Zero Feedback Impedance Amplifiers

[bocka]

Dear Susan, Dear John,

thanks a lot for your responeses about the RF amp. It's just a quick idea the same as using a lot of OPs and parallel them. I know the RF design is completely different from power amp design. But 40MHz and about 2MHz what I got with the fastest designs developed for audio seems not so far away.

On topic - Susan you really ashamed me. When I started 25 years ago building amps first I learned is to use feedback. Well that had some tradeoffs I build more oscillators than amps. Another 10 years later - my amps are more stable that time - I read, that a lot feedback will cause some audible problems. So I removed feedback with the effect of more complex circuits. I also build some tube amps, no power but headphone and preamp. Yes they are looking very nice, that glowing is really impressive. And that gave me the inspiration to make semiconductor amps better. Well, some power regulation / stabilisation here some CCS over there, I think my latest designs are not much more simple than that commercial stuff (I repaired a lot of them in the past years!)

And now you came with a two transistor design, added two transformer in the in- and output. Just like that golden time where transistors cost more than transformers. And you tell me it sounds good. Shall I really believe this? As well as tube amps they are all from an ancient time. Cum on we're living in a time of highly integrated semiconductor chips, computers, where billions of transistors are in and connected in a very complex manner, this cannot be true. Why we need then the use of the other billion transistors? They cannot be out of job!

Using some high quality parts, the less as possible in the signal path and made it in the RIGHT other way round. Susan - your design is really cool.

[janneman]

Susan,

On the other hand, while I am sitting here listening to Norah Jones through my Nemisis monoblocks, I wonder why one would need per channel no less than 2 transformers and two transistors, when half of that would be perfectly adequate...

But, Susan, seriously, are you familiar with the Nemesis? Do you hasve an opinion on how your amp would compare to that, soundwise?

Jan Didden

[Upupa Epops]

Jan, one is too much - maybe best result give you horn and shelack with 78 rpm .

[Circlotron]

Hmmm... I wonder if this amplifier was originally made with P-channel fets but later got changed to an N-channel cct so that's what it is now?

[Brian Clark]

Wahey! Only just read this thread. I love that topology Susan! My present single-ended amp is in the same spirit but uses a single 801A valve in each channel with a Lundahl amorphous core stepup on the input as driver. Works a treat and everybody who hears it wants me to build them one. Need sensitive speaks though

Semper Musica!
Semper Simplistica!

Brian.

[Susan-Parker]

Hi everyone,

Many thanks for your continued interest.

I have reasembled my test setup to be able to make some of the measurments requested and incorporated this in updates to my amplifier web page:

http://www.susan-parker.co.uk/zeus.htm

which now have links to two transformer spec pages.

Some highlights...

Input impedance = nominal 150 ohms parallel and 600 ohms series wiring of primarys.

Termination resistance is 120 k ohm.

Output impedance is 2.78 ohms at 1 kHz.

Laminations are 0.35 mm thick.

There is also inductance etc. for those happy bunnies who can make working Spice transformer models. Of course one also needs working mosfet models, but that is another thread.

Best wishes,
Susan.

[Susan-Parker]

Hi Graham.

Originally posted by Graham Maynard
[B]Hi Susan,

What's inside your amp?
In your case that really is a redundant question; but oh how much thinking you must already have done about so little !!!

Thank you for your appreciation. And over a decade in time, sigh.

I'm moving over to your thread here, and I note that there already are some worthy contributions. I'd love to hear your amp, but alas I live north of Belfast so it is not likely I can call round for a demo.

Guess I will have to put some bits together and send you a parcel. I assume you have power supplies n' stuff.

In jcx's other thread I suggested that your input transformer would need to be something rather special, I was thinking of a need for about 1:20+20 of grain orientation + split windings. I see you have described it as 1;5+5, which allows you to use bell wire for the input cable. However, I think figure-8 pairing has a natural impedance closer to 100 rather than 600 ohms, though 600 ohm driving should not have too great an effect at audio frequencies if the cable is only of 'domestic' length. If signal source push-pull/balanced driven at say series 56 ohms per core wrt to ground, the cable should have an unchallengable hf phase response, and could be loaded with a resistive portion across the transformer primary as well as secondary. ( I wrote this yesterday and see that Hugh too has come up with lower impedance and higher ratio, maybe there is room for more investigation here.)

Input transformer input impedance at 1 kHz:
Parallel = 165 Ω.
Series = 667 Ω.

I normally use parallel mode whcih gives me a 1:10. Series is 1:5.

As you have probably noticed my amplifer is a power stage, and does require a seperate driver after the attenuator section.

This is a different split from normal but based on technical reason not some glitzy marketing speil.

I have no wish to upset or dampen your enthusiasm, but as with PMA I too would wonder about phase accuracy, because an amplifier can sound good enough for listening satisfaction (and easily better a Quad set-up) without it being as accurate as is possible with other designs that are equally capable of resolving swaying or breathing musicians. You mention a 300kHz bandwidth; is this -3dB, and did you ever run an audio frequency phase check ? I suspect transformer core losses, which become both measurable and so audible when you A/B compare mains types pressed into audio usage when compared with properly grain orientated versions.

The windings of my output transformer is different and doesn't model the same, however I would like to use better materials. If I ask Sowter to make a pair I guess I will get properly grain orientated laminations?

I also want to try using some mu-metal interleaving.

With regard to driving the loudspeakers directly from source + c.t. transformer connections used as a choke I recon that per half hf voltage due to loss of coupling between halves and per half inductance might be theorectically less natural than via your separate core coupled winding, but it would be worth auditioning in case it fortuitously compensates for transformer imperfections.

With quad filar winding the coupling is very good, far better than a conventional winding topology. There will be some loss but I am not sure exactly how much. I would have to try out this alternative but overall I prefer the seperate output windings as this gives more flexability in load matching.

Did you ever measure the output resistance/impedance at say 1kHz ? Have you managed to achieve a decent loudspeaker damping figure without NFB ? Have you tried listening whilst adjusting the quiescent bias and found a level at which further increase does not improve reproduction, or did you set for a reasonable dissipation ? Have you ever subjected the amplifier to distortion test set measurement ?

2.78 ohm at 1 kHz.

I set for reasonable disapation. There is some change as the bias kicks in.

I have some test gear (including an HP Audio Analyser) but need to sort out and match some power devices - plus the time to do this as I have some short term work pressures.

Like johnnyx I could ask myself if I need to change direction, as inded I have already have following Pavel's new circuit release, but as I prefer transformerless power, regard 100W-8 ohms a requirement, and as it is possible to minimise feedback induced problems, then I would need to know phase, distortion and output impedance figures of your design before I could be convinced. I would rather have an accurate feedback amplifier and then pre-amp modify its signal for listening that suits my ears, than have an amplifier that is easy to listen to but which compromises waveform accuracy in a way that can never be subsequently adjusted out.

I think that the output distorsion levels are such that it is subserviant to the speaker driver/crossover distorsions. The ultimate distorsion after all being the combination of the two, which is why I believe valve level distorsion whilst "awful" on paper often don't sound as bad as one could expect.

I understand the 100 W - 8 ohms but in my experience I rarely use more than 5 W or so except for daytime classical listening (the loud bits).

A lot depends on the speaker sensitivity.

I don't beleive that it is possible to completly remove the loudspeaker's reactive and frequency effects on the feedback network of the amplifier. It can be minimised for a certain range of conditions and tuned to a preference, but it will always be there.

The only wayI see to make a feedback amplifer work without artifact is to have two output stages, one with a non inductive load of say 12 ohms (or whatever would be a good overall match to the nominal loudspeaker impedance) and used for the amplifier feedback, and the second slaved but WITHOUT any feedback connection which is used to drive the loudspeaker.

Obvioulsy the output devices have to be matched and closely mounted on a "nice big heatsink".

As to my transformer design - it's a third type, neither valve or solid state, and doesn't to my ear have the same sound as either. Personally what I hear most is the quality of the source material rather than a "valve" or "solid state" sound.

But one becomes used to what one listens to, which I would agree doesn't mean the same as better (or worse for that matter).

I must state that I have not heard this kind of amplifier, and so I cannot state whether it is cleaner sounding than a feedback amplifier or not, so please do not take my points as being either an attempt to be definitive or derogatory.

Perfectly appreciated.

You like the Line Source too ! What is your driver ? Mine is Visaton FRS-8, far from perfect, but cheap and clean over mid range and needs a central top end tweeter, though I'm always on the lookout for other suggestions.

I was looking at the Jordan full range drivers - which I use in my marble spheres - but they are a bit bright at the top end with my amplifiers.

I am thinking about other drivers - more conventional but less expensive.

Many thanks for your time and interest.

Best wishes,
Susan.


P.S.

Sorry it's late now, I will have to reply to other posts tomorrow.

[dimitri]

http://www.stereophile.com/amplifica...40/index6.html

Speaking of single-ended amplifiers, it's worth noting the work of Paris-based Jean Hiraga at L'Audiophile over 10 years ago. He developed a single-ended, transformer-coupled transistor power amplifier called Nemesis, with no other active devices whatsoever other than a single power device (namely a 25SK 135 power FET). The Nemesis's utter simplicity helped it achieve a sound that closely approached that of a classic WE 300B single-ended triode amplifier.—Martin Colloms

Audiophile n°34, janvier 1985
Audiophile n°35, printemps 1985

[Susan-Parker]

Hi Dimitri,

Thank you for the reference...

Originally posted by dimitri
http://www.stereophile.com/amplifica...40/index6.html

Speaking of single-ended amplifiers, it's worth noting the work of Paris-based Jean Hiraga at L'Audiophile over 10 years ago. He developed a single-ended, transformer-coupled transistor power amplifier called Nemesis, with no other active devices whatsoever other than a single power device (namely a 25SK 135 power FET). The Nemesis's utter simplicity helped it achieve a sound that closely approached that of a classic WE 300B single-ended triode amplifier.—Martin Colloms

Audiophile n°34, janvier 1985
Audiophile n°35, printemps 1985

... which makes interesting reading.

However the Nemesis is a valve design with the power fet instead of a tube, and the gain stage is the FET transfer characteristics, which as Eva has pointed out are a bit shakey.

These single ended designs by their nature have an ASYMETRIC output impedance, which will affect the reproduced sound quality. They can only drive in one direction, the other direction is dictated by the back EMF of the transformer or by the current sink.

Again this is not necessarily good or bad, it's just how it is.

My amplifier uses an impedance gain stage, which does not add any noise. It is literally "wire with gain". To do this voltage amplification I trade off a low impedance input for a high impedance output.

Of course, as has been pointed out by several people, transformers are not linear in the same way as a feedback loop amplifier. But then I am not aiming for 0.0001% THD into a non-inductive 8 ohm resistor - because listening to a resistor is very boring!

My amplifier is push-pull off a single supply rail which gives a number of advantages.

1. Very high supply ripple rejection - small (i.e. cheep) smoothing cap.
2. Active drive in both directions at all times - constant impedance.
3. Low quiescent bias - circa 500 to 600 mA - less heat.

... but I am beginning to repeat myself, which is also boring.

Sufficient to say I am not trying to reproduce a Triode sound (nor a solid state sound for that matter). I am trying to reproduce the sound encoded in the original recording - no more, no less.

It should be remembered that the average signal for a classical CD with large peaks is only around 12 bits resolution, i.e. 1 in 4096, which is 0.025% distorsion before one even starts with anything else like the woefully inadequate sampling rate of 44 kHz.

I know this because I built my own CD player using a CD-ROM drive and striped out the digital audio data from the signal processing chip to feed into my own Burr Brown DACs to then drive my amps directly.

All this oversamplig filter stuff is a fudge, and introduces it's own signature to the sound. It can be a very good fudge, and be more pleasant to listen to, but that doesn't negate the fact that it is modifying the original sound.

Any item that can be interchanged with another of similar function and that results in a different sound will be doing this, with the one exception of the single item in the group that isn't modifying the sound. But that one probably won't be recognised as such, given that each will be subtly (or not so subtly) different anyway.

So at the end of the day it comes down to personal choise and preferance (and spelling) and what works best for one person isn't necessarily going to be good for another.

Best wishes,
Susan.

[Upupa Epops]

Whatabout damping factor, Susan ? Do you think, that several ohms is good ? By this value is bass like snowball throwed on tent .