distortion From ESL at low level

[geraldfryjr]

Hey there, I haven't been on in a while and I am glad to see that you have finally been able to gets some parameter measurement on your transformer.

If you have a signal generator you should try to do the resonance test's as well this will tell you the transformers capacitance and leakage inductance.

With that data and the ones you now have you can easily model your transformer in a Spice program.

I did this with my data and started working on a passive crossover as well.

I had a few versions that seemed to turn out okay but they would have been rather costly implement due to the several very large capacitors that were required, However it is a place to start.

The transformers saturation curve follows that of a second order filter and in order to keep the core from saturating below the crossover frequency at least a 12db per octave filter must be used.

I haven't read through all of the recent posts yet so sorry if any of this info has been repeated.

As much as a passive design seems simple, it is not!!
It can cost much more than a simple active design would cost, but, it is still something that is good to know as it helps you to understand the complex impedance that these things create.

Cheers!!!

jer

 

[WrineX]

Had you measured the impedance with this setup? I would think impedance will drop to a very low value(not much above the series resistance) around 1kHz.

Looking at your previous impedance measurements again, I noticed that the magnitude of your impedance peak is not nearly as high as I am used to seeing. Normally I will see Zmax of 100 to 500 ohm. Yours was < 20ohm. With no load connected, the impedance peak is due to parallel resonance between primary inductance and winding capacitance. The magnitude of the impedance peak is an indicator of core(and other) losses. All other things being equal, a better core will result in larger impedance peak value. If there were no losses, the magnitude would trend toward infinite. Your low value of Zmax may be why esltransformer suggested there may be something wrong with the transformer.

The low value of Zmax mentioned above indicates there are higher than average losses in the transformer that are also damping your 2nd order crossover producing the smaller response hump.(most likely also part of why you were getting distortion with only a series capacitor)


One measurement without would tell us the primary inductance.
If you take another measurement with the panel connected and know what the panel capacitance is, we could also estimate the leakage inductance.

I could provide some suggestions for winding your own transformer with the wire you mentioned.
I would need to know:
1) core type and dimensions
2) bobbin dimensions, and number of bobbins
3) ESL capacitance to be driven
4) Desired crossover frequency

ok ill give it a shot

i will look into your wish list hope i can gather all info next week, its been bussy around here Thx so far everyone for the reply's . looking forward to wind some simple tannies that can be crossed at around 300-500 hz i noticed the antect could be used as well but it should be possible to wind my own for less money ?

 

[WrineX]

ok no measurements made, but i do must say it is finally working, i still had this nasty hump at the crossover fequency of the second order filter, i now added a 1.2 oHm resistor in paralel with the caps, the funny thing is it does not influance the high end response like the resistor directly on the trannie does. if i increase say to 15 ohm i get the hump back more or less. with the resistor the rolloff looks also better, not as steep, its only a 12 dB octave filter but when undamped it looked more like 18-24. so finally this thread is done. and i know now that i still can use cheapy trannies with the same problem for tweeters without to many problems.

thank you all for the input !!! love the forum and love the amount of knowledge floating around here!

 

[geraldfryjr]

Yes, The saturation curve of a transformer core requires at least a 12db per octave filter as they are exactly the same slope.

I am glad that you finally got it working now!!!

jer

 

[geraldfryjr]

A 6db per octave filter will allow the core to saturate at the lowest frequency's if its cutoff frequency is low enough.
Where as a 12db filter won't for the very same frequency as in the example above.

The resistor will only influence the high end if it is only in series with the primary winding.

jer

 

[geraldfryjr]

Here is the section I meant to post earlier if I haven't already.
It is about taking the parameters you have found on your transformer and modeling them in a Spice Program.

I used this format to start work on a passive crossover, but I haven't found those files to show yet.

Here is the thread,

http://www.diyaudio.com/forums/plan...truct-cube-louver-acoustat-3.html#post2177098

jer

 

[bolserst]

ok no measurements made, but i do must say it is finally working, i still had this nasty hump at the crossover fequency of the second order filter, i now added a 1.2 oHm resistor in paralel with the caps, the funny thing is it does not influance the high end response like the resistor directly on the trannie does.

Congrats on getting your 2nd order network response damped to your liking.
The reason that the resistance in parallel with your crossover capacitor does not influence the high end response is that the capacitor effectively bypasses the resistor at high frequencies. Basically, the capacitor looks like a short circuit to high frequencies so they take the path of least resistance through the capacitor rather than through the resistor on their way to the transformer.

A 6db per octave filter will allow the core to saturate at the lowest frequency's if its cutoff frequency is low enough. Where as a 12db filter won't for the very same frequency as in the example above.

Just to be clear, flux density in a transformer core is proportional to applied voltage and inversely proportional to frequency. Suppose your transformer doesn't saturate with your maximum input voltage at say 500Hz. A 500Hz, 6dB/oct HP filter will roll off the response so that the voltage reaching the transformer is halved each time the frequency is halved. So, the flux density in the transformer core would remain the same resulting in no core saturation problems.
More details here: http://www.diyaudio.com/forums/planars-exotics/161485-step-up-transformer-design-4.html#post2096503

That being said, you can't simply put a capacitor in series with the transformer and achieve a 6dB/oct filter response. The desired 6dB/oct response is only achieved if the load has a constant magnitude and low/constant phase angle...which a transformer definitely does not. One option is to swamp the transformer impedance changes.
See example of Acoustat 6dB/oct HP filter here: http://www.diyaudio.com/forums/plan...tin-logan-cls-power-supply-3.html#post2882393).
However, with transformers having low primary inductance this may require very low value high wattage resistors.

The next simplest passive crossover option is using a damped 2nd order filter(12dB/oct) as Wrinex has shown. One interesting issue with the 2nd order filter is that if they are not properly damped they can sometimes cause the core to saturate at a frequency where there was no saturation problems without the filter. Basically, a 6dB peak in the filter response doubles the voltage applied to the transformer at that frequency and thus doubles the flux density. For example, applying 30Vrms to the input of the under-damped 2nd order filter can result in 60Vrms(or more) applied to the transformer primary. It's not "free" voltage gain though, it comes at the cost of double the amplifier current. As mentioned previously, this peak can be useful for equalizing the inherent ESL dipole roll-off of. You just need to make sure that the crossover frequency is set high enough to handle the increased applied voltage at the frequency of the filter peak.

 

[bolserst]

i still had this nasty hump at the crossover fequency of the second order filter, i now added a 1.2 oHm resistor in paralel with the caps...

I got to thinking about your method of damping the filter today with a resistor in parallel with the capacitor. I have drawn what I think is your circuit in Figure (A). Adjusting the value of Rx would change the crossover damping; lower values give better damping. The series resistor Rhf(if included) could be used to adjust the high frequency response.

If I understand your posts correctly, your crossover parts are:
Lx = 0.25mH
Cx = 69uF
Rx = 1.2ohm

Have you measured the impedance?
With these part values, you probably have a very low impedance below the crossover frequency as shown by the Red Curve in the plot.

You should be able to achieve nearly identical response with circuit (B).
But, with no low value resistor bypassing the capacitor the impedance will be a much easier load for the amplifier(Blue Curve). An Rx value between 1 and 4 ohm should give you a nice range of damping options; higher values increase damping. As with circuit (A), the series resistor Rhf can be included if you need to adjust the high frequency response.

Note: The curves in the plot are for illustrative purposes, not an exact modeling of your crossover/transformer.

 

[bolserst]

hmmmmm....
I wonder if your distortion problem will come back if circuit (B) is used with larger values of Rx.

 

[esltransformer]

What about the value of L from the step-up transformer?

Specially for toroids the value change is huge.

I got to thinking about your method of damping the filter today with a resistor in parallel with the capacitor. I have drawn what I think is your circuit in Figure (A). Adjusting the value of Rx would change the crossover damping; lower values give better damping. The series resistor Rhf(if included) could be used to adjust the high frequency response.

If I understand your posts correctly, your crossover parts are:
Lx = 0.25mH
Cx = 69uF
Rx = 1.2ohm

Have you measured the impedance?
With these part values, you probably have a very low impedance below the crossover frequency as shown by the Red Curve in the plot.

You should be able to achieve nearly identical response with circuit (B).
But, with no low value resistor bypassing the capacitor the impedance will be a much easier load for the amplifier(Blue Curve). An Rx value between 1 and 4 ohm should give you a nice range of damping options; higher values increase damping. As with circuit (A), the series resistor Rhf can be included if you need to adjust the high frequency response.

Note: The curves in the plot are for illustrative purposes, not an exact modeling of your crossover/transformer.

 

[invisible force]

problem with transformers regarding distortion is, that the main inductance is strongly dependent on signal level. This is going like this : starting from a low value (say 50mH for a bass transformer, this you measure with a L meter) inductance wil rise with rising voltage at the primary winding up to the point when core is starting to saturate (say 500mH).
This gives the current a distored part !
If the source (amplifier) and the primary winding had zero Ohms DC resistance this distored current would not (!!!) show up in distored voltage, but all the linear impedances in series to the transformer (cables, dc resistance of primary winding, amp output resistance, crossover network (!), capacitor) create a nonlinear voltage drop because of this non linear current flow through them.
This voltage drop is subtracted from the hopefully distortion free amplifier output and will result in a distored input voltage for the (inner) transformer inductance (measuring at primary winding with a thd meter will not measure distored voltage drop of its dc resistance).
so keep main inductance as high as possible from the start (lowers the distored current amplitude) and/or keep serial elements small lowers distortion.
Hope this was helpful...

 

[bentoronto]

But is it right to introduce a permanent "distortion" (a high-pass filter) to pre-empt a potential distortion (core saturation)? Is there a logical error here?

If something disastrous happened when the core got saturated, you might introduce some kind of limiting. Unbeknownst to many golden-eared audiophiles, there are brief peaks in live music (esp. womens choirs due to the stats of music) that will inevitably be off the scale of all HiFi systems. Likewise, occasional instances of saturation may be of small importance to musical enjoyment, even if saturation makes nasty-sounding sorts of distortion.

While the freq math is unassailable, the reality of anticipating when the core will be saturated by the confluence of loudness and frequency is pretty hard to guess and even depends a bit on the kind of music being played.

Ben

 

[bolserst]

What about the value of L from the step-up transformer?
Specially for toroids the value change is huge.

To avoid the large changes in the transformer primary inductance from compromising the crossover response you need to swamp it(make it unimportant) by placing a low impedance in parallel with it. For a 1st order crossover this is done by placing a low value resistor across the primary as in the Acoustat circuit I posted. For a 2nd order crossover this is done with a low value inductor across the primary; usually adding some resistance in series with the inductor is desireable for proper damping of the filter response.

Note the very low inductance value (0.25mH) Wrinex is putting in parallel with the primary. Even if the transformer primary inductance varies wildly on its own, once the inductor is place in parallel with it, the impedance across the primary will not stray far from that of a 0.25mH inductance.

problem with transformers regarding distortion is, that the main inductance is strongly dependent on signal level...so keep main inductance as high as possible from the start (lowers the distored current amplitude) and/or keep serial elements small lowers distortion.

This is absolutely true and was what I pointed out early in this thread, although perhaps not as clearly as you have.
http://www.diyaudio.com/forums/planars-exotics/226637-distortion-esl-low-level-4.html#post3317465

In this case, the OP is wanting to use a passive HP filter with an existing transformer he already owns.
Its primary impedance is unusually low, <1mH, so the only options to keep distortion low are:
1) drive it directly from a low impedance amplifier as you suggested, and use an active crossover
2) place a low impedance across the primary to swamp its effect as part of a passive crossover

 

[bolserst]

But is it right to introduce a permanent "distortion" (a high-pass filter) to pre-empt a potential distortion (core saturation)? Is there a logical error here?

I'm not sure I follow. The OP wants to HP filter his ESL panel so he can cross it over to a dynamic driver for the bass. Whether or not you consider a crossover between two separate transducers as a permanent "distortion", you would want to do it without adding other known distortion artifacts to the ESL...right? If you have been following this thread, you know that Wrinex is dealing with other core related distortions besides just core saturation; distortions that were happening at low volumes levels not high.

...occasional instances of saturation may be of small importance to musical enjoyment, even if saturation makes nasty-sounding sorts of distortion...the reality of anticipating when the core will be saturated by the confluence of loudness and frequency is pretty hard to guess and even depends a bit on the kind of music being played.

While I agree that core saturation related distortion is difficult to predict for music signals, once you have heard it you will want to avoid it. It introduces a spray of higher order harmonics that forces hands over ears. Avoiding it could be as simple as not listening to those particular tracks that cause it above a certain volume setting...or a limiter as you suggest. But, if you can adjust the crossover design to avoid it I think that is a better solution.

BTW, during core saturation the transformer inductance falls to near zero. The result is that the amplifier will be trying to drive a near short...limited only by the series resistance of the wiring and primary coil. If your amplifier does not have over current protection built in you can easily destroy the output stage during transitory core saturation events. Sadly I do have "experimental" evidence to support this behavior.

 

[invisible force]

another point of view :
the parallel resonance frequency, caused by the transformer inductance parallel to transformed ESL capacity (+ layer winding capacity ..) is much higher for low listening volume (transformer inductance is low) ! parallel resonance means low current, this means low distortion...so it is logical : for low voltage distortion for frequencies below this parallel resonance is higher than for moderately increased listening volume..have fun

 

[esltransformer]

There are other / better solutions but as long people insist on low budget transformers it stays difficult to get good sound/ performance from an esl.

One of the thinks to overcome the problem is to use an airgap.

To avoid the large changes in the transformer primary inductance from compromising the crossover response you need to swamp it(make it unimportant) by placing a low impedance in parallel with it. For a 1st order crossover this is done by placing a low value resistor across the primary as in the Acoustat circuit I posted. For a 2nd order crossover this is done with a low value inductor across the primary; usually adding some resistance in series with the inductor is desireable for proper damping of the filter response.

Note the very low inductance value (0.25mH) Wrinex is putting in parallel with the primary. Even if the transformer primary inductance varies wildly on its own, once the inductor is place in parallel with it, the impedance across the primary will not stray far from that of a 0.25mH inductance.


This is absolutely true and was what I pointed out early in this thread, although perhaps not as clearly as you have.
http://www.diyaudio.com/forums/planars-exotics/226637-distortion-esl-low-level-4.html#post3317465

In this case, the OP is wanting to use a passive HP filter with an existing transformer he already owns.
Its primary impedance is unusually low, <1mH, so the only options to keep distortion low are:
1) drive it directly from a low impedance amplifier as you suggested, and use an active crossover
2) place a low impedance across the primary to swamp its effect as part of a passive crossover

 

[bolserst]

another point of view :
the parallel resonance frequency, caused by the transformer inductance parallel to transformed ESL capacity (+ layer winding capacity ..) is much higher for low listening volume (transformer inductance is low) ! parallel resonance means low current, this means low distortion...so it is logical : for low voltage distortion for frequencies below this parallel resonance is higher than for moderately increased listening volume..have fun

Makes sense to me
This was what I was trying to describe in post#39:
“...poor transformer iron has very low initial permeability and would result in very low primary inductance for low input voltages. So, only a small portion of the input voltage appears across the primary. This also means that the majority of the voltage drop due to any current distortion caused by the core will show up across the capacitor and distort the voltage fed to the primary. As voltage amplitude increases, the primary inductance increases and a greater portion of the input voltage will appear across the primary and distortion is reduced...”

There are other / better solutions but as long people insist on low budget transformers it stays difficult to get good sound/ performance from an esl.

I agree that simply purchasing a better transformer will neatly side-step a lot of performance problems. However, I find it worthwhile to understand the exact causes of performance problems with non-optimal transformers and determine ways to work around them. You may discover something useful to apply to better transformers that would not have been obvious since the distortion mode was not as prominent in the better transformer. Always new things to learn…

 

[esltransformer]

Makes sense to me
This was what I was trying to describe in post#39:
“...poor transformer iron has very low initial permeability and would result in very low primary inductance for low input voltages. So, only a small portion of the input voltage appears across the primary. This also means that the majority of the voltage drop due to any current distortion caused by the core will show up across the capacitor and distort the voltage fed to the primary. As voltage amplitude increases, the primary inductance increases and a greater portion of the input voltage will appear across the primary and distortion is reduced...”



I agree that simply purchasing a better transformer will neatly side-step a lot of performance problems. However, I find it worthwhile to understand the exact causes of performance problems with non-optimal transformers and determine ways to work around them. You may discover something useful to apply to better transformers that would not have been obvious since the distortion mode was not as prominent in the better transformer. Always new things to learn…

It almost sounds that there is a solution but there isn't.
You only can live with the problems and you can make it as less a possible (if possible).

Creating a potential super loudspeaker and saving money on a step-up transformer is i.m.o.very weird.

 

[WrineX]

true but i think allot of people only have money for a not to costly almost potential super loudspeaker

i mean if this small fix of 1 resistor fixes it for a cheap tranny then so be it, it may not sound as good as the far more expensive good trannie. but stil it sounds rahter good and far better then most speakers (mid high frequency).

so stil something gained, maybe not the maximum but hell it was sort of fun to build (although not for the past year)

 

[geraldfryjr]

My little system may not be a super super system, But it sounds Great!!!
And, I use inexpensive widely available transformers as well.

My ears and THD test's ( as low as .01% and lower) have shown that it is far better and superior than any commercial speaker I have ever layed any amount of money down for!!!

Plus the enjoyment of knowing I built and designed it is Priceless !!!

FWIW

jer