Mixed feedback drive circuits for audio output transformers

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
> Is this really what we want in audio business?

Often real studios and networks need the isolation that only a transformer can give, without the distortion or the cost/weight of low-low-distortion iron.

> Does someone has studio experience with the sound of this circuit?

It is fairly common in the big German broadcast systems. A transformer the size of your thumb can show 0.01% THD to the lowest audio frequency.
 
I tinkered with a negative impedance converter several months ago, to see if I could hear any improvement with something like an Edcor output transformer, I didn't feel like there was an impressively huge difference between the NIC and the regular driver.
 
See also in this connection the Cabot (?)-Ap patent on using a noninductive copper winding on the trafo to temp compensate the positive feedback drive. UREI licensed this for some number of years, and its use in Ap's allows galvanic isolation while fine-tuning the negative Z drive for low output Z and distortion.
 
I was curious about that also and tried some tests to drive a LL1539 with a balanced mixed feedback circuit
http://www.lundahl.se/wp-content/uploads/datasheets/1539.pdf

Here is the Lundahl paper about mixed feedback drive http://www.lundahl.se/wp-content/uploads/datasheets/feedback.pdf

Gain = 2 • RF/ RIN • Transformer turns ratio.
Select RR ~ RIN
Select CR such that 1 / (2π • RR • CR ) << FMIN , the lowest desired output frequency.
Optimum R <= Rprimary • (RIN/RF) / 2, where Rprimary is copper resistance of primary winding(s).
Select R for good THD at LF, and for good square wave response.

  What is the copper resistance of a winding? is it the dc resistance we measure with an ohmmeter? is it the "static resistance" given in the transformer datasheet ? (maybe both are the same and only thing?)
 
Michael Krusch said:
http://www.lundahl.se/pdfs/papers/feedbck.pdf
Using mixed feedback drive circuits with audio output transformers have two major advantages:
1. Transformer-caused distortion is reduced (or almost eliminated)
...

Is this really what we want in audio business? Does someone has studio experience with the sound of this circuit?
I have routinely used this in most of the products I designed. This is what I want in my audio business, when faced with the apparent necessity of including a transformer in the product (most of the times for marketing reasons).
Audio Precision use this in their über-performance audio test sets.
Neve and Focusrite have used this technique (or alternatively tertiary NFB, which gives the same results) quite often.

I've always felt the "transformer sound" is a nuisance, that muddies the midrange. Using NZ or tertiary NFB has only benefits for me; lowers THD, extends BW and provides active hum shielding.
 
Note transformers designed for this approach don't have the inductance of an impedance matching design, and generally will not perform well if this approach isn't used. 
 
saint gillis said:
Ok!
But they say in the LL1539 data sheet :
Static resistance of each primary half (4 -- 1&6 or 3 -- 2&5 respectively): 20 Ω
And I measure 47Ohm with the ohmeter between 4 & 1-6  ...
My mistake! I though 1&6 and 2&5 were connected inside but they re not, so yes it is 20Ohms with the ohmeter!
 
saint gillis said:
So here's how I ve driven the LL1539 :
http://img15.hostingpics.net/pics/971105mixedfeedback.png

Does it look correct to you? Are there things I can improve?
The negative resistance here is about 12r, which  compensates only about half of the DC resistance, so the improvement in performance will be marginal. Increaing the 5.6r to 9.1 would compensate 95% of the DCr.
In order to bring significant improvement, the negative resistance must be  very close to the DCr of the primary, so trimming is often necessary. The issue there is that the closer you get to perfect cancellation, the closer you get to overcompensation, i.e. unstability, and any variation, in particular in temperature may lead the system to oscillate.
That's why I eventually abandoned this technique in favor of tertiary NFB, which provides better overall performance and requires no trimming.
 
abbey road d enfer said:
The negative resistance here is about 12r, which  compensates only about half of the DC resistance, so the improvement in performance will be marginal. Increaing the 5.6r to 9.1 would compensate 95% of the DCr.
In order to bring significant improvement, the negative resistance must be  very close to the DCr of the primary, so trimming is often necessary. The issue there is that the closer you get to perfect cancellation, the closer you get to overcompensation, i.e. unstability, and any variation, in particular in temperature may lead the system to oscillate.
That's why I eventually abandoned this technique in favor of tertiary NFB, which provides better overall performance and requires no trimming.

Yes I realized that oscillation occurured with a negative resistance higher than 12 Ohms.
I guess tertiary NFB needs special transformers, can it be implemented with a LL1539?
 
saint gillis said:
abbey road d enfer said:
The negative resistance here is about 12r, which  compensates only about half of the DC resistance, so the improvement in performance will be marginal. Increaing the 5.6r to 9.1 would compensate 95% of the DCr.
In order to bring significant improvement, the negative resistance must be  very close to the DCr of the primary, so trimming is often necessary. The issue there is that the closer you get to perfect cancellation, the closer you get to overcompensation, i.e. unstability, and any variation, in particular in temperature may lead the system to oscillate.
That's why I eventually abandoned this technique in favor of tertiary NFB, which provides better overall performance and requires no trimming.

Yes I realized that oscillation occurured with a negative resistance higher than 12 Ohms.
In addition to the basic stability theorem, there are parasitic capacitances that play games; I used a capacitor that progressively reversed to voltage drive at HF. I guess it could be done in the published topology by inserting an RC low-pass before the non-inverting input of the upper opamp.
I guess tertiary NFB needs special transformers, can it be implemented with a LL1539?
You may try; I used a custom-designed xfmr that was originally specified for 4x150ohms. I found out that using one of the windings for tertiary NFB was perfectly adequate and a 5532 could drive it at full level (+20dBu) down to 22Hz.
Then you're faced with the alternative of using two of the windings in parallels for drive and one for output, or the contrary. I ended up using two windings for the output and only for drive, but IIRC, there was not much operational difference. In fact in several applications I used two separate secondaries, for duplicated outputs.
 

Latest posts

Back
Top