Making an electrostatic headphone driver.

[owenhamburg]

Hi All,

Had a snooze then could not resist wiring a second transformer to each channel, and now the sound is much more stable but still not exactly sensitive. A 1:20 step up ratio is not enough for these Stax head phones and wiring a second transformer to each channel to step up the voltage has it very clear the low step up ratio was at least part of the problem the Amplifier is not happy, and getting quiet hot but the majority of the distortion is gone. I am enjoying the sound but its definitely a prototype and not a final design. I think I will try more transformers per side and make some measurements of voltage and current on the primary.

I guess if I keep up on this high voltage game I am going to need a good way to measure high voltages. I was thinking maybe a large ladder of resistors on a plane bit of breadboard, and filling my Flac player with some test tones.

Thanks again for the encouragement guys.

[owenhamburg]

Sound is now very good.

So I thought a little light in the bass when Random DJ played Promo I need not have feared I would loose bass.

Its very clear my old Meizo M3 (china ipod mini clone, more Hifi and well made but needs a screen protector.) is a better source compared to the android device I was testing with. I now consider my android device unsuited for testing. Not only can my old Meizo M3 play flac audio and the android cannot, but some of the sounds I thought where amplifier noises where caused by the android player and its poor ability to drive at high volumes. I am rather surprised how bad it is, I never tested it at high output before without sensitive headphones that where too loud for me.

Comparing the sound with AKG K500, from a zero dac head amp, the Stax is Maybe a little boosting highs and low in the bass but the sound is defiantly already better than the AKG K500's with that moving coil sound. The AKG's have reasonable new K701 ear pads that I like more than the Stax.

Things are now very listenable, but not portable, next is a test mounting, still measurements have to be planned soon so I can make some estimates before I buy a set of transformers but I think the experiments now it should be more step up than I have now, probably about the 1:50 step up ratio that was measured for Stax. I assume I would like to target 8 Ohm amplifiers after I see the results of some measurements. So next experiment with 4-6 transformers for each channel prepared (fortunately I think I have more of these transformers than I thought) and attached to terminal blocks with the correct polarity, to see what loads the amplifier best. This is too heavy bulky and impractical for a long term solution but having all these measurements will help.

I think I might go for some transformers from hereges the prices are good, and it stocks indel's transformers suggested by Bazukaz. I will have to add it to the European components for ESL building thread when I have I have had some success.

This project is beginning to smell of success

Although the experiments are not yet done, and measurements not started I have a sneaky feeling, the powers involved are not so high and the costs can be quiet low.

[geraldfryjr]

Here are a few links to a transformer test Jig that I made to give you a few ideas of how to measure such voltages safely for ESL's

A TEST JIG FOR FINDING ESL STEP-UP TRANSFORMER PARAMETERS

and the latest update,

A TEST JIG FOR FINDING ESL STEP-UP TRANSFORMER PARAMETERS

I use these same techniques in my variable HV bias supply here,

Does anyone have schematics of a varible HV power supply

I hope that these help to give you some ideas on how to accurately measure such parameters.
I just started the transformer jig thread and have not gotten into the testing procedures yet but they are coming soon.

jer

[Bazukaz]

Hi,

Some more comments about measurements. Do not use a carbon composition or a single small carbon film resistor to form the resistor divider. These are designed for limited voltages(normally 200-300V) and may cause excessive distortion if that is exceeded(especially true for carbon composition ones).
If high voltage metal film resistors are not available then a string of regular 0.25W metal film resistors can be used as well(lets say, 5-10 pieces).
If you build a voltage divider of, lets say 1000:1 with resistances of 500KOhms : 500 Ohms then input impedance of the sound card should not have a significant influence to measurements and still not load the power amp much with a turns ratio of 1:50. This would result in output impedance equivalent to 500Ohms which is pretty much a standard.

Regards,
Lukas.

[esltransformer]

Originally Posted by owenhamburg

Dear esltransformer,

That is very interesting data though I am not sure I understand it all yet. It seems you have made an impressive transformer though, and I would be interested in what you thought of off the shelf alternatives as winding a transformer seems to require tools and I already have a lot of tools when I move house next (which might be quiet soon)

The original transformer is 1:50 has a -3db point at about 10 Hz under the test graph 1. It would be interesting to know the power you where using, are situation Blue and Red with and without a 1 ohm resistor in series like graph 2? It is also interesting to see the 1 Ohm series resistor and how this improves things, but again would be interesting to see the power involved. The final graph I think explains that the capacity for bass and how the transformers can saturate and how yours passes more VA before saturation at 10Hz. Please correct me if I misunderstand as I am new to all the details.

Since your graph states clearly that the transformer is a 1:50 ratio I can now look at transformers to buy in an informed way? Do I still need the input peak current and peak voltage delivered to your headphones at a little too loud volume, or does the Bass Responce graph tell me the VA I need?

Thankyou

Owen

First i have to tell that the original Stax transformer in the SRD-7 has a 1:60 step-up. Mine has 1:50.
A esl transformer is a push-pull transformer so it has a + and - fase. In fig. 1 you see that the Stax transformer has not equal fase response for the high frequencies. This is one of the reasons i would like to have a better transformer. Secondly i would like to have a transformer with very low losses. Third i like a transformer-esl combination wich is easy to drive with any amplifier so impedance should be high even at 10Hz or 20kHz

I use for my headphone 500V polarization. I not use a 1 Ohm resistor but a 0.47 Ohm, i just use a 1 Ohm for measurements. Adding a serie resistor is not good for the low frequency response and also give more distortion.

[geraldfryjr]

Yes, Very good points.

In my test jig I only had 20 X1megohm resistors to work with or else I would have used more.
It is good until 10kv too not exceed the 1/4 watt power ratings with a maximum 5watts total and 500v across each resistor.

Radio Shack doesn't specify exactly what the voltage coefficient of their resistors are.
In extensive searches I have found ratings typically 200v to 350v and even as high as 500v for a 1/4 watt resistors on some models.
I have used some 10Megohm resistors at higher voltages before and they seem to hold up very well.
Until their protective paint coating fails and arcs through and either burns it up or opens it up like a fuse!

In my power supply circuit they max out at about 400v per resistor but this supply ended up having a larger voltage output than what I was anticipating and I rarely run it above 10Kv in normal use.
Even at 13.5Kv it doesn't drift and stays rock solid.
Maybe one day I will build a larger one with even more voltage but this one more than suits my needs.

For the transformer test jig I rated it for a maximum of 10Kv for safe use and it will rarely see any voltages above 5Kv for measurement use.
This will keep the voltages at and below 250v across each resistor for accuracy reasons.

It gets very hard to contain any voltages near 10Kv and above as it gets everywhere.
It even gets into the ground side of things, luckily I haven't had any issues of killing equipment or components yet!!!
But a good arc discharge off of my supply at 13.5Kv will sometimes shutdown my computer that is plugged into the same power strip.

Here is a link to a really nice resistor divider calculator that displays all of the factors discussed and effects of the measurement devices loading on the accuracy of the results.

Voltage Divider

Typically you want your bottom resistor as low as you can get it with out exceeding the power ratings of your selection of resistors for your divide ratio.
This will lessen the effect of measurement device loading inaccuracies.

As well as having a high enough resistance in the chain to not draw too much current dragging down the source you are trying to measure such as very low current bias supply's.

Using a buffered output is almost essential but not in all cases.
It depends on the input impedance of your measuring device.

I was finding errors of 500v to 1000v and more in some of my earlier measurements due to the loading of the meter and pulling my voltage source down due to lack of resistance in the chain.

FWIW

jer

[owenhamburg]

Thinking about what I have learnt so far.


0) The loudest anyone could want has transients that are met by 20 Vpp in the current setup, 10Vpp swings in the drive are normal. So with a 1:20 step up 200 Vpp is normal listening with these Stax Lambda Pros and 400Vpp for too loud, but maybe If we can run with a higher bias we can get away with 2 transformers per channel. Giving both a rated safety headroom against saturation of almost nothing so 3 or 4 transformers are needed per channel, 4 has the advantage in that the bias can be set to the middle if their is an advantage to this. More experiments to come, even though I doubt it makes any difference as everything is floating in relation to the bias.

1) I have no suitable resistors for measuring primary current at this moment.

2) Lukas explained how the transformer has a voltage rating, and this rating is lower than the maximum voltage the transformer should ever see from the amplifier, The reason being that saturation of the transformer magnetically if this value is passed causing a collapse in current drawn. Assuming some headroom of 50% this would suggest 4 transformers per channel. This combined with the test situations was likely the reason why one step up transformer of 22:230 was so bad sounding as I had to drive it above 22 V to sound audible and also distorted.

3) Much to my good fortune exactly the step up ratio 4 of my transformers will lead to 1:40. Lukas has been days ahead of me in realizing this already, I probably have nearly the right transformers. Only the Duel low tension winding's is a shame, single low tension would have been nicer, but maybe if the resistor is per winding problem will be minimized. At the other end of the worth checking spectrum are the 12V giving approximately 1:80 step up. I suspect 15 V rms transformers would be the sweet spot for this task giving a step up at 1:60 while still providing ample headroom for 3V peak signals with excessive volume.

4) esltransformer believes that series resistors makes the bass sound worse by adding distortion, but benefit by taming the Q of the high frequency resonance caused by the capacative and inductive components of the load, as shown by his graphs. For his environment 0.47 Ohm is the best compromise.

5) The headphones seem to change in volume dependent and frequency balance way dependent on the cables positioning, a box would make me happier and also make all the moving things around less lethal. The sensitivity of the cable is probably due to capacitance and the lead and some better mechanism or resolution is needed, maybe this is an extra benefit of a series resistances will solve this a bit/enough. The resistor is an obvious area for tweaking and reasonably safe too and could be used with each low tension winding to avoid parasitic currents getting large.

I should order at least a few smaller resistors as the smallest high power resistor in my stock is 4.7 Ohm attached to a crossover that came free with some speaker plugs on ebay, and its values is too much to be much point when I am understanding the problem domain.

6) I have 9 identical transformers to play with that are 22:230 (and 1 which is powering 2 dacs and in production, and nearly the next thing to do is organize 4 of them per channel giving 1:40 step up.

7) esltransformer chose to use a 1:50 step up ratio to make an easy load for the amplifier. Stax chose to use a 1:70 thing that would be upgraded by a pair of Torrods. 22V to 12 V gives 1:40 and 1:80 respectively.

8) I am not generating enough sensitivity with 1:20 step up, but the confusion caused by using more than 2 transformers when most have no polarization markings, is not practical as it makes too much care with oscilloscopes I need an easier way to resolve this preferably low tech.

9) esltransformer has made his own hand made transformer for his Stax and chose to make it 1:50 step up ratio. The original was 1:70 step up in a 1970's stax step up transformer model. Lukas suggests step up ratio is not something to be concerned about but saturation voltage being passed.

5) We do not know the current delivered to the cables to the headphone either peak or RMS, Playing both quiet or at dangerously loud volumes we cannot calculate the number to transformers needed per channel.

For design tools :

Originally Posted by geraldfryjr

In my test jig I only had 20 X1megohm resistors to work with or else I would have used more.
It is good until 10kv too not exceed the 1/4 watt power ratings with a maximum 5watts total and 500v across each resistor.

This seems good advice. I also like your power supply thank you for sharing with us.

[owenhamburg]

Thank you all for all your help,

I hope the 3 Stax Sockets I ordered comes soon, separating the headphones from my test set up would be nice. Anyway mounting all the transformers on a board is the next stage as it sounds OK at the moment , and I want to be able to hear it as memory is very distorting. When the mounting is finished then, working out the polarities of the transformers, wiring up the terminals,

The plan is now to buy some series resistors, both for measuring current, it would be nice to know if I need more or less than 15 or 30 VA or if the costs will mount with larger transformers.

I also need to search for some solution to free up as many of the identical transformers as possible, a good search in my box of old ac adapters is needed as I will be buying more transformers rather than less in the future so I will hold back for now.

[Bazukaz]

Originally Posted by owenhamburg

Thinking about what I have learnt so far.
2) Lukas explained how the transformer has a voltage rating, and this rating is lower than the maximum voltage the transformer should ever see from the amplifier, The reason being that saturation of the transformer magnetically if this value is passed causing a collapse in current drawn. Assuming some headroom of 50% this would suggest 4 transformers per channel. This combined with the test situations was likely the reason why one step up transformer of 22:230 was so bad sounding as I had to drive it above 22 V to sound audible and also distorted.

Core saturation happens when almost all microscopic magnetic zones called "Domains" become oriented in a specific direction by magnetic field. No amount of current can increase magnetization by much. The current draw tends to increase very rapidly and not "collapse" when this point is reached with increasing voltage as long as the source can supply it.
Max. voltage/frequency relationship due to core saturation of the transformer is more or less linear. So transformers you have should be able to handle at least 22Vrms@50Hz, or 11V@25Hz, or 44V@100Hz. Giving that most of music content does not have much of energy at frequencies below about 40Hz, IMO 22:230V transformer should be pretty much suitable for full range operation with an amplifier producing ~20Vrms.
However, as I already mentioned, even small amounts of DC current from the amp can partially magnetize the core so much less is left for audio signal.
So stay away from custom-tuned audiophile's amplifiers from which all of the DC blocking capacitors are removed from signal path
Another simple test for DC current is placing a large electrolytic cap(lets say, few thousands of uF @50V or so) in series with a transformer. It will completely block DC(not like resistor) , so if there is a sudden and obvious improvement you will know the problem even without an ampere meter(not all of them are suitable to measure DC in this situation; as they have a built in resistor which can alter the current a lot by itself).

Originally Posted by owenhamburg

4) esltransformer believes that series resistors makes the bass sound worse by adding distortion, but benefit by taming the Q of the high frequency resonance caused by the capacative and inductive components of the load, as shown by his graphs. For his environment 0.47 Ohm is the best compromise.

The magnetization curve of the transformer core is non-linear. This effect is much more prominent in low frequencies where core magnetization is high. The low impedance(mainly caused by feedback) of a typical discrete amplifier compensates for this : despite current draw varies, the voltage on the transformer secondary follows the voltage at amplifier's output(remember transformer's formula : U1/U2 = N1/N2).
Therefore I view tube amplifiers a bad choice for driving an ESL; their relatively high output impedance causes increased distortion from step-up transformer, in addition to attenuating high frequencies.

Originally Posted by owenhamburg

8) I am not generating enough sensitivity with 1:20 step up, but the confusion caused by using more than 2 transformers when most have no polarization markings, is not practical as it makes too much care with oscilloscopes I need an easier way to resolve this preferably low tech.

You don't need a scope to determine polarity of transformers. Primary goes in parallel(this case from amplifier) and secondary in series. After connecting, run some steady state signal through the trafos and measure voltage across secondaries via voltmeter. If you wire them in wrong polarity, voltage is subtracted otherwise summed. Just be careful, you may be having hundreds of volts there.
After having been hit by a small ESL panel a long time ago I started to greatly respect step-up transformers and their capabilities

[owenhamburg]

Dear all,

I put the Bias back up to 230 X 3 volts from 230 X 2 volts and with this Bias I can half the stators drive voltage down to 2.5 Volts so so making ther output voltage with around 50V. Anouther day of listening plus measurements should decide this issue but I hope this will make 2 transformers per channel fine for Hifi. Mozart's Symphony No. 38 is now sounding like my Quad speakers, we definately have a big improvement with Higher bias, I estimate with a target bias of 580 V a voltage trippler is about 20% over specified bias rather the voltage doubler which is 20% below bias.