[wdiabc]

This is the final circuit I have, see page 5 for details:
 

 
 
 
 
 
 
Update 2: Somehow I managed to kill two MPSA42 transistors in the circuit. One is for soft start one is burnt when running at 300V and bearing 450mw for a to92 seems to be too much. Now I replaced all four transistors with TIP500, which can stand 400V, and they dissipates about 1w without heat sink, so I will be good for a while. 
 
 
 
Updated again with 5 cell lipo supply, on man, what a difference. My old lambda is singing with 600v peak to peak voltage swing! 
 

 
 
 
Showing how this little happily drives the ECR500 now. Scope is showing 50v/div, so it looks like 250v peak to peak playing music in my normal listening level, will distort at full output. However its the mighty inefficient ecr500, for stax normal bias version lambda, it singg without any distortion.

 
 
PSU

 
 
Schematic, note Vcc can be changed from 150v up to 300v, but current is always 6ma per channel, so loading resistor needs to be adjusted accordingly. 
Input can be balanced or not, does not matter. At the output, it will always be balanced. 
One concern I have is, some people like to ground all their devices together, cause ground loop noise or whatever people may call it. In this circuit, the input is floating. 
needs 2v peak to peak to reach full output swing. 
Re=330 ohm. 

 
 
Testing with one channel: 

 
Hi all,
 
[updated with edits: I raised the power supply to +300V by doing a voltage multiplier from switch power supply and feed it with 20 volts, output swing is now 600V peak to peak, with idle power consumption near 4 watts, man what a difference! ]
 
Recently I have been trying to work on a very cheap stax amp that aims to drive electrostatic headphones. Thanks to the help of the headfiers here, especially to Frank, who offered to send me the 6pin female connectors for free.
 
So far, this thing costs nothing compared to other fancy builds:
all the parts needed for this amp, excluding enclosure is around $20. So cheap that I feel many people won't even bother do look at the measurements. However the results are really good and I really like the sound of this cheap amp. 
 
Technical Details:
Power supply: battery or 12V supply, so it can be powered by 3 cell lipo or 12v wall adaptor. ( now I am good with 19V )
Output: 600v peak to peak.
Idle Power: 300V *12ma=3.6 watts, that translates to 6 ma per channel, and 3ma per leg of each differential pair.( surprisingly to92 package with 600mw heat dissipation is more than enough, so I am not using any head sinks at all). 
Input impedance: 2k 
FR: DC to 15Khz db, rolls off to -3db at 20Khz
 
Please offer some thoughts and comments. I need some help building the enclosure, will really appreciate if anyone can provide some resources. this thing can be made quite small so mental project enclosure is currently my top choice, just want to know how you guys drill holds for these enclosures? what tools do I need? 

 

[wdiabc]

quick discussion:
 
Power Supply: I want to design the amp so imperfect power supplies can be used. We have been trying to deal with imperfect power supplies for a long time, and ideal power supplies do not exist. Some class A amps uses very large cap banks( 20,000 uf, which are very costly and honestly, may be an overkill), so I designed the circuit to be class A with differential pairs, so the power supply will only be supplying a constant DC current, filtered by a ccs. 
 
Output: Personally I did some measurements, I feel an output of 300 V peak to peak seems to drive the old STAX lambda well. However, for those new 580v bias staxs, 300v pp may not be enough. But this circuit can be modified to take 400V power supply and has the potential to swing 800v pp, which to me is more than enough to drive these staxs. 
 
Input power is only 3-4 watts, since my current setup is 150V DC loaded at 12 ma for two channels. ( 3ma for each leg of the differential pair). That is only 1.8W total power consumption, and if we use batteries, only 150ma is required, so a 2100mah li-po battery can go for more than 10 hours. I also play with a lot of RC car stuff, so the charger and battery wont be an issue for me, but those who don't have lipo chargers and batteries, these can cost $100( $20 for the battery, $80 for a decent charger) But 12v wall adaptor can be obtained on ebay for only $5 ( these HK sellers) 
 
A quick note here: eXStatA only uses 3ma each leg, if you look at the circuit. So I feel 3ma is more than enough to drive the headphone. 
 
Input Impedance: currently around 1K or 2K I guess, didnt really measure it, but now most DAC and preamps can do very well with 600 ohm load. I try to make the input impedance small so the interconnect cables' inductance and capacitance won't impact the sound that much. ( Trying to avoid those insanely expensive cables) 
 
FR is a little tricky to measure, I recently got a tek 222 scope so it will be very helpful to take some measurements.  

 

[kevin gilmore]

definitely interested in seeing the schematics.
 
I have seen decent quality lipo chargers for far less I have the
link at home somewhere. The RC batteries are definitely the
way to go, really good power density.
 
Assuming you are using a high frequency switcher to make
the high voltages, even smaller caps rated for voltage are
more than sufficient. Small series inductors help.
 
Low input impedance is not an issue these days.

 

[wdiabc]

A quick discussion about the noises and so called 'dynamic ranges' and THD:
 
Here is a measurement of fiio E10, which is a cheap DAC.
 

Fiio E10:	Sine Wave	Perfect 20-20K, rolls of a little beyond 18K
	Square	Small Ringing, but well controlled
	Clip	Clips clean, but overshoot with inductive load
	Noise	1mv(5mv little spikes) noise floor at lineout, at 200mv max output level, lineout is very clean

In comparison, I have measured a popular Class D amp, which I use to drive the energizer to power up ECR500 and stax lambda: 
 

Class D Amp:	Sine Wave	Class D, made of small steps when loaded with 8 ohm, 20-18K flat 2V PP, 18K-20K 2.5V pp output
	Square	at 1Khz, at almost full output, looks OK, but low output has 100mv HF noise and Rings badly, at 100HZ, distorts like an OTL amp
	Clip	Clips clean, but overshoot with inductive load
	Noise:	50mv noise floor, constant ****ty, at 800K-1000K, somewhat sine wave

 
So a quick conclusion:
A fiio E10 is considered to have clean sound: it has about 200 mv peak to peak line out, with a noise floor of 1mv. so output is around 200 times to the noise floor. I am bad with math so I will just use this new ratio I come up with: ( noise floor / output peak to peak)  and E10 is at 1/200
 
For the Class D amp, noise floor is at 100mv peak to peak, vs its output of 24V peak to peak. so the ratio here is 1/240, very interestingly, in this ratio, this nasty distorted class D amp wins over fiio E10.
 
For this little I designed, I measured the noise floor at around 200mv peak to peak, with output of 300 peak to peak, then this ratio is 1/1500, it is a winner, hands down. 
 
 
 
Appendix
Class D amp output wave form: made of small steps, becuase it is basically a high frequency switch, like an imperfect function generator. I don't have the square output shot for 100hz square wave, but it looks nasty: 
 

 
 
Class D noise floor at + - 50mv. 

 
 
 
 
DIY AMP: Output, 1Khz, square wave, notice how this DIY amp perfectly captured the imperfect ringing coming from fiio e10, and these little ringings are at 30khz so this amp is very well extended into high frequencies. 
 

 
 
Input from fiio e10, please notice the little ringing at 1khz:

 

[wdiabc]

Thanks for the reply. I try to keep the input impedance low because super high impedance like 1M will cause the cable to have a significant impact on the sound. 

 

[nikongod]

Cool to see new projects! 
 
Low input impedance is usually used to avoid noise. Considering that so many newer sources are able to drive it, its a shame more DIYers don't take advantage of this. 
 
Cables are usually very low DCR and fairly low capacitance. Even cheap ones from the dollar store are very low DCR. 
 
When measuring the switching/class-D amplifier did you have a load attached?

 

[kevin gilmore]

Where did you acquire the high voltage power supply board?
I want to acquire some. Then I will modify to make +/-300.
 
With mpsa92 you can go to 300v.
 
Some people have a problem with lifting the dc voltage of the stators.
A significant shock to ground from one of the output voltages can be
present. not a good thing. You can always add output capacitors.
 
Lifting the voltage this way requires an even higher bias voltage.
So for 150 vdc, you need 730 volts for the bias.
 
So a voltage trippler of the top of the secondary should work.

 

[wdiabc]

Cool to see new projects! 

Low input impedance is usually used to avoid noise. Considering that so many newer sources are able to drive it, its a shame more DIYers don't take advantage of this. 

Cables are usually very low DCR and fairly low capacitance. Even cheap ones from the dollar store are very low DCR. 

When measuring the switching/class-D amplifier did you have a load attached?

Yes it's loaded with te ecr500 transformer box, which is 8ohm. But it is somewhat resistive plus inductive load.

 

[wdiabc]

Where did you acquire the high voltage power supply board?
I want to acquire some. Then I will modify to make +/-300.

With mpsa92 you can go to 300v.

Some people have a problem with lifting the dc voltage of the stators.
A significant shock to ground from one of the output voltages can be
present. not a good thing. You can always add output capacitors.

Lifting the voltage this way requires an even higher bias voltage.
So for 150 vdc, you need 730 volts for the bias.

I got it on eBay. The 12v to 110v inverter, this cheap one spits RF out badly. And with the voltage multiplier its 270v no load and with 6 ma load it went down to 240v. I wish I could go to 400v with a few tip50 at hand. However I am also trying to get some cheapo solution for the HV supply. I need 400v at 12ma output, merely 5w though, however such things are hard to find nowadays.

PM me if you need the link to the eBay item.

 

[wdiabc]

Where did you acquire the high voltage power supply board?
I want to acquire some. Then I will modify to make +/-300.

With mpsa92 you can go to 300v.

Some people have a problem with lifting the dc voltage of the stators.
A significant shock to ground from one of the output voltages can be
present. not a good thing. You can always add output capacitors.

Lifting the voltage this way requires an even higher bias voltage.
So for 150 vdc, you need 730 volts for the bias.

Bias is not a problem. I have the 230v bias version. And it can be properly driven even if I only feed it with - 120v bias( by connecting the bias directly to the ground)

Or I can simply use another voltage multiplier with a 5M resistor to restrict the current, like the bias circuit in stax srd transformer boxes.

 

[kevin gilmore]

That's a 50 or 60 hz switcher. Not what I am looking for.
Besides which it makes nasty square waves.

Total efficiency going to be under 50%

Still you have some interesting ideas

 

[wdiabc]

That's a 50 or 60 hz switcher. Not what I am looking for.
Besides which it makes nasty square waves.

Total efficiency going to be under 50%

Still you have some interesting ideas

Surprisingly, this is a cheap inverter from HK and its a 50khz fast switching circuit. If you take a look at this circuit, 50hz switcher will need a huge transformer. The efficiency of this circuit maybe around 80-90%. The to220 transistor does not get hot at all.

 

[wakibaki]

Cool.
 
I always like to see 'dirt cheap' in a thread title.
 
w

 

[jezz]

Nice design; it reminds me of the Son of Zen. If you haven't already, you could improve distortion a bit by matching the output transistors for hFE.

 

[wdiabc]

Quote:

Nice design; it reminds me of the Son of Zen. If you haven't already, you could improve distortion a bit by matching the output transistors for hFE.

Thank you. Yes its more of a fully balanced Zen. No matching is required if you look close of the design of this amp. I just need to carefully adjust the main load resistor so the voltage drop is always half of Vcc, which will ensure maximum output level