Drive an audio transformer from an AOP. Good or bad idea?

[abbey road d enfer]

"Zero" is more of a abstraction than a reality.

"Zero-ohm" drive is indeed a colloquialism for a circuit that has a very low source impedance compared to the impedance it drives. Concretely it means that, whatever the actual impedance of the intended load, the output voltage will be constant (within limits dictated by the application), and this, as long as in linear operation.

A strong OP amp with negative feedback can make a drive impedance very low, but adding some series resistance protects against bad loads.

When the resistance is included in the global NFB loop, the criterion for "zero-ohm" drive is maintained. Zero-ohm drive does not imply there is no limit to the output capability.

Obviously massive feedback has its own issues, like suppression of even order harmonics.

Why only even order? Do you mean that suppressing even-order harmonics is not a good thing?

 

[Matt Syson]

Yes to the 'brain fade' 2 squared is 4 so the Audix drivers were to handle 150 Ohms not 50. There is so much 'twaddle' written on the internet it is hard to remember the real details.

 

[jokeramik]

How so? The load goes with the square of the ratio - 600R load with 1:2 would be 150R.
https://en.wikipedia.org/wiki/Transformer

But let's be realistic - we don't have 600R loads out there. This is loooong gone, modern line inputs are in the 10k area.
Mic inputs can be down to 1k, standard in the 2k area but more and more I see 3k and higher in datasheets. (RME 3,4k, Midas 10k!) This makes sense to enable condenser mics for high output without distortion.
Microphones don't use step up transformers and you only have about 10mA over the phantom power line anyways. (OPA1692 could be used and is able to send 50mA - way more as you have available)

Many modern OPA can output 50mA (OPA1612) - 18V peak in 600R would need 30mA.
Just take care you have no DC at the output and you should be fine to drive a transformer.

I think the main problem is to get a source with low impedance in order to achieve the lowest possible introduction of noise and distortion. That's why I always use 50 ohm max. output impedance. This way there are very few problems with unbalanced connections. The V 781 of the RFZ of the GDR, for example, is between 3 ohm and 15 ohm output impedance. Also the MV 810/2, made by VEB Funkwerk Kölleda have a very low output impedance. All output-transformerss of RFZ and VEB Funkwerk Kölleda are 1:1 transformers with this low impedance.

The other problem is that step-up transformers increase the voltage linearly, but the resistance behaves quadratically.This is exploited in microphone output transformers to get from the relatively high internal resistance of the anode circuit to a low output resistance. With all the problems and difficulties that can result from this.

Best regards!

 

[Cqwet Dbdfte]

"Zero-ohm" drive is indeed a colloquialism for a circuit that has a very low source impedance compared to the impedance it drives. Concretely it means that, whatever the actual impedance of the intended load, the output voltage will be constant (within limits dictated by the application), and this, as long as in linear operation.

When the resistance is included in the global NFB loop, the criterion for "zero-ohm" drive is maintained. Zero-ohm drive does not imply there is no limit to the output capability.

Why only even order? Do you mean that suppressing even-order harmonics is not a good thing?

A bit off topic, but suppressing even order leaving odd order intact is not ideal. Push-pull transformers, audio and RF, "cancel" even order harmonics. In single ended power amps this does not happen, and perceived loudness depends on a harmonic series with descending even and odd harmonics. Not so much for low level amps, but the concept may be used to effect in sound generation.
Nelson Pass, Peter Baxandall, John Lindsley-Hood gets credit for their work in this area.

 

[jokeramik]

Don´t forget Douglas Self.
Best regards!

 

[Mr. Moscode]

My younger experienced ears tell me high NFB ckts suppress low level info, reverb tails, room tone, etc. That may be due to the suppression of harmonics mentioned, I don't know. IMHO, distortion down to .0001% from the use of NFB is a rabbit hole.

 

[soliloqueen]

My younger experienced ears tell me high NFB ckts suppress low level info, reverb tails, room tone, etc. That may be due to the suppression of harmonics mentioned, I don't know. IMHO, distortion down to .0001% from the use of NFB is a rabbit hole.

it may not be that they're suppressed, but just that they're not exaggerated. if the SNR of a circuit is higher, doesn't everything kind of get stretched "vertically" in some ways? or rather, they don't get squashed? even though there's no real mechanism for this and the dynamic range of audio is not at all the same as the DR of a picture or display, I can't deny that I have also subjectively experienced some loss of perceived space in lower noise circuits and recordings in general. the background of a truly clean amplifier circuit is so...blank. maybe performance at or near the noise floor isn't linear somehow, so the tails get picked up or represented more? i've noticed low bit depth recordings with noise shaping also increase space, and plugins that enhance percussion with noise bursts apply this psychoacoustic effect practically. personally, i do agree with you that they sound better, but i'm not sure if it's because something is being lost. maybe something pleasant is being added.

 

[ChrisMM]

it may not be that they're suppressed, but just that they're not exaggerated. if the SNR of a circuit is higher, doesn't everything kind of get stretched "vertically" in some ways? or rather, they don't get squashed? even though there's no real mechanism for this and the dynamic range of audio is not at all the same as the DR of a picture or display, I can't deny that I have also subjectively experienced some loss of perceived space in lower noise circuits and recordings in general. the background of a truly clean amplifier circuit is so...blank. maybe performance at or near the noise floor isn't linear somehow, so the tails get picked up or represented more? i've noticed low bit depth recordings with noise shaping also increase space, and plugins that enhance percussion with noise bursts apply this psychoacoustic effect practically. personally, i do agree with you that they sound better, but i'm not sure if it's because something is being lost. maybe something pleasant is being added.

I've wondered if very clean / low noise recordings may cause a kind of uncanny valley effect. That is to say, it sounds very close to reality, but is still not quite real; the combination of which we tend to experience as unsettling. For that reason we may subjectively prefer recordings that are objectively more noisy / distorted. Just a thought. I won't expound on it here since it's a bit off topic for the thread, but I find this stuff incredibly interesting.

-- Chris

 

[JF]

I keep it with Rupert Neve - a very low noise floor is NEEDED so you can hear all the fine details and reverb tails. There is an interview where he tells that they always get a circuit clean and with low distortion before they even start to give it a listen. Cause it's not worth to listen to noise or distortion.

I know the uncanny valley with cutting of audio tracks. When you have total silence between cuts your ears/feelings/brain get "sucked in", it feels very unnatural. But that's a beginners mistake, you should not find this situation with music out there. You always have "ambience" in audio tracks, even when it's very low it's enough to give orientation.
(In film audio they record silence for exact this reason to have ambience to cut in when needed. It's funny when you attend it for the first time. "Everybody shut up the next 5 Minutes - we are recording silence")

But it's fur sure not the job of your equipment to provide a noise floor! I want to hear everything of my source signal, not just the upper 85%.

NFB is a great thing when used properly! Which means it should work through the audio bandwidth (or at least to 10kHz?). Old/simple OPAs often have strong feedback at low frequencies but go down already significant at 1kHz and higher. Have a look at closed loop THD over frequency. PSRR or open loop gain is also interesting.
Not sure where the limit is when this get's audible, but it's not tooo hard to take care of, so I try to just do that.

 

[ricardo]

A nice transformer is made by SRS. A bit expensive.
https://www.thinksrs.com/products/sr554.html

This is actually quite a poor transformer for audio. I've used it as an illustration of various transformer evils

 

[Cqwet Dbdfte]

This is actually quite a poor transformer for audio. I've used it as an illustration of various transformer evils

How so?

 

[ricardo]

This is actually quite a poor transformer for audio. I've used it as an illustration of various transformer evils

How so?

Have a look at their published data for various impedances.

 

[Cqwet Dbdfte]

Have a look at their published data for various impedances.

This?
A high step ratio transformer, like this 100:1, is used for a low source impedance, like a 2 - 3 Ohm MC cartridge, pretty flat 20 - 20k.
Don't try this with a 600Ohm source.

 

[ricardo]

The 'negative impedance' approach to transformer distortion reduction was the subject of a German patent many years ago and indeed Studer and others used the principles. Before or after Ken Farrar of Calrec (RIP Ken) may never be known.

This was used by Calrec in the late 70s which is about the time of the Lundahl app. note. We were very close to them at the time as we were switching form Sowters to Lundahl cos the size. IIRC, there's a Lundahl patent too.

Have you got a link to the German patent and perhaps Studer circuits with dates?

Indeed Calder Recordings may never have happened if the inspiration of capacitor Microphones had not been an article in Wireless World.

A brief history of Calrec (from my somewhat fading memory) is https://groupdiy.com/threads/cadac-vs-calrec.64588/post-818047

Debenham, Robinson & Stebbings appeared in Tape Recorder Magazine later merged with HiFi News.

Ken Farrar did quite a bit of lateral thinking and came up with a tuneable Wein Bridge oscillator that only needed a single gang pot for tuning

This was actually by Great Guru Baxandall
A Low-Distortion Acoustic-Measurement Oscillator Using Semiconductor Junctions as Variable-Tuning Elements

Used for production testing at Calrec. (& KEF for nearly a decade) I had the (dubious?) honour of having to bring some of GGB's ideas to small scale production at Calrec. He was a true genius and taught me many things. But his idea of 'ready for production' dates back to working on early RADAR in WWII and relied heavily on 'hand carved from solid BS by virgins' methods.

But ALL his stuff is worth re-reading in your old age to pick out pearls of wisdom which you didn't see on first reading. John Vanderkooy agrees with me on this.
A Low-Distortion Fast-Settling Audio Oscillator: A Tribute to the late Peter J. Baxandall, Audio Analog Expert

He revisited LN design towards the end of his life and that article is on the WWW somewhere. It has a Transformer Design for LN subtext which is particularly valuable as the knowledge and expertise on this subject is fast becoming extinct and he was a true master.

 

[ricardo]

The problem with this circuit is that the current sense resistor must follow the primary resistance variations with temperature to provide optimum distortion compensation.

Yes. But there is a more serious problem with the Calrec/Lundahl circuit. You can't cancel out all the distortion as the whole thing becomes unstable.

As 'Zo' approaches zero, LF bandwidth is extended but the caps in this circuit give rise to a huge VLF peak. You really need to precede this output stage with a HP filter.

Don't think the AP circuit has this problem (no capacitors). Might have to leave beach bum mode & sim this but the Great Barrier Reef calls.

My choice of a tertiary winding makes this unnecessary, and the circuit is not more complex.
Bill Whitlock (member CMRR) may chime on it, as he was stakeholder in the AP/Jensen collab.

Abbey, have you posted this anywhere? We (Calrec) didn't want a tertiary winding. That's the main reason I think the original idea was Ken Farrar's cos Lundahl would have used another winding.

 

[abbey road d enfer]

Yes. But there is a more serious problem with the Calrec/Lundahl circuit. You can't cancel out all the distortion as the whole thing becomes unstable.

As 'Zo' approaches zero, LF bandwidth is extended but the caps in this circuit give rise to a huge VLF peak. You really need to precede this output stage with a HP filter.

That's one of the issues. The other is that the more you decrease distortion, the closer you are to unstability. If for some reason the sum of negative drive impedance and DC resistance of primary goes below zero (negative) the system oscillates.

Abbey, have you posted this anywhere?

There for example:
https://groupdiy.com/threads/scv-electronics.65116/#post-908523

We (Calrec) didn't want a tertiary winding.

Why? the pros largely overthrow the cons IMO.

 

[13engrsapper]

cost is my guess on the why.

That's one of the issues. The other is that the more you decrease distortion, the closer you are to unstability. If for some reason the sum of negative drive impedance and DC resistance of primary goes below zero (negative) the system oscillates.

There for example:
https://groupdiy.com/threads/scv-electronics.65116/#post-908523

Why? the pros largely overthrow the cons IMO.

I know I have seen some Beyer type transformers in something Studer - But who made the rest of them? Studer themselves?

 

[abbey road d enfer]

cost is my guess on the why.

Compared to a normal xfmr, a quadfilar is marginally more costly, maybe 20% more.

I know I have seen some Beyer type transformers in something Studer - But who made the rest of them? Studer themselves?

AFAIK Beyer subcontracts their xfmrs, probably from the same supplier that also makes Neutrik xfmrs.

 

[zamproject]

I know I have seen some Beyer type transformers in something Studer

Yes there is Beyer branded Xformer in early Studer desk
Later it's a Studer internal ref, like the -macaroni- input Xfmr we see in A80 or x89 series desk, as many output transformers, I never see those form anywhere else, so maybe at Studer golden age and big plant they have they own machinery to manufacture them ?!?
For the Neutrik form (input Xfmr) that appear in the late 70' and in all Studer product later, a guy close to Studer veterans told me once it's a studer manufacturing... and later they transfer the technology and processes to neutrik... but it's the same guy claiming that studer A1/2/3 monitoring were studer design wile it's a subcontract from the company that now manufacture the PSI line.

-edit- did i say macaroni...non ravioli !!!

 

[13engrsapper]

I have piles of those - in modules, lol