The purpose of this article is to describe a universal amplifier capable of driving a number of electrostatic speaker ( E.S.L ) panels from various designs, without the use of coupling transformers. From the outset I would like to make clear that this module is capable of delivering Lethal- yes let me say it again Lethal, quantities ofelectrical energy if the appropriate precautions are not taken.DO NOT attempt this project if you don't have at least some elementary knowledge of electricity. Hopefully we will be able to license some kits or completed modules at least in the U.S , through the D.I.Y speaker page via Dylan Kelly.
Firstly let us take a look at some of the requirements of driving ESL loads.
- An ESL is essentially a capacitive load.
- A suitable amplifier must charge this capacitive load continually.
- The amp must be capable of delivering sufficient current to complete this charge / discharge cycle.
- For fidelity, the amp must supply a large voltage swing across the ESL load at reasonably high frequencies.
The amplifier in this design is capable of at least a 1000V swing and is able to drive loads as high as 5nF or more. The absolute maximums are a function of the final current selected and the supply voltage chosen and these in turn are limited by the heatsink / power supply and component combinations. The circuit has some flexibility in the components chosen for the final design and endeavors not to use exotic and expensive devices.
As it stands, the amp is able to drive a number of individual panels in parallel. In practice these panel loads may vary from about 75 to 300pF depending on the size and construction of the panel so that for example 10 panels of 100pF will give of load 1000pF. In a later article I will detail some of the issues of materials involved in building a number of the DIY ESL designs that have been published in the past, and there have been many. As well as detailing information on materials, such as the dielectric strength of insulating materials used as spacers in ESL panels , I will also include a comprehensive bibliography of previous ESL designs.
These brief specifications cover 3 areas.
- General voltages and current requirements of the design.
- Some calculated performance limitations.
- Some distortion measurements of a number of prototype units.
Voltage and current: power supply
Each module requires a 1000V D.C supply at 40 mA. ( With choice of output mosfet and heatsink design 1250V at 50 mA should be possible for driving larger panel arrays; and probably even higher! ). So the total current for each channel is 80 mA since module drives one side of an ESL plate i.e 2 modules per mono channel. For a stereo pair the 160 mA or up to 200 mA if the 50 mA value is chosen. The xformer will need then, to be about 150 to 200W, and will be described in detail with its schematic later in the article.
I have designed both switch mode and analogue power supplies for the modules, the switch mode design was found to be difficult due to the lack of vacuum impregnation available to the author for the transformer.The H.V xformer secondary insulation continually broke down after a shortish time ( 1 to 3 days ). The analogue design uses a commercially available xformer. Both designs will be included but with no responsibility for the SM version. The one advantage of the SM design is that all voltages including the 2KV diaphragm bias voltage are derived from the one xformer. It would be of course possible to leave out the 2KV bias voltage winding.
The maximum power bandwidth of the amplifier is limited according to the equation :
dV / dt = V / RC
where V is the mid point voltage ( see circuit description)
C is the capacitance of ESL
R is the effective resistive mosfet load
The solution has units of volts per unit time in this case V/uS and is the effective slew rate of the amp. The amp can only supply its full voltage load to about 5KHz ( depending on mosfet drain current ), i.e for a midpoint voltage of 500V a drain current of 50 mA and resistive 10K load the dV/dt is 10V/uS. This is of course not the same as the frequency response. There is not much high amplitude content in music above this frequency which is going to cause this amplifier to run into practical full power bandwidth problems.
Preliminary distortion characteristics
for 50V peak to peak into 1000pF at 10KHz = 0.017%
for 50V peak to peak into 1000pF at 20KHz = 0.017%
for 250 V peak to peak into 1000pF at 1KHz = 0.0005%
Active load version
U1 and associated components form a high impedance buffer while U2 provides either a non-inverting or inverting mode depending on position of jumpers of JP1. U3 provides two functions, the first is voltage gain to provide the correct signal levels for the final stage and second to provide an offset voltage for 1/2 Vcc operating point at the output.U4 and Q1 provide the gate drive signal for mosfet Q2, while Q3 and its components provide an active load for Q1. The output stage operates in class A and VR2 through servo loop amplifier U5, sets the gate turn on voltage for Q3. The operating voltage level for U5 is set by voltage regulator diode D1 and decoupled by C11 and C12. C6, C7, R25,R22 and C8 and C9 provide overall compensation while local feedback around U4, Q1 and Q2 is provided by R13,R14,R11 and C4.
Passive load version
A circuit description is probably not necessary here, but briefly the transformer T1 supplies the bridge rectifier Br1 to charge the capacitor chain C1-C3 and voltage divider chain R1-R3 to produce ~ 1000 to 1200V D.C according to the xformer secondary voltage. An approximate upper and lower limit to produce these voltages at up to 200mA are shown on the diagrams. R1-3 serve to equalize the voltage across each capacitor and may not be strictly necessary. If 850V A.C secondary is chosen for the H.V supply, then C1-C3 need to be increased to a rating of 450V. The primary AC supply and fuse values will change according to local conditions. The choke in version 1 will give greater smoothing but in practice the second version proved satisfactory. Notice that the choke is inserted in the earthy side of the circuit so that the windings do not see the H.V and the insulation does not risk breakdown.
This amplifier needs to be built on a circuit board and the layout negatives have been included for those who are able to make their own circuit boards. The layout should not be particularly critical, although the prototype ( not on circuit board, was rather touchy ). There are many DIY pcb kits available and these boards are quite small and although they have a ground plane and are double sided (the small number of through holes are made by components or wire pins), should be quite easy to manufacture. I hope that someone may pick up this project and supply kits at a reasonable price.
Now some details regarding components:
- All resistors are 1/4W metal film types
- Low value caps should be NPO types and the rest either ceramic or preferably polypropylene.
- On board H.V decoupling capacitor ( not shown on schematic) 0.01uF 1.5KV poypropylene or ceramic.
- There is a link needed on the solder side of the pcb between the +18V rail ( U4 pin 7 ) and the collector of Q2 ( 2N5210)-. Make sure the link does not short to other solder points!!
Start the assembly by cutting small pieces of solid wire for the through board pins ( see notes under pcb overlay ) and solder these top and bottom. Next place the smaller components diode, resistors and solder these, top and bottom where necessary. Now the capacitors, and transistors may be soldered to the board and finally the IC sockets added. Remember to add the link under the circuit board as described above. The Mosfets are placed with metal tabs facing away from the other components on the pcb, with the source pin nearest the bottom edge of the board as it appears in the overlay below.
Note that the mosfet drain is connected to the tab and hence the H.V supply, so that it must be insulated from the heatsink by both an insulating washer on the tab and a plastic washer around the hole through which the bolt securing the mosfet to the heatsink is inserted. Failure to accomplish this could result in the heatsink being live. For this reason the heatsink be earthed for safety.
The insulating washer kits are generally available from the semiconductor supplier and I prefer the SILpad type for the drain and a long nylon for the bolt insulator. The length of nylon insulator must be at least the same as the thickness of mosfet tab or longer. Alternatively a non-conductive bolt, i.e. glass filled nylon or some other plastic can be used.
Although the LM318 is a relatively old design its performance is still quite creditable, and has an excellent slew rate (National Semiconductor brand if possible). Newer designs, such as the Analog Devices SSM 2131P, and those especially with increased slew rate may be experimented with, but the compensation might change appreciably. Any 75W mosfet of 1000V and 1A drain current should work without change but if using another type select for as high a transconductance as possible at currents around 50mA, also some minor change to the compensation components might be necessary. The reason for the large number of resistors in the feedback chain is that resistors suffer from voltage coefficient and this leads to non-linearity and hence increased distortion. So the lower the voltage impressed across each the resistor the less the non-linearity. The resistors must be metal film or equivalent types, not carbon film and the same applies to the resistors that form R23. The LM310 is a low distortion high impedance type and could be substituted but the TLC271 as U5 would require careful substitution as it is used in a single supply, floating configuration.The best way of testing the unit once built, is to operate the amp with a somewhat lower high voltage rail before attempting to hook it up to 1KV supply. Perhaps 100- 300V supply with some current limiting or at least well fused and VR1 is sets output to 1/2Vcc and VR2 adjusted to turn on the gate of Q3 and bridge across 13 of the 15 resistors of R22. < /p>
- The resistors used in the position R1-3 must be tin oxide H.V types rated at least at1.5KV each.
- It may be useful to put in a slow start component in the primary of the xformer i.e a thermistor of about 1K when cold to about 10-20 R at 40 degrees C or so and 5 watt capacity. The final values will depend on the thermistor charateristics and supply voltage. The characteristics should be checked carefully against the manufacturers data.
- If there is a chance of wide local supply fluctuations then either a lower value of xformer secondary should be used or higher breakdown voltage mosfets used.
- Paper/oil types are most suitable for C1-3 and a single unit at say 8 to 16 uF / 1500V D.C are generally available but are larger and consume more chassis space. The paper types are also non-polarized.
I cannot stress enough, the care that must be taken in handling and measuring the high voltage sections of this project, namely the power supply and sections of the amp that have the H.V applied to them.
These voltages can cause electrocution. Do not touch them with the open hand, and always use probes!!!
If testing with meter or C.R.O always use a H.V probe or you may damage the instrument.
Each amp pair will provide from between 25 to 50mA into the load on each phase of the driving signal depending on the ESL load to be driven. The lower currents will enable smaller heatsinks to be used for the mosfets and it is worth noting that because of the class A operation the heat dissipation is constant. The dissipation for the four transistors per channel, remember they operate in push-pull, is 4 x 25W = 100W total @ 50 mA drain current. Some builders may want to try selecting mosfets for voltage breakdown and use 1200V dc rail or higher. In this case the dissipation will increase further to 120W or more. A suitable type with a breakdown voltage of 1500V is the Hitachi 2SK 1317 n-channel mosfet.
Finally those who wish to build the passive and simpler version should note that the drain load resistors ( 10K for 1000V rail ) will dissipate 25W each and the same for each mosfet. 50W aluminium types are satisfactory but must be mounted on a heatsink. Use non-inductive types if available and as cost guide 50W standard types cost ~$5 Aus. and 100W types which could be mounted on the chassis without heatsink ~$12 Aus. They will have however a greater inductance in standard form.Just omit components R15-23, C10-12, VR2, D1,Q3and U5 and link the source pin of Q1 to the H.V pin.The circuit is then that of the passive load version with R15 being a power resistor and mounted off the pcb ( on heatsink ).
These pages will continue to be updated with further information. Please inquire first through these pages as any comments you may have or replies I may give should go to the bass digest or any other link Dylan sees as the most suitable. Follow up projects will include a high performance no frills pre-amp to drive the ESLor other amps and a review of the materials technology and design considerations in the building of DIY ESL's and a practical easy, low cost design.
When printing the schematics of the amplifier, set the page margins of your printer to no marginor at least 0.1'' , and check with print preview if using Netscape, otherwise the images may be clipped. Finally, I have a small quantity of the 1KV power mosfets and would be willing to assist at something near cost plus postage if these devices prove difficult to procure. This might be arranged through Dylan if he is able.
The component and solder side negatives above are exact size and if printed on a good quality inkjet or laser printer may be used to effect a 1:1 negative on Scotchcal or other sensitive emulsion film for the production of a double sided circuit board. The component overlay was made larger than actual size for adequate viewing.
- All resistors that enter the board at the ground plane must be soldered top and bottom.
- All connections shown on the overlay as a dot enclosed by a circle are connections top to bottom and must be soldered both sides (see text).
|Resistors (All 250mW metal film 1%)|
|Trimmers (10 turn cermet)|
|Capacitors (Ceramic 100V NPO)|
Polypropylene 1.5KV (see text)
Polypropylene 63V MKP
|Polyester 63V MKS|
Monolithic ceramic 50V
|1000uF 16V RB||1|
|Single in line 3 way header|
|Single in line 2 way header|
8 pin D.I.L sockets ( 4)
Alternative switching H.V power supply
This schematic is an alternative to the analog power supply circuits presented in the text above. The cost and flexibilty of this circuit, along with greater availability of components over high voltage transformers and large high voltage capacitors used in the analog design may appeal.
Care should be exercised in building this high voltage circuit
- C5 needs to be 385-400V rating for 240V A.C mains operation.
- T1 is core material is Philips type 3C85 and former size ETD44.
- No of primary turns of T1 needs to be changed to 2x75 turns.
- R15 is 22K for 240V operation.
- Wire gauges for the secondary winding should be selected to accomodate 500mA.
- ESL bias winding is optional and a separate supply used instead.
- Care should be taken to adequately insulate the primary and secondary windings with Kapton tape and insulating varnish.
- L1 may be wound on any suitable ferrite core for operation for frequencies under 200KHz or so. A commercially produced inductor may be used if rated for a least 300mA and voltage.
The schematic diagragms contained in this article may be freely distributed among amateur constructors, but no commercial use of any of this material may be made without the written consent of the author.
An Electrostatic Loudspeaker Ultra low distortion pre-ampflier (coming soon)