Transformer output and feedback

[bentoronto]

Two related questions:

1. What is the correct step-up transformer output voltage curve for getting constant loudness across the frequency spectrum out of a specific cell? Or does it vary with various factors in various cells?

2. What is the potential benefit of running feedback from the transformer output back to some point in the amp in order to ensure the transformer is putting a proper signal into the cell?

Thanks.
Ben

 

[bolserst]

A1. It depends on the size/shape of the ESL panels and if they are segmented or not. In the case of a small un-segmented panel, you would need stator voltage falling -6dB/oct to get a flat acoustic response at the listening position. Un-segmented line sources would need a voltage falling -3dB/oct. Segmentation is usually designed to provide flat acoustic response for a flat voltage input.

A2. The main benefit for running feedback around the transformer would be to reduce distortion. If the transformer is driven by an amplifier with low output impedance, distortion will be minimal and predominately a low frequency problem. Since this is also the area where distortion starts rising in the ESL panel, one might consider wrapping the feedback loop around the ESL as well. Both Beveridge and Wright provided details in their patents on how to accomplish this, although I think only Beveridge sold products with the concept implemented.

Any tailoring of the frequency response is more easily accomplished with upstream analog or digital filters.

 

[bentoronto]

Much appreciated, as always.

I gather from A1, that we can look at "total" output into a reverberant space as constant with voltage and frequency. But due to beaming, the loudness at a given spot on-axis, would need less and less voltage with rising beaming in an anechoic-like space. And varying in an orderly manner with points in-between.

As for A2, glad to hear not worth fussing over minor distortions in the step-up transformer.

But I'm not sure how much more "wrapping" to include the cells can be done once you're downstream to tie-in to the step-up transformer secondary (except with a mike)?

Ben

 

[bolserst]

I gather from A1, that we can look at "total" output into a reverberant space as constant with voltage and frequency. But due to beaming, the loudness at a given spot on-axis, would need less and less voltage with rising beaming in an anechoic-like space. And varying in an orderly manner with points in-between.

Correct, constant applied voltage results in constant radiated acoustic power from dipole ESL panels.
For frequencies where the wavelength is shorter than the panel dimensions this acoustic energy is progressively focused into a tighter and tighter beam resulting in SPL increasing with rising frequency. For frequencies where the wavelength is longer than the panel dimensions, the opposite phase acoustic energy radiated from the rear of the panel starts to diffract around the panel and combine/cancel out the front radiation. The lower the frequency, the closer the phase difference between the two source will be to 180deg and the more cancellation takes place. This is the so called dipole cancellation resulting in SPL decreasing with falling frequency.

Some plots depicting this behavior here:
http://www.diyaudio.com/forums/plan...se-cancellation-esl-speakers.html#post2856955

As for A2, glad to hear not worth fussing over minor distortions in the step-up transformer.

Note that this is only the case if the transformer is driven from a low impedance source. If you use a damping resistor > 1 – 2 ohm or use a passive crossover, then core related transformer distortion can be something to fuss over.

But I'm not sure how much more "wrapping" to include the cells can be done once you're downstream to tie-in to the step-up transformer secondary (except with a mike)?

In simple terms, the Beveridge method injects a low level RF signal on to the diaphragm. The RF signal is capacitively coupled to the stators, but will be AM modulated by the diaphragm motion. The difference between the demodulated stator RF signals will give you a signal that directly follows the average position of the entire diaphragm, neatly side-stepping any diaphragm resonance mode issues. You then need to differentiate this signal to get something proportional to far field SPL before closing the feedback loop. The Wright method works in the opposite direction, injecting two separate RF frequencies(one to each stator), and then picks up both signals on the diaphragm and compares their demodulated magnitude to get diaphragm position.

If interested, patents are: Beveridge (US 3668335), Wright (US 3135838)

I forgot to mention that Wright also suggested using an auxiliary winding on the transformer to provide feedback in one of his later patents. (US 3668336)
Hmmmm...that patent number is conspicuously close to the Beveridge patent mentioned above

 

[bentoronto]

Beveridge RF motional feedback patent (mostly with tubes!) very interesting. Thanks for citation. More elaborate than I'd consider.

Would the following be a sensible take-away message: no great advantages to having feedback originating downstream of the transformer since (1) panels and transformers do not add much distortion and (2) tonal irregularities can be addressed with EQ.

Ben

 

[bolserst]

Yes, that summarizes my experience.
(1) If the step-up transform is driven from a low impedance source, and ESL panel is operated in constant charge mode, there is just not much non-linearities there that warrant the complication of adding feedback. Beveridge operated his full range ESLs in constant voltage mode, so feedback made sense to correct square-law force distortion.
(2) Tonal or response irregularities are best addressed with segmentation or EQ.