bigger rooms.
Power Supply Section
Teflon Board
We used Teflon 2mm thick the same size as the original PCB. The capacitors are mounted on top and the circuit connected together using the diodes on the bottom. This was one of the first modifications we carried out so used all the original components. Highly recommended!
Resistor
I have been using the original 500M resistor (R5) feeding the diaphragm. Now experimenting with the HV thick film resistors from RS, 500M 7.5kV, type 296-0645. The Caddock MX450 (used for some time by others) seems to be as most Caddocks, grainy and closed in compared to the original. If there isn’t a problem, then the original is hard to beat.
Mains lead
We have experimented with mains lead feeding the interface and find there are benefits. I have used reconstituted mains lead (solid core twin and earth) wrapped around “XLO style” a PTFE tube former. Richard and a few others use RG214 RF cable, Andy uses XLO mains cable on his Quad ESL63s.
Conclusions
Most of the above modifications refine the sound and generally increase the speakers resolution, bandwidth, dynamic range and ultimate loudness – in other words the voice of the speaker changes dramatically, we believe in the right direction. Some of the modifications on their own may reveal an unpleasant result but that usually means their has been another problem area exposed OR you are just hearing more of the truth upstream.
Sourcing Parts
The Ohmite resistors can be hard to get outside of the values North Creek supply. Use the Ohmite Website Worldwide Inventory Search Tool to locate stock. Note: Check Ohmite values for resistors as they use non-standard values.
Further Tweaks
Interface Housing
The interface is very susceptable to vibration which is exacerbated by the fact the Model series use it as a major frame support component. In our early frames, we used the interface to weigh down the speaker. When we isolated the interface from the vibration, we experienced a less confused sound. Moving the interface out of the rear sound wave improved things further. Mounting it on damping material and a single hard point such as a cone improves things again.
EHT Modification
In messing about with mains regeneration, it occured to us that the mains and/or the multiplier was messing with the sound. The multiplier has no effective smoothing so we added capacitance to the output to smooth things out a bit. The effect was far greater than we could have imagined. First up we used the spare treble HT coupling caps of 0.01uF as we had them lying around having replaced them. Later we tried various sizes of PP film/foil SEC capacitors; the same as we used for the audio part. Currently I’m still using 0.033uF rated 5kV having tried 0.01uF of the same type. Not sure if there is any difference; see pic below for placement. It changes the speaker dramatically removing a greyness to the sound, opening it up dynamically seeming to allow them to play louder and more extended at both ends. Quite remarkable!
Note that this added capacitor will store energy after the mains has been switched off and could prove to be very dangerous. Only do this mod if you are absolutely sure you know what to do as a mistake could be fatal!
Revised R4 position, interface box, Black Gate size comparison, teflon cct board and carpet eye view of EHT mod capacitor.
Hot Rod 1+1s
This is the current state of my 1+1s, the parts used and response
I thought Hot Rod describes my speakers configuration fairly accurately. The idea is to reduce the speaker to its rawist state so that nothing exists other than to perform the task intended in its purist form and executed as best as I am able or can afford.
Audio Interface
| Circuit Identifier | Component | Manufacturer |
|---|---|---|
| R1 | 6R 50W rheostat | Ohmite or original |
| R2, 3 | 4 off 25k ohm 10W 1% WW | Ohmite 80 Series |
| R4 | 1 ohm 10W 1% WW These can get hot when playing very loud. | North |
| C1, C2, C3 | 200uF Zen 100V | North |
| C4, C5 | 0.01uF PP film/foil, 5000V | SEC |
Power Supply
| Circuit Identifier | Component | Manufacturer |
|---|---|---|
| C6 – C10 | 3300pF 2000V | BC Components, 356-3406 Farnell |
| D1 – D5 | originals | Acoustat |
| R5 | original | Acoustat |
Measurements
I was lucky to have a friend over that likes to measure things. This is what Jack Bouska measured. Click diagrams to enlarge.
Current room is 14’9″L x 11’8″W x8’7″H. There are 6 DIY “CornerTune” type cushions in the various corners and 4 DIY “RoomTune” type devices behind/between the speakers to control the reverb field. As there is no furniture, there are two single bed mattresses at the first reflection points to break up slap echo and a duvet over the window to reduce it’s effects (bass leakage and high end emphasis/blurring). The speakers are about 4′ from the rear wall and 2’6″ from the side walls. The listening chair is about 3′ in front of the rear wall with damping behind the head to stop rear wall reflections hitting the ear. Overall the room is live without being overly damped. We moved things around a bit but it didn’t make things better (the measurements) from that position determined by ear.
The low frequencies suffer from room modes and overly long reverb time. From about 100Hz up the RT is about 0.3 to 0.4s – good for a listening room however below that the RT gets up to about 0.6s or more and so needs treating. A major room mode at 38Hz needs taming. Having only been in this room for a week before the measurements and with no bass treatment, it’s not too surprising to see the results as they are. Very pleasing is the lack of broad energy storage in the waterfall plots and the superb impulse response.
Of interest is the room mode analysis and it’s match with bumpy bass response. I will be trying to remedy this with various acoustic devices.
Frequency Response, 2mS Impulse, Waterfall – Low Frequency note: time axis = 300mS. Level 50 to 100dB, Waterfall – Low Frequency note: time axis = 300mS. Level 50 to 100dB, Waterfall – High Frequency note: time axis = 1.5mS Level 70 to 100dB and Room Modal Analysis
A Warning
Acoustat ESLs were no doubt designed to sound they way they do for a reason. We have taken that design and made it sound different and in our opinion, our designs/upgrades improve the speaker and bring the design up to date. In the process of doing that, the speaker can become a microscope and may highlight a problem elsewhere which the speaker could be blamed for. Many a time, we have thought about selling them on and getting something else easier to live with but each time, we find the speaker to be telling the truth.
If at some point in the future the speaker may be sold, ensure any modification can be removed and the speaker returned to standard. Check and double-check any wiring modifications as the transformers are virtually irreplaceable.