Sound Quality Vs. Measurements

[Wavebourn]

Originally Posted by ThorstenL

Hi,



Then this will severely disturb you:



That's a 1KHz squarewave, on axis, at 1KHz for a Quad ESL-63...

Ciao T

Quite pretty picture for speakers!

[bcarso]

Originally Posted by tvrgeek

[snip]

bcarso. you mean the old Radio Shack book of 200 circuits was not everything in the world? Gasp, how we were miss-led! Well,everything new has not already been written in software either, even though some managers seem to think so. Apache provides foundation, not solutions. Problem is, the colleges are teaching only how it integrate Java libraries, not how to program. It is getting darn hard to hire real software engineers. Or at least ones who can pass the drug test.

When I left UCLA to join the family business, in 1985 BH (Before Harman), my father, who was a very clever but almost wholly intuitive engineer, gave me a design job right off the bat. It was a testing fixture for telephones and telephone accessories, and he had promised it to a big guy at Service Merchandise Corp., a company that bought a lot of Amco products. He had gotten as far as to identify some chips that decoded touch tones and dial pulses, and had something of a start on generating the 20Hz ringing voltages. Also in the mix was a tester for modular cables.

So I started to think about it, and scribbling things. I could see that he was a bit ill-at-ease with this approach, and never being one to conceal frustration, finally lost patience (of which he had very little) and said G$%@#%*&&! Why don't you look in these books? By which he meant these doorstop-sized compendia edited by whomever, full of circuits collected from manufacturers' app notes, "ideas for design" magazine sections, and the ilk. "Free" circuits that --- surprise --- are usually worth about what you pay for them. So I grudgingly did consult one and another, found one circuit somewhat usable but suboptimal for the cable tester (it used a bunch of 555 chips), explained to George why my approach was superior, and was left alone for a while.

Later, when the thing was put together and worked, he remarked (not to me but to someone who told me) that he was amazed that I had simply designed something on paper --- and it had worked!

One of the designs that he had gleaned from the pages of Graf was what he affectionately referred to as a "standard flip-flop". It was deployed extensively in products and as well became the core of what he made for a friend as a little side business, after he retired, a box of relays and relay drivers that simply toggled on and off in response to a remote pushbutton closure. Problem was the flip-flop was hardly standard, although it could be seen distantly as an approximation to a D flip-flop, where the capacitances internally that make them work are rarely shown explicitly. In any event, as the installations using these boxes had longer wire runs, problems arose. I pleaded with him to let me design a robust solution. Nope. Wouldn't hear of it. So I had to keep devising patches to bail him out.

[dvv]

Is it all the same WHERE inisde of an amp is most distortion produced?

Logically, since everything before the final output stage runs in class A, most distortion should be produced by the power transistors, but I wonder is it always just so?

A poorly judged second stage of the voltage amp, for example, is capable of generating some not so obvious levels of distortion you don't even notice in classic static measurements. Everybody usually gets the voltage right, stick in a 250 or 350 V transistor and you're done, but what about its bias current? I have seen examples of admittedly inustrial work in which the signal required say 5 mA, but the transistor was biased at only 3.5 mA.

How soon, or how too late, has the response of the voltage stages been rolled off in relation to the capabilities of the current amplification part of the amp?

Just how REALLY constant is the loading of the voltage gain stages? Most people do without, but since Otala, Harman/Kardon for example has always used resistors to swamp the natural impedance of the current gain amp, which makes their amps rather free of any funnies exercised by the load - just as Otala did way back in 1976. As a result of that, and of other little things, their humble integrated amps rated at 85W/8 Oms will produce peaks (t=40 mS) of over 500 Watts into 1 Ohm and stay perfectly stable. With no inductor parallelled by a resistor at the output (e.g. model HK 680, overall NFB just 12 dB). Krell does the same in their FPB series at least.

[abraxalito]

Originally Posted by bcarso

With semis you may have a point. I'm thinking about systems like amplifiers, DVD players, etc. The redundancy of chassis, connectors, etc. and the inability to judiciously take advantage of a output on board A for a signal needed by board B almost inevitably leads to suboptimal designs.

I think in analog terms its far more often the case your point is true than in digital. Even so when I was at a console company they had a hybrid module for doing line driving duties from what I recall. Perhaps there was also an input balancing module, I can't remember. In conceiving of my modules, they're digital.

But how did it sound (let alone how much does it cost)? The answer:

Good enough. At least in the minds of the managers. And of course for marketing, Hey! It's gotta be great --- it's DIGITAL!

I think the halo has probably worn off digital a bit now. In my experience mass-market digital stuff sucks sound-wise, probably for the reason that you hinted at - that the analog bits have been done by digital designers, the analog ones long having departed (or retired). After all, its only app notes isn't it?

[Wavebourn]

Where most distortions are produced? The answer is simple: where we need more of power amplification. 2 major places: mic pres and power amps.

[dvv]

Originally Posted by tvrgeek

That's what they learned in business school. Is the Troy Institute For Maneuverability still around? They taught us massive lessons on cost management. I look at the mechanical assembly of even mass produced products and just shake my head. Keeps people employed I guess. Little screws of different size. A trained engineer could save some of these companies a lot of money. Hint: If you only need to put round pegs in round holes, you can't screw up.

Wanna bet?

Remember Einstein:

Only two things have no limits - space and human stupidity. I'm not sure about space.

[dvv]

Originally Posted by Wavebourn

Where most distortions are produced? The answer is simple: where we need more of power amplification. 2 major places: mic pres and power amps.

I meant INSIDE a power amp.

[bcarso]

Originally Posted by wahab

Only if the reference signal is a pure sine , then adding
0.1% of H2 can be audible but if the reference signal
is a musical one it s doubtfull that implementing the same H2
ratio would have any audibility.

Yes, hearing low-order HD with pure tones is tough. But how about the difference tone resulting from, say, 19kHz and 20kHz attendant on second harmonic-generating nonlinearities? And how about the low frequency (really d.c., but always limited by some electronic or physical highpass in the system) energy associated with 2nd?

Certainly higher-order distortions get more audible and weirder. But I do get a little tired of reading, accompanying the residual plots in the Stereophile sidebars, the commentary that so-and-so's distortion is predominately of the 2nd and 3rd, and thus subjectively benign.

[Wavebourn]

Originally Posted by dvv

I meant INSIDE a power amp.

The same answer.

[Wavebourn]

Originally Posted by bcarso

Yes, hearing low-order HD with pure tones is tough. But how about the difference tone resulting from, say, 19kHz and 20kHz attendant on second harmonic-generating nonlinearities? And how about the low frequency (really d.c., but always limited by some electronic or physical highpass in the system) energy associated with 2nd?

When character of electronic distortions is similar to distortions mechanical it is hard to tell, is the sound louder, or more distorted. "Own **** smells less".