Sound Quality Vs. Measurements

[tvrgeek]

T,
Not ready to build a 50W class A amp, yet.

[VladimirK]

Speaking about optimum amount of global NFB, I would propose to pay more attention on how exactly the NFB is arranged in a given schematics.
In all cases I experimented with, current type NFB (tenths of mA flow through the NFB resistive divider) always lead to more resolved and articulated bass, compared to voltage NFB (tenths of uA flow) of the same amount.

[ThorstenL]

Hi,

Originally Posted by tvrgeek

I have a cheap ( Behringer) digital crossover because it is easy for quick testing on the bench for designing speakers. It is no where near a high fidelity unit. Not even close to their $100 analog ones which are surprisingly not as bad as one may think. Not what one can build though.

I just spend time on an install with Funktion One's version of the XTA digital crossovers. Sound quality is much worse than just driving the Amp's directly from the PC and EQ'ing the Funktion One R1 tops's of the Club in the PC. The XTA's A2D and D2A have a significant sonic footprint even with the whole thing on bypass. And XTA's stuff is reckoned by many miles above Behringer.

In the end I used it, because we had ordered it, I needed the availability of delayed feeds (there is nearly 60mS covering all the speakers in the club) and it was very easy to EQ the system to both measure and sound right.

Originally Posted by tvrgeek

So distortion added by an inductor is what the model does. When I model just a passive filter, I also get quite high distortion. What the model is assuming I have no idea.

Are you using a Non-linear model for an Aircore inductor?

Originally Posted by tvrgeek

A lot of designs take some feedback before the outputs in that the drivers are not just tied together but the center point also tied to the output.

That is not "feedback before the outputs", instead it is a Class AB Driver stage (usually a bad idea).

Originally Posted by tvrgeek

I guess I am not following how to add local feedback to the outputs themselves. Larger emitter resistors?

Emitter/Source followers already have 100% NFB, so you cannot apply more. Tubes usually operate as amplifiers and not followers, similar topologies exist in solid state audio. James Strickland's "Trans Nova" Circuit is an example of an Amplifier that combines a local loop around the output stage with positive feedback inside the driver stage and global looped feedback.

Ciao T

[ThorstenL]

Vladimir,

Originally Posted by VladimirK

Speaking about optimum amount of global NFB, I would propose to pay more attention on how exactly the NFB is arranged in a given schematics.
In all cases I experimented with, current type NFB (tenths of mA flow through the NFB resistive divider) always lead to more resolved and articulated bass, compared to voltage NFB (tenths of uA flow) of the same amount.

In my Tube circuits it invariably is current feedback, but in solid state I have used both. My experience is that the amount needed relates to topology.

For Class A SE and Push Pull Triode I find looped NFB is entirely optional, unless the speaker requires extra damping over that offered by the the natural output impedance of the output stage or greater linearity is required.

The deeper we get into "Class AB" the more NFB around the output stage seems needed to overcome the sonic problems Class AB adds.

Ciao T

[dvv]

Originally Posted by ThorstenL

Hi,



I have done such things using Tubes.

In one case the system consisted out of two ESL Panels (appx. 100Hz - 500Hz and 500Hz to above 20KHz) per channel plus a dual 18"Dipole Sub.

The crossover filters and EQ was designed into the Amplifier, which where class A DHT Push-Pull, 300B for the Upper Mid / Treble and 845 for the upper bass / lower mid, non-lopped feedback of course. The Bass Amp was a 1KW Mosfet Module with the lowpass designed in as Sallen Key filter and the dipole EQ in the feedback loop.

The Tube Amplifiers employed normal output transformers and could be switched full-range & flat for driving more conventional speakers, but in system they drove the ESL Stators directly via capacitive coupling from the Anodes.

High pass filters combined the interstage coupling cap's and others to form a 3rd order HPF, while the lowpass & eq was done first order (assymetric slope XO).

If your output stage is Class AB you will need some NFB.

Well, it seems my thinking was not too far off, then. Of course, in that case you have a symbiotic relationship with the speakers used, the arrangement will hold true for that one model only, and everything else would be a lottery.

So far my work leads me to believe that the correct number for NFB around a optimum or overbiased Class AB Output stage is around 12dB, this holding for Tubes as much as for Transistors.

More seems to lead to less favourable harmonic distribution and transient behaviour in technical terms and to a more edgy and constipated, subjectively, as NFB ratio's are increased, less seems to cause a sound that is overly soft & wooly and "inconsistent with level". My experimentation remains limited though and other means (like Feed Forward Error Correction) may be employed instead of NFB (and IME with benefit).

Class A Amplifiers can use local degeneration and/or loops only without negative issues, IME.

Ciao T

Ah, it seems that Richard Miller and H/K agree with you, because that's precisely the global NFB my H/K 680 integrated amps has.

Of course, that doesn't make it the perfect amp, but it is rather good.

[homemodder]

Originally Posted by dvv

The reason why I use so few Japanese transistors is quite simply that they are unavailable to me where I live. And frankly, I'm sick and tired of having to import every little ninny each and every time.

As a point of interest, I believe that transistors do have a sound of their own, i.e. that they are NOT absolutely neutral in all respects. As most people (I think?), I tend to use devices I am well acquainted with, but more thjan that, devices which I can easily get hold of.

The local market has been taken over many decades ago by European manufacturers, and consequently, I am well acquainted with many (but certainly not all, not even most) European transistors. The only non-European company which rules with a strong hand locally is Motorola/ON Semi, they've been present here since the early 70ies.

I have yet to see and hear a transistor with better HF performance than the BF 720/721. They are grossly misrepresented in catalogs, in real life they do a LOT better than stated, assuming only you buy them from Siemens of Philips (no biggie, Farnell has them). Their only inconvenience is that they are intended for SMD, but, with a little patience nd a fine soldering iron tip, that's not really a big problem.

Of course, if I had a decent selection of Japanese transistors readily available, I would certainly give them a much more serious look.

Oddly enough, Dan d'Agostino didn't mind having many of them set in parallel, and nobody can call him an amateur. Not at his prices.

For the bf721 range I can vouch, I bought over a thousand pairs from zetex, for use in vas and cascodes, if high frequency distortion is what youre mostly targeting these are better than most japanese types.

[VladimirK]

Originally Posted by ThorstenL

Vladimir, In my Tube circuits it invariably is current feedback, but in solid state I have used both. My experience is that the amount needed relates to topology.

For Class A SE and Push Pull Triode I find looped NFB is entirely optional, unless the speaker requires extra damping over that offered by the the natural output impedance of the output stage or greater linearity is required.

The deeper we get into "Class AB" the more NFB around the output stage seems needed to overcome the sonic problems Class AB adds.

Thorsten, amount of possible schematics variations is huge, and it is almost useless to speak with big generalization. In more specific cases, I used to listen to several modifications of schematics, all of them had similar SE class A transistor output. Some of them had 0,1...0,2 Ohms output impedance modulus, without NFB, the output impedance was determined mainly by transconductance. Other had low transconductance MOSFETs and even power jFETs at the output, and even WITH NFB APPLIED their output impedance was 0,4...0,6 Ohms. On contrary to general expectations, bass detality and articulation is definitely better with NFB versions. Speakers are PMC EB1i. Specific amp topology rules for specific speakers. The Yamaha B2-x vintage amp (0,04 Ohms Rout) gives nothing good for bass articulation, compared to my low-wattage amps with current NFB.

[ThorstenL]

Hi,

Originally Posted by VladimirK

Thorsten, amount of possible schematics variations is huge

Not really. Basic schematics use only a few variations of building blocks.

Originally Posted by VladimirK

Speakers are PMC EB1i.

Probably one of the Speakers least suited to be operated with appreciable source impedance, even ATC are not that bad... Your example illustrates the qualifications I had made.

Ciao T

[dvv]

Originally Posted by homemodder

For the bf721 range I can vouch, I bought over a thousand pairs from zetex, for use in vas and cascodes, if high frequency distortion is what youre mostly targeting these are better than most japanese types.

Zetex is NOT the source to buy them from. I can understand your frustration, I had a few Zetex made units and they were very different from the same but from Siemens.

These are very manufacture sensitive devices, I think more so than we are generally used to.

The source to buy them from is Siemens, or at worst, Philips. NOBODY else.

BTW, the key reason I like them so much, when from Siemens, is precisely because there is so little HF distortion. And, truth be told, their power handling of 1.8 Watts does come in handy, too.

On a more general scale, I refuse to buy Zetex products. Their nominally same products most often differ quite a bit from the "originals". I neither have the time, nor the will to research Zetex' ideas of how it should work, Siemens may be a boring German company, but they are reliable and very consistent; if that's boring, pass me your share of it if you don't want it.

[jan.didden]

May I post with something related to the topic of this thread?

I spend some time recently doing controlled listening test between an SACD signal, unprocessed, and through a 15kHz brickwall-filtered, heavily phase-shifted version. Yes we did hear a difference (or most of us did, anyway).
Read my little report here.

Comments?

jan