Sound Quality Vs. Measurements

Status
Not open for further replies.
I was thinking abouit this . If we take all of our commersial amps and solder in some nice bits of copper we might do some good . Doubtless the cross-section has some effect on sound ? I would focus on the Ground alone . The output transistors probably are more trouble than the existing amp arrangements so no great need to change that ( 0R22 emitter resistors ) ?

This says something in favour of the amps like LAB 47 . It's 9 mm feedback path is irrelevant as far as I can see . If the distance from it 's Ground is short and low resistance , that makes sense . The argument might be this ? Skin effect probably is important . Ideally 1 mm cross sections ? As I pointed out yesterday even a minute piece of 0.8 mm is loosing 0.03V at 0.2A . Setting aside skin effect we need a compact layout so as to make best use of the copper . Ideally a star point or a well thought out bus-bar . I say a well designed class AB can outperform a class A . In the PSU department it can . The high standing current is said to saturate the PSU ( if class A ) . Having the least standing current should produce less potenial difference in the 0 V rails ( must in fact ) . That will change things and I can only think for the better ? The output stage Dvv uses should promote very low distortion at very low standing currents ( triples , typically 7 mA per pair can be enough ) .

One thing people get wrong in design . Yes something should work . The practicality of getting it to work might have inconvenient consequences . It is a bit like the mid engined car and the boxer engine . We so nearly don't need to do it that we pretend it is a caprice . To work on a car is another matter or pay for the repairs when someone does . The front engined rear gearbox is intelligent . I heard Clarkson say rear wheel drive costs more to make and that's why BMW is better because they pay the extra . Rubbish . It will be more expensive now as it is less widely used . Initially the front wheel drive cost more as it had more difficult engineering . Clarkson forgets it was to get maximum interior space that dictated the change . Some argue BMW stayed with rear drive because it made them more money and was not without some merit . The money was not to retool and build a niche increasingly made vacant . They stayed with class AB if you like ?

One of the most inconvenient things about space rockets is the difference between weight of fuel and required thrust is a delicate balance ( almost not possible ) . Alan Bond ( and Val Cleaver et al ) in the UK solved it a bit . His early rockets were like a Coke can . It was stressed by the fuel . Problem is it was useless for a defense rocket as it took 30 minutes to fill and was stored empty . Shame we were so narrow as for commercial rockets it was ideal . The UK government always has a big supply of stupid pills to take every day . The latest Bond adventure might put some who read this into space one day . What is wrong with the world , penniless Britain of 1947 put jets into aircraft , now we can't even build locomotives . We had a saying here " the last one to leave the country please remember to switch the light off " . Irony is people flock to come here !
 
The beauty of this is we might have it done in 20 minutes . The problem with PCB's is they are almost 2 D . Dead-bug is no big deal . I have some offcuts of Nordost 1.7 mm 60 micron silver plated OFC . That seems Ideal. Do a before and after for voltages .

In the UK we call that 2.5 sq mm twin and earth for electricity . It costs pennies . The silver helps a bit . I think it is 4 N's if Nordost .

( d x d x 22 ) /28 ( is OK for that ) (1.7 x 1.7 x 22 ) / 28 = 2.27 sq mm

22/ 7 = Pi 22/28 is related to Pi d sq/4
 
Last edited:
Strictly speaking for myself, I find that no two models can be treated the same. Each is a story unto itself.

Changing the output transistors is a can of worms - no sense in doing that unless you also redo the protection circuits. If not, the amp will continue to behave as if you've done nothing. Your only possible gain might be in thermal behavior, given that more powerful transistors have a different derating curve.

What I find makes the most difference per buck is putting in some high or higher quality electrolytics, also adding some size as well. For example, take out the 15,000uF caps and put in some 22,000uF, preferably of known good quality. If nothing else, you will get a better controlled bass.

Another thing which often works well (as for example in Marantz' 7000 series integrated amps) is to change the mundane NJR op amps with some quality fare from say Analog Devices. It makes the sound more transparent, been done often enough.

But fiddling around with ground is, in my view, a highly risky business. If they got it wrong in the first go, you will be spitting blood by the time you redo it the right way. You MIGHT play around with an existing potential equalizing resistor value, if you've just changed the caps, it might shift a value or two up or down, say from 15 Ohms to 10-20 Ohms, something like that.

You might wish to change its internal wiring from original copper to say pure silver, or high grade OFC copper, but that one you have to play by ear.

Or, change a weedy little wire from the amp output to the speaker binding post with something more substantial of known good quality. On occasion, this can audibly change the sound for the better, at least for more substantial, with more gravitas.

Or, as in case of my own HK 6550, you might throw away the crappy volume pot and install an ALPS Blue in its place. This produced literally stunning difference in my case.
 
It's 9 mm feedback path is irrelevant as far as I can see . If the distance from it 's Ground is short and low resistance , that makes sense . The argument might be this ? Skin effect probably is important
IME the feedback path is extremely critical, as is grounding and power supplies. Skin effect to me is one of those "fashion" fads, I never once contemplated that being relevant.

To me, it's looking for the real weaknesses, and fixing 'em. I'm not interested in a 10% improvement, I want a 10 x step up in behaviour in the areas of weakness - this I believe is where the real gains are made, the ones that count, that one's ears are particularly sensitive to ...

Frank
 
hi guys an occasional reader/poster on here came here after a search on Perreaux brought this thread up. noticed it was an interesting title.
I have come across a guy who has a PMF5150B a MOSFET Power amp long story short I checked Perreaux schematics which I have on file but never used in anger.

they make some serious and significant claims on their site. checked the drawings...

It is the "Hitachi" app circuitwith a tiny variation but it is often very liked and applauded for sonic quality. so 5 small signal BJT's driving several pairs of comp MOSFETS is how it done eh???

just goes to show sonics are subjective and cumulative depending on what else is in the chain personal taste also counts in this equation.
But the Perreaux guys must be "generous" with truth too, as they claim to be the 1st "World company" to use MOSFET's perhaps they invented them then and not Hitachi or in indeed loosely Sony (for their Vertical FET's
in a nutshell specs can mean little if it dont sound "nice" or "musical" or just plain how you like it!!!!
 
...
But the Perreaux guys must be "generous" with truth too, as they claim to be the 1st "World company" to use MOSFET's perhaps they invented them then and not Hitachi or in indeed loosely Sony (for their Vertical FET's
in a nutshell specs can mean little if it dont sound "nice" or "musical" or just plain how you like it!!!!

That particular version of MOSFETs perhaps, but if memory serves, Hitachi was the first on the market with MOSFETs in thei HCA/HMA pre and main amps some years before anyone had ever even heard of Perraux outside New Zealand.

That said, and repeating that I am not a great fan of MOSFETs, I have to admit Perraux's 100 WPC MOSFET amp was one of the few such amps I really liked. Heard it in London in the early 80ies. No idea why, it just sounded "right" to me driving big KEF speakers in a deačer's showroom. They sort of "clicked together".

As for "being generous with the truth", aren't they all?
 
IME the feedback path is extremely critical, as is grounding and power supplies. Skin effect to me is one of those "fashion" fads, I never once contemplated that being relevant.

To me, it's looking for the real weaknesses, and fixing 'em. I'm not interested in a 10% improvement, I want a 10 x step up in behaviour in the areas of weakness - this I believe is where the real gains are made, the ones that count, that one's ears are particularly sensitive to ...

Frank

But it all adds up, Frank, 5% there and 10% here, in the end, you might wind up with a significantly improved sound.
 
yes fully aware of Hitachi's entry into comp market and HMA's.. just Perreaux's website regarding their "Histoty" refutes this claiming to be "First"( suggest a read made me laugh) but as an experimental designer as one does when studying I used the Sony devices first though not A MOSFET just a FET of the vertical type way before 2SK134/J48's.
My point if Perreuax will tell "Porkies" about being "1st" what other economies do the give or any other maker for that matter still even if the "Specs" are real or made up by the Marketing dept it has little or nothing to do with the spec in this case just Subjective opinion or perception which IMO is all that countsI rent and install Pro-audio not one item gets on my stock list unless I like it regardless of specs. a lot of kit has arguably good or excellent specs but sounds rubbish. I can remember talking to amongst others Bob Carver about his reseach into "Transfer Function" which has very little to do with measured results we would usually associate with performance the same topology with different parameters but "measuring" virtually the same sounding so very different.
Specs mean little about real-world performance to the ear hence your liking the Perreauh/KEF combo using a basic 5 BJT consisting of only a differential pair a current source and a curent mirror driving that bank of MOSFETSwith virtually no stablising of the driver part running on unregulated rails which would usually give a "bad" spec. ...So measured specs or sonics er toss a coin!!!
 
My memory says it was Sony who used V-FET's ( TA 5650 was a domestic version and very nice ) . MOS FET's v V FET's . One can always find unique selling points . Mr H C Lin patent for these FET's is dated 1969 ( Silver Springs MA ) .

Feedback paths being 9 mm was the critical parameter I referred to . I can not imagine the speed of light comes into this ? That is I suspect the belief system that is being promoted here ? If I am wrong let me know . If you say it gets other things right I did already say that . The way feedback is arranged is critical . the spectrum analyzer will say that . Doing all the right things for all the wrong explanations . I do this sometimes . I discover something and keep my mouth shut . It is then promoted using existing concepts .
 
Feedback paths being 9 mm was the critical parameter I referred to . I can not imagine the speed of light comes into this ? That is I suspect the belief system that is being promoted here ? If I am wrong let me know . If you say it gets other things right I did already say that . The way feedback is arranged is critical . the spectrum analyzer will say that . Doing all the right things for all the wrong explanations
The vagaries of electronics come into it - an extremely short path means that stray inductance, capacitance and RF pickup is minimised, the reason why SMD is good for really high speed and sensitive circuitry. My DIY gainclone used about the same distance, for the same reasons.

If someone wants to come with hokey reasons, for marketing, then that's their call ...

@dvv: yes, everything counts, but I look at audio in a subtractive, not additive, light. To me, every system is really "perfect", on the "inside", but to get that happening the problems have to be subtracted. Find the worst problem, subtract that effect by a fix or workaround, find the next, etc; continue until the SQ is acceptable. I find I get, for me, dramatic improvements with that approach.

Frank
 
Frank , thanks for that . I want to get as much discussion on this as I can . I feel the short feedback path is important . Short in itself may be incidental ( resistance or time effects that is ) ?

Very interested in the supposed emergence of upper harmonics if using greater amounts of feedback . The expression all other things being equal springs to mind . And they never are in real life . An amplifier happy with 20 dB of feedback may not be so happy at 40 dB . Already the pattern of distortion changes due to action require for stability . I was also interested in the effects of noise . Could it be that any distortion which is in noise effectively doesn't exist ? That is like dither it has asked the transistor to make a choice ? We are talking crossover distortion if so . Other distortions do not interest me as much . Usually these are well under control .
 
You don't necessarily need thick metal, except when your design assumes superconductors everywhere (like Spice does?). If you assume every connection has some resistance and inductance, and design accordingly, then you can use ordinary wires. So the bigger improvement you get by scattering copper, the worse was the original design.

I will drink to that . I guess some extra copper in the 0V connections usually will do no harm ? It will cost nothing in terms of materials and time . Keeping standing current down is useful ? Or maybe not . If reveled at higher current it might be that alone that alerts someone to a problem . If so turn the bias up and have a measure . Turn it down latter .
 
Nige,

By definition, shorter signal paths are welcome, no matter what you are making, and with which components. Obviously, this also applies to the feedback path. However, that may not always be practical to do; it's like sorting things in a cupboard, not everything can be up front, something has to be down below.

Which means that planning your outlay is, in my view, by far the most important single item influencing the end result.

Folding a circuit - having a part of the circuit on a verically mounted separate board, its pins being soldered down in the "motherboard", is sometimes a good way to go to keep everything compact.

Another aspect is to plan for double sided PCB, with the top side having a lot of ground plane on it. Ask the RF boys, they excell at it because they have to. That also allows you for wider traces for the rest.

Lastly, what about the actual glass fibre boards? The industry standard is 30-35 microns of copper, and it's known to work. You can move n up to the 50 micron class, or like me, go up to the 70 micron class. 70 micron copper is about what you can buy easily, the 120 micron ans 150 micron are very hard to come by, and are damn expensive.

The killer end move is having all your through holes metalized. That is about as good a contact as you can hope for without resorting to NASA technology. It's not very expensive.

Before soldering, clean every contact leg of whatever with some cotton and medicinal alcohol. I know they look shiny and clean, but in fact may not be quite so. If you want to go bonkers extreme, sandpaper each contact leg with very fine sandpaper, clean up with said alcohol and gently spead solder from top to bottom BEFORE soldering on anything. That should if not stop corrosion, then at least that will delay it by a couple of years and give you the best possible solder point ever.

Obviously, you will be doing that only on the final form of whatever. It takes a lot of time.
 
Last edited:
My memory says it was Sony who used V-FET's ( TA 5650 was a domestic version and very nice ) . MOS FET's v V FET's . One can always find unique selling points . Mr H C Lin patent for these FET's is dated 1969 ( Silver Springs MA ) ...

.

In 1975, Sony introduced their two integrated VFET amplifiers, models TA-8650 and TA-5650.

In the same season, Yamaha introduced their C1 and B1 pre and power amp, also based on VFETs.

It's a moot point who was actualy THE first, as both appeared in the same season, based on completely indepndent research.

The rise of Hitachi's MOSFEts marked the first truly economically viable power devices, which were found to be quite reliable and easier to work with that previous versions from other sources. First models were the 6500 series, which lived on until 1989 in different guises and were much more elaborate than the original version.

Unfortunately, their versions of VFETs were very expensive to manufacture, and both suffered from reliability issues for the first year of production or so.
 
nigel pearson said:
I guess some extra copper in the 0V connections usually will do no harm ?
Adding copper could couple circuits which should be separate, if a Tee junction gets too broad so currents flow in a slightly different way. As in other branches of engineering, there is no magic bullet. To modify a circuit successfully, you generally have to be smarter than the original designer so you can spot his mistakes/economies and improve them. If you are smarter than him, why not design your own circuit?
 
I'm with dvv on where the real gains can be made -- I have only a low level interest in topologies; whether MOSFETs or BJTs, or TMC or TPC, etc, is "better" - these are all in the area of the 5 or 10% gains I mentioned before. It's about being really fussy about getting what you've finally decided on using to really work properly, that's where the major action is, as far as I'm concerned.

Technically brilliant circuits can sound awful; mid-fi "muck" can take your breath away if they're properly tuned -- it's as simple as that ...

Frank
 
Status
Not open for further replies.