Sound Quality Vs. Measurements

Status
Not open for further replies.
It beats me ! I did notice in some set ups it sounds boomy if in wrong phase .

If anyone reads the Slew PDF I would be interested if anyone supports using Complimentary Feedback Pairs for the LTP input . The writer of the PDF uses a single input CmfbP if my memory is correct . The design I like is 90% the original Hitachi amp , 10 % me . I like the fact it gives 100 W and near op amp distortion at 50 kHz with only 7 transistor devices ( Hitachi , mine 10 ) . The pair I probably will use are 2SA1085 and 2N2504 . I recently tried a cascode VAS . It was no better and no worse .That is important as suitable transistors are now rare . 2 x 2SD756 + 2 x 2sB716 replaced by 2 x BC550C+ 2 x MPSA 42 + 2 x MPSA 92 . If was fantastically stable .

Have I not been clear, Nige?

Transistors are just as subject to fads and fashon as anything else. Over the decades, the vigorous Japanese semiconductor industry has reached a level of brainwashing that one is led to believe an amp without their products is simply almost not possible.

Nonsense.

Of course they have some very good semicondictors, and anyone disbelieving that should take another hard look at what's being offered. But by the same token, many have never encountered, or have forgotten, that quite a few non-Japanese manufacturers have as much, of not mor, to offer as well.

I realize the European BC series is badly suited to Amerian needs because they are mostly lower voltage, up to about 80V (BC 639/640), and the really low noise models don't go higher than 45V (BC 550/560), or 65V at best (BC 546/556). But where they stop, Motorola/ON Semi takes over.

How does it work? Well, let me just say this - around 80% of Krell amps are made using Motorola/ON Semi MPS series, there are over 120 of them in the voltage gain stage of EACH channel. And the rest is also by the same team, MJE 15030/15031 as predrivers, MJL series as drivers and output devices.

The voltage reg and output stages use custom MJL transistors, which are simply their well known output devices repackaged from plastic to metal TO-3 cans for more power.

High quality, high power devices have ALWAYS been America's forte, and specifically Motorola's forte. The Japanese never caught up, and Europe made only a few attempts worth mentioning over the decades, of which most are no longer being made.
 
Not really . When the Hitachi amp design came along the transistors were specially made to do the jobs . By combining BC550C and MPSA 42 I can get close . As a cascode the 45 V is not a problem used with MPSA 42 . Look at 2SD 765 ( PDF below ) . If is a bit like a very good 2N5551 . Then look at specifics . It more resembles the BC550 C yet in A version is good for 140 V . Cob is 1.6pF . Gain is > 250 ( 400 typical ) . 0.75W

2SA1085 is an interesting device . 0.5 nV noise , 120 V 0.4 W and high gain .

http://www.datasheetcatalog.org/datasheet/HitachiSemiconductor/mXwuuwq.pdf
 
Not really . When the Hitachi amp design came along the transistors were specially made to do the jobs . By combining BC550C and MPSA 42 I can get close . As a cascode the 45 V is not a problem used with MPSA 42 . Look at 2SD 765 ( PDF below ) . If is a bit like a very good 2N5551 . Then look at specifics . It more resembles the BC550 C yet in A version is good for 140 V . Cob is 1.6pF . Gain is > 250 ( 400 typical ) . 0.75W

2SA1085 is an interesting device . 0.5 nV noise , 120 V 0.4 W and high gain .

http://www.datasheetcatalog.org/datasheet/HitachiSemiconductor/mXwuuwq.pdf

But Nige, these are all small signal devces, not power trannies.

And while you seem convinced of MOSFET's qualities as an audio powwer output device, frankly, I am not. They all have problems with large currents, another case of excellent Data Sheet but not nearly so excellent real life perofmrance. True, the latest generations do much better, and the prices have gone down, I suspect mostly because of IRF policies of cheap devices.

As for behaving "like a very good xxx ...", pray tell why would use something that behaves like when I can use the original? As a matter of principle, not this one case?
 
The power transistor are nothing of the rest is less that it should be . Although I have never seen the 2SD756 and 2SB716 used with bipolar outputs I would expect them to be superb . If a complimentary feedback pair was made with your favourite Motorola outputs and drivers with suitable Vbe bias adjuster if should be out of this world good . My reading leads me to think in term of power delivery , current and purity at high frequencies it should be unbeatable . I like MOSFET's as they do 90% what you want without hassle . If matched sets are bought which is automatic if buying at trade price 25 pieces then the current delivery can be excellent using multiple pairs . Ron 0.8 R , absolutely no requirement for source resistors if matched . By the time you get to 4 devices the difference between bipolar and MOSFET is small . I note you use 0R47 so that is not far away .

As far as I know I have most of the 2SD756 that exist in the world . They are like 300 B 1941 stock
 
Last edited:
The power transistor are nothing of the rest is less that it should be . Although I have never seen the 2SD756 and 2SB716 used with bipolar outputs I would expect them to be superb . If a complimentary feedback pair was made with your favourite Motorola outputs and drivers with suitable Vbe bias adjuster if should be out of this world good . My reading leads me to think in term of power delivery , current and purity at high frequencies it should be unbeatable . I like MOSFET's as they do 90% what you want without hassle . If matched sets are bought which is automatic if buying at trade price 25 pieces then the current delivery can be excellent using multiple pairs . Ron 0.8 R , absolutely no requirement for source resistors if matched . By the time you get to 4 devices the difference between bipolar and MOSFET is small . I note you use 0R47 so that is not far away .

As far as I know I have most of the 2SD756 that exist in the world . They are like 300 B 1941 stock

Agreed on the view that power transistors are not much good without a good driver and possible a predriver.

You are wrong in my case, I haven't used a 0.47 Ohm emitter resitor in decades. I typically use 0.33 Ohms, Motorola MPC71 or MPC72 by default, and sometimes, 0.27 or even 0.22 Ohms. I currently have around 60 pcs of MPC71 0.33 Ohms on stock.

This preference is purely practical. They are not wirewound but use metal oxyde, and are hence not inductive to any significan degree, they are available in a wide range in 5W (MPC71) or 7W(MPC72) versions, unlike most other similar offerings they are rated at 5% (most others 10%) tolerance, and experience shows they will handle more than their power rating. In the last 30 years, not one has ever failed me.
 
diyAudio Member RIP
Joined 2005
What is missing from Keith's comment is how was this wide bandwidth achieved. Was is through inhrenetly wide bandwidth design, with an open loop full power bandwidth of say 100 kHz, then aided by relatively low global NFB of say 20-26 dB, a mix of more modest open loop bandwidth and more global NFB, or the traditional way, with lots of global NFB?

The only Spectral schematic I've ever seen a fragment of is a very early preamp, in one of John Linsley-Hood's books. But I recall some remarks Keith made about negative feedback some years ago in an interview and then some elsewhere, and he mentioned a relatively modest amount of global feedback, of the amount you tend to prefer, and as well that the input devices were small-geometry JFETs. He said the latter were slightly higher noise but permitted the amplifier to be faster, indeed specifically faster-settling.
 
diyAudio Member RIP
Joined 2005
Another remark or two about settling time: except for the nonrealizable totally flat response with frequency, the optimal open-loop response for fast settling to high accuracy is an integrator, that is, a response with a 6dB/octave falloff with frequency; for settling to less than perfect accuracy an amp with flat gain to the 6dB rolloff is the first compromise.

Real amps often approximate this in ways that are not exactly ideal in this respect. Some of the more successful opamps like the JE990 depart from it to get other advantages in distortion at high frequencies. Scott Wurcer remarked that the split compensation shown by Self and others, for enclosing the output stage within the pole-splitting network of the second stage (colloquially the "VAS", although Putzeys agitates to change this nomenclature), is shunned at ADI because of the deleterious effects on settling time.

Aside from anecdotal reports, with which we must include KOJ's (although I would personally give him a lot of credence :) ), I don't know of any controlled tests that discriminate among amps with different settling behavior.

It sounds like, from others' recent remarks, as if Bruno is exploring considerable departures from canonical fast settling singularity patterns of late, and must certainly be doing so in the belief that it is advantageous for audio.
 
diyAudio Member RIP
Joined 2005
given what is known about musical signals, audio transducers, human hearing it really doesn't seem likely that fast settling by ~ MHz gain intercept electronic amplifier time scales is required

I'm inclined to agree, at least as far as my own perceptual efficacies are concerned. But to paraphrase Lloyd Bentsen, I know Keith Johnson, and I'm no Keith Johnson.
 
In my admittedly modest experience, extremely wide bandwidth, say 1 MHz response amps will also tend to overshoot, sometimes badly, and to ring a bit using square waves up to 20 kHz.

I also found out that when such an amp is endowed with an input filter in the 200-300 kHz region, overshoot all but disappears and the ringing is for all practical purposes done away with.

In other words, the Studer/reVox and Sony approach again. Make it as widedbandwidth as you can, then limit its input signal rate of change. A cutoff point of 200 kHz is the bandwidth of 24/196 and of course, 10 times above our limit of hearing. In my case, at 59, more than 10 times.:p

A trade-off is the added phase shift of the input filter. One of the reasons we want wide bandwidth is so that our phase shift, expressed as a square wave tilt at 20 kHz, is as small as possible, and the input filter will degrade it somewhat. By how much remains to be seen/heard, but in my view it's a good trade-off.

Actually, the above may well be one of the reasons why the British audio industry is so commited to bandwidth limiting to 60 or 70 kHz. I feel this is a bit too low, but I can see the reasoning.
 
No , I think you and I are always in the crows nest .

I caught the tail end of discussion about DNA today whilst shaving . I suspect the gentleman talking would rank higher than any of us even allowing for different disciplines etc . He made an interesting statement . He was the one who discovered that a Chimpanzee is 95% the same DNA as us . He had thought 60% as a guess ( how many here would admit to guessing ) . He then went on to say having been asked why people differ between lets say Africa and the so called West he said it can not be in the DNA as that is more or less identical . Thus it must be food and encountering different animals plus farming that makes us different . My instinct is to say if we get everything that describes us from 5% of our DNA then how small does small need to be? My bet is Einstein wasn't so different in DNA ? Hi fi I know for a fact is very small differences with large outcomes . It is very complicated . I do accept that sometimes making a circuit more complex actually reduces the problems . That is the focus .

I think in 1971 Peter Blomley sent a letter to Wireless world about a novel approach to amplifier design . Alas I can not find a reference to it worth showing ( Perhaps D Self ) . From memory it split the amplifier earlier than usual to create more stable class B operation . I think it said the two halves working independently ? The signal was feed if I am correct into a double common base stage fed by two current sources ( current mirrors ) . I think it's virtue was it could not be set up incorrectly as long as set to the standard recipe . Again from memory Mullard used the circuit for their budget priced modules , Even future in my memory that he worked for them ? Anyone ever tried it ?
 
I make a lot of guessess. Not my fault that a fair number of them turns out to pretty correct.

But guessing is invariably linked to personal experience. For example, I'm willing to bet that John's guess would be much more accurate than mine and most of us here simply because he's been at it much longer and much deeper than most of us. He's seen more than most of us, tried more also, all of which served to build his experience to a point where his guess will most often turn out to be right, or very near right.

I think that's quite normal and only to be expected. I have the same thing with my good wife - she's been teaching medicine and doing basic research in her lab for 33 years, she's practically breathing her field (pharmacology). So if you ask her why do you react to a drug in a certain way, chances are 99% that she will give you the right answer, after she asks a few questions. And she may well astound you with a claim or two, which may appear to be totally crazy, but listen to her, try it out and it won't be crazy at all. After 28 years of marriage, I have long stopped wondering or being amazed.

As for DNA, much as we have learnt about it, we are still just scratching the surface. But at least we're on our way.

What you said about the unnamed gentleman, sound awfully like a balanced topology, where you have two standalone amps working in parallel, after a common input stage, usually a dual FET. One side drives the plus side of the load, the other the minus side.

If that's it, then I can send you some Sansui schematics, they were big on it with their Double Diamond DC topology.

Sony's approach is rather interesting. In one of their models from high up on their range list (I forgot the model number, but I should have its Service Diagram, so if anyone wants a peek, just let know), they start out just like a common, regualr amp regarding the voltage gain stage, but then switch to fully balanced for the current gain stage, qwith lots and lots of power transistors.

Not knocking it, but so far, I am quite happy with a classic fully complementary layout. thank you.
 
looked at the link a.wayne posted earlier with Steven Guttenberg and the two other guys (sorry, not a names person).
I frankly believe now that Steve has no clue. he brings absolutely nothing new to the discussion and simply keeps repeating flawed arguments. why they needed to have an 1 hour long talk to reach the surprising conclusion that a FR plot doesn't tell everything, beats me...
well, based on that I conjecture that 500 years from now they'll have similar discussions about why certain brain implants sound better.
 
Sometimes, it takes a long discussion to arrive at well known conclusion.

Here, for example, we had a discussion on why should we use FETs for our input stages. It took like 20 pages for someone to put forth the somewhat obvious argument that two BJTs in differential mode, when given enough bias current and when their gain has been degenerated by emitter resistors to something below 10:1 actually starts to act more like a FET than your typical BJT.

While still not quite the same thing, it's getting awful near to it. And no voltage issues, either.

But it was still fun. :D
 
Hi Mr Push Pull . It took myself and Dvv weeks to assert that maybe medium speed output transistors sound better . That is a radical concept as it sits between the only thing wrong with 2N3055 is it's lack of voltage and go as fast as you like it never will be fast enough . I countered that by saying perhaps we should use complimentary feedback pairs ( CmfbP ) of virtually anything you please but slug them a little bit to mimic the conditions of the slower device .

Douglas Self says the EF output has the advantage if done correctly to empty the output devices of minority carrier charges . MOSFET's do not have them so the problem goes away . I fancy that the Blomley amplifier can better use the CmfbP ?

There is a very primitive kit sold by Velleman . The output devices ( TIP 142 /7 perhaps ) are slugged with a capacitor each . One critic said obviously a bad design to need that . We don't say that of the VAS cap usually do we ? In fact Self says get Cob as low as you can and make the VAS cap as high quality as you can . As he says the external cap is bound to be better than what a BD139 gives you gratis . Some say that single input transistor stages are more stable . Can't say I noticed that given a low Cob VAS .

Einsteins brain was 1.2 Kg , most are 1.4 . There is a daft body to brain mass ratio to express something ? Glad in Audio we are not that badly unknowing .

I see DVV and me as two salty old sea dogs out in all weathers keeping this ship going . OK we will never be Captain , our scares and weathered faces say we might be experts in the trade of the sea . Sometimes it gets like the Tale of the ancient mariner , is the Albatross class D , or what I call cake mix amplifiers ( you buy a packet and say you baked a cake ) ? Dvv I hope you did that poem when in Somerset ?
 
Hi Mr Push Pull . It took myself and Dvv weeks to assert that maybe medium speed output transistors sound better . That is a radical concept as it sits between the only thing wrong with 2N3055 is it's lack of voltage and go as fast as you like it never will be fast enough . I countered that by saying perhaps we should use complimentary feedback pairs ( CmfbP ) of virtually anything you please but slug them a little bit to mimic the conditions of the slower device .

Douglas Self says the EF output has the advantage if done correctly to empty the output devices of minority carrier charges . MOSFET's do not have them so the problem goes away . I fancy that the Blomley amplifier can better use the CmfbP ?

There is a very primitive kit sold by Velleman . The output devices ( TIP 142 /7 perhaps ) are slugged with a capacitor each . One critic said obviously a bad design to need that . We don't say that of the VAS cap usually do we ? In fact Self says get Cob as low as you can and make the VAS cap as high quality as you can . As he says the external cap is bound to be better than what a BD139 gives you gratis . Some say that single input transistor stages are more stable . Can't say I noticed that given a low Cob VAS .

Einsteins brain was 1.2 Kg , most are 1.4 . There is a daft body to brain mass ratio to express something ? Glad in Audio we are not that badly unknowing .

I see DVV and me as two salty old sea dogs out in all weathers keeping this ship going . OK we will never be Captain , our scares and weathered faces say we might be experts in the trade of the sea . Sometimes it gets like the Tale of the ancient mariner , is the Albatross class D , or what I call cake mix amplifiers ( you buy a packet and say you baked a cake ) ? Dvv I hope you did that poem when in Somerset ?

DVV is salty ....... :)
 
Hi Mr Push Pull . It took myself and Dvv weeks to assert that maybe medium speed output transistors sound better . That is a radical concept as it sits between the only thing wrong with 2N3055 is it's lack of voltage and go as fast as you like it never will be fast enough .
2N3055... you reminded me of a board collecting dust (clearly visible) in a drawer. I made it in high school, some 17 years ago. the only thing preventing me from binning it are its neat looks, my PCB etching and soldering skills seem to have been superior at 16.
please ignore the two compensation diodes that are obviously supposed to be physically bonded to the output transistors (I only said my soldering skills were good). and the 8-pin device is the poor old 741. based on a 70's Uher amp schematic.
oh and the long cap leads are because original parts were large axial types.
P1040325.jpg
 
Last edited:
Status
Not open for further replies.