First, the second point - I reiterate once more that the differences in sound were not amazing, stunning, large, etc, I leave the biblical proportions to the magaizines. Not even illuminating. But obviously sufficient to be detected, in my view, very reliably, and to be heard as improvement.
Here's how we did it, step by step.
First, we built two bipolar units. They were assembled and static measurements were performed to establish just how similar they are. They turned out to be well neigh udentical, for example, THD 20-20,000 Hz on one was measured as 0.002%, and on the other sligthly more, as the number couldn't make up its mind whether it was 0.002 or 0.003% (BTW, the limit of our test equipment is 0.001%), one's output impedance was 0.2 Ohms and the other's 0.022 Ohms, and so it went. All measurements for both of these, as well as for the FET/MOSFET unit, were made under identical conditions, absolutely nothing was even touched, let alone changed.
It was tested on both our heads with copper wiring first. Then the copper wiring gave way to silver wiring on ONE if the units, and another round of measurements was made, the same as the first one, to make sure all was well and to see if the change would bring about any easily noticeable change. In terms of figures, it did nothing, the numbers remained the same, at least those we did.
Then we tested it all over again, using exactly the same music material as before, and oddly enough, got the same results. My friend was obviously not very comfortable with the Koss cans.
In the next round, we repeated all of the above with the two FET/MOSFET units. After making sure we could not tell the difference between bipolar amp A against the bipolar amp B, nor between FET/MOSFET amp A and B, we were ready for what we aimed for.
First round goes to bipolar A and bipolar B, A with copper and B with silver wiring. We got 10/10 right for me and 9/10 for my friend and associate. The switch box was placed behind the two amps, so as to minimize wiring, which was reduced to just 25 cm (10 inches) for all four RCA to RCA connectors. Even fully turned around, one could not see the box itself, let alone its current settings. To complicate things, switching was done with a randomly chosen number of A/B switches, up to 7 switches per one go and was logged accordingly.
We then repeated the same procedure, in the same manner, with the FET/MOSFET amps. Here, I scored 10/10 again, but my friend slipped from 9/10 to 10/10.
To me, that is absolutely conclusive.
Just so you know, both versions were designed to be happy with as low as 5 Ohms, the impedance of a highly regarded AKG headphone, but just in case, for this test both drove a much more typical load of 30 Ohms. Sennheiser cans are rated at 58 Ohms, and Koss cans are rated at 100 Ohms. Not that it mattered much to them, the bipoalr version uses 50W devices at its output, and the FET/MOSFET version uses 70W output devices. Both run in pure class A for any dynamic headphone known to me currently on the market, their output stages being biased for 90 mA static bias current, and very much requiring their own heat sinks. Both use fully complementary topologies, with a SEPP output stage. Both use 20 dB of overall NFB and an input filter limits their bandwidth to 400 kHz (otherwise, it would just hit 1.2 MHz, and I feel this is unnecessarily high), all figures ref. 7 Vrms output into 30 Ohms.
ALL transistors are hand matched, all resistors are 1% metal film or 0.5% in critical parts. All components originate from well known world class companies (Beyschlag/Vishay, Wima, Fisher & Tausche, Panasonic, Siemens, Elna, Philips, IRF, Motorola/ON Semi, ALPS, etc) and have been purchased from very reliable sources (i.e. the real deal, not Chinese copies). The entire circuit is fed by two shunt type regulators, in turn fed by their own custom toroidal transformers, for a truo dual mono configuration. PCB boards use 70 microns of copper, top side carrying ground mostly.
* * *
Sy, you can voice three sentences in Serbian, all of which would get you shot? I don't think so. While we Serbs are a very hot blooded sort indeed, we can surpise by being incredibly tolerant of anyone even attempting to speak our language. We have to, since our grammar would knock out a brigade of Delta Force people, it's even worse than the German grammar, and most of us natives go through life not really knowing it. With us, it's not only what you say, but even more important is how you say it.
I suppose it's a matter of personal definition to say the glass is half empty, or that the glass is half full, but I think, and you can possibly agree or disagree, that people are basically just people the whole world around. I choose to think of our inevitable differences as of our joint racial heritage being by that much richer; haven't met a culture yet from which I could not learn something good. Or found something similar to my own background, but with an interesting twist of local flavor.
Unforunately, even if I'be been to the USA three times, I never managed to get off the East Coast. Never got to Texas, and I sure would like to, because from what I know of Americans, and having grown up in part among them (albeit among US Аir Force staff and their kids in Ankara, Turkey), I did notice a considerable level of similarity between US Southeners and Sebs, which only raised my curiosity levels. Some day ...
Folks,
While we are on the subject of blind tests for preference (NOTE, I said preference), a few years back we did one I would like for someone to explain.
We tested three options of power supplies for a power amplifier output stage.
The Amplifier Output stage uses a special form of bridge tied load.
This output section has 8pcs Elna Silmic II 1,000uF local PSU capacitors, with a smidgen of inductance and resistance deliberatley designed into the Cables between the Output Board and the main PSU. Changes to this main PSU is what was tested.
FWIW, the cable uses solid core silverplated copper in FEP Teflon, ribbon cable style, it basically star-wires out from the main PSU.
The PSU contains a high power zenner follower regulator with around 100mOhm output impedance under quiescent conditions, falling as output power rises.
The frontend is tube based, so a classic hybrid amp, without global looped feedback.
Three identical production units where pulled from the line and tested. We selected amplifiers with levels less than 0.1dB different (our specifications allow +/-0.2dB deviations).
One unit was unchanged from the production line.
One had the number of the power supply capcitors before the zenner follower doubled using the same capacitors (A Japanese Brand and Series that came out on top of the listening tests for something that could fit the space on the PCB). Doubling the number of capacitors entailed stacking two PSU PCB's on top of each other.
The third Amplifier was instead modified with capacitors from a different Vendor (Rubycon) that allowed the original PCB to be retained by being taller and offered slightly more than twice the capacitance of the originals.
All Amplifiers where tested using AMR's AP2 to have less than 0.1dB level difference and to otherwise measure as identical as can be expected with any reliability. No appreciable differences in frequency response, THD, IMD or any other common metrics where measurable. You could not tell one graph from the other in a fully sighted (20/20) test.
Amplifiers had no identifying marks other than serial numbers on the back, which where not visible to the listening panel.
The Source (CD-Player) and Speakers including all other set-up was completely identical for each presentation. The speakers are fairly high impedance (do not draw huge currents), essentially full range (sub 30Hz to ultrasonic frequencies) and levels where set so the the amplifiers where far from clipping. The listening space was fairly heavily acoustically treated, solid concrete floor and wall, suspended ceiling.
A hidden from sight assistant, who neither knew what was being tested or which amplifier was modified if any switched the cables between amplifiers (Speakon for Speaker, XLR for signal). All three amplifiers where plugged in using identical length mains cables into a star-wired distribution strip next to each other, where on for a considerable time before the test and all for the same amount of time.
The listening panel was unaware what modification was being tested, just that there where three choices and they where expected to express preference ranking for the three choices. The panel consisted of three listeners experienced with both the protocol, the system and the music selected (we have used them many times before).
The panel where also given time to familiarise them again with the systems sound using the "reference" (unmodified) Amplifier, the modified amplifiers where not played during this. They where seated at the same time in the same room, however seating is such that it is not possible to notice or observe another listener without visible efforts.
I was "out of competition" in a position to observe the panel and did not notice anyone peeking. I had an acoustically very bad spot (off to one side) but I also kept score myself. This (listeners in the same room at the same time) is arguably one point where improvements could be made, we do not have this luxury and I notice that is has never stopped the ABX Mafia in their tests.
Presentation was five rounds of three presentation where no repetitions where made (e.g. each presentation featured all three choices in a random sequence - predetermined using Excel's Rand function). Each presentation contained four pieces of music selected to highlight different areas of performance and listners where familiar with these pieces on the tests system from the pre-test session and from prior occasions.
Following the 15 * 4 Trials in total (quite wearying if I may say so) we found the all listeners (myself included) had expressed strong disapproval one Amplifier, identifying it correctly in each and every case, with all music. The assistant from behind the system also complained that this amplifier (he knew which one it was of course - as in what was the serial number and where in the rack was it placed, though not how it was different) also mentioned that he strongly disliked this amplifier.
That was the Amplifier fitted with Rubycon Capacitors.
Between the other two amplifiers the preference was obvious where "heavy bass" pieces where played, the one with more capacitance was preferred, with "Girl & Guitar" music no obvious preference was notable from the limited dataset.
Personally I had been able to quite reliably (17/20 IIRC) pick my "reference" amplifier and had preferred it on all music, despite the poor acoustic position I was in. However, I have NEVER liked Rubicon Electrolytics (including BG's) each time I have tried before them before and I was extremely familiar with the "reference" amplifiers sound, having listened probably 10 hours + each week very critically to this system as part of QC work.
The outcome of this test was that we tested the "double decker" supply with a number of our Distributors (where we thought we could get away with it fully blind). One was truly outraged that of two Amp's he had received one was "so much better" than the other and his own demo unit (you may guess which he considered better) until we fessed up and offered him free upgrade kits for his stock...
Again feedback in what was a mostly blind test (some distributors knew there was a difference) was strongly in favour of modified units. So eventually we upgraded the production amplifiers and offered free upgrades for recent units and "at our cost" ones to earlier machines.
If you liked for me to propose any mechanism or physical process that would explain the outcome (including my own surety picking "my" design and preferring it whereas other clearly did not prefer it) I would have to admit to being clueless.
I really do not know and I did not expect this outcome (I was quite certain that my fairly elaborate PSU regulation and cleverly implemented extra filtering would swamp anything like the double capacitance or capacitor types "into inaudibility", but it did not do so.
I would not generalise beyond the actual individual amplifier design tested, but I suspect there are many lessens hidden in such a small and simple test.
BTW, in my experience differences between mains cables or interconnects of differing construction, R/L/C/G etc. are large than what we experienced in this test, while metallurgy for otherwise identical cables ranks lower.
Ciao T
While we are on the subject of blind tests for preference (NOTE, I said preference), a few years back we did one I would like for someone to explain.
We tested three options of power supplies for a power amplifier output stage.
The Amplifier Output stage uses a special form of bridge tied load.
This output section has 8pcs Elna Silmic II 1,000uF local PSU capacitors, with a smidgen of inductance and resistance deliberatley designed into the Cables between the Output Board and the main PSU. Changes to this main PSU is what was tested.
FWIW, the cable uses solid core silverplated copper in FEP Teflon, ribbon cable style, it basically star-wires out from the main PSU.
The PSU contains a high power zenner follower regulator with around 100mOhm output impedance under quiescent conditions, falling as output power rises.
The frontend is tube based, so a classic hybrid amp, without global looped feedback.
Three identical production units where pulled from the line and tested. We selected amplifiers with levels less than 0.1dB different (our specifications allow +/-0.2dB deviations).
One unit was unchanged from the production line.
One had the number of the power supply capcitors before the zenner follower doubled using the same capacitors (A Japanese Brand and Series that came out on top of the listening tests for something that could fit the space on the PCB). Doubling the number of capacitors entailed stacking two PSU PCB's on top of each other.
The third Amplifier was instead modified with capacitors from a different Vendor (Rubycon) that allowed the original PCB to be retained by being taller and offered slightly more than twice the capacitance of the originals.
All Amplifiers where tested using AMR's AP2 to have less than 0.1dB level difference and to otherwise measure as identical as can be expected with any reliability. No appreciable differences in frequency response, THD, IMD or any other common metrics where measurable. You could not tell one graph from the other in a fully sighted (20/20) test.
Amplifiers had no identifying marks other than serial numbers on the back, which where not visible to the listening panel.
The Source (CD-Player) and Speakers including all other set-up was completely identical for each presentation. The speakers are fairly high impedance (do not draw huge currents), essentially full range (sub 30Hz to ultrasonic frequencies) and levels where set so the the amplifiers where far from clipping. The listening space was fairly heavily acoustically treated, solid concrete floor and wall, suspended ceiling.
A hidden from sight assistant, who neither knew what was being tested or which amplifier was modified if any switched the cables between amplifiers (Speakon for Speaker, XLR for signal). All three amplifiers where plugged in using identical length mains cables into a star-wired distribution strip next to each other, where on for a considerable time before the test and all for the same amount of time.
The listening panel was unaware what modification was being tested, just that there where three choices and they where expected to express preference ranking for the three choices. The panel consisted of three listeners experienced with both the protocol, the system and the music selected (we have used them many times before).
The panel where also given time to familiarise them again with the systems sound using the "reference" (unmodified) Amplifier, the modified amplifiers where not played during this. They where seated at the same time in the same room, however seating is such that it is not possible to notice or observe another listener without visible efforts.
I was "out of competition" in a position to observe the panel and did not notice anyone peeking. I had an acoustically very bad spot (off to one side) but I also kept score myself. This (listeners in the same room at the same time) is arguably one point where improvements could be made, we do not have this luxury and I notice that is has never stopped the ABX Mafia in their tests.
Presentation was five rounds of three presentation where no repetitions where made (e.g. each presentation featured all three choices in a random sequence - predetermined using Excel's Rand function). Each presentation contained four pieces of music selected to highlight different areas of performance and listners where familiar with these pieces on the tests system from the pre-test session and from prior occasions.
Following the 15 * 4 Trials in total (quite wearying if I may say so) we found the all listeners (myself included) had expressed strong disapproval one Amplifier, identifying it correctly in each and every case, with all music. The assistant from behind the system also complained that this amplifier (he knew which one it was of course - as in what was the serial number and where in the rack was it placed, though not how it was different) also mentioned that he strongly disliked this amplifier.
That was the Amplifier fitted with Rubycon Capacitors.
Between the other two amplifiers the preference was obvious where "heavy bass" pieces where played, the one with more capacitance was preferred, with "Girl & Guitar" music no obvious preference was notable from the limited dataset.
Personally I had been able to quite reliably (17/20 IIRC) pick my "reference" amplifier and had preferred it on all music, despite the poor acoustic position I was in. However, I have NEVER liked Rubicon Electrolytics (including BG's) each time I have tried before them before and I was extremely familiar with the "reference" amplifiers sound, having listened probably 10 hours + each week very critically to this system as part of QC work.
The outcome of this test was that we tested the "double decker" supply with a number of our Distributors (where we thought we could get away with it fully blind). One was truly outraged that of two Amp's he had received one was "so much better" than the other and his own demo unit (you may guess which he considered better) until we fessed up and offered him free upgrade kits for his stock...
Again feedback in what was a mostly blind test (some distributors knew there was a difference) was strongly in favour of modified units. So eventually we upgraded the production amplifiers and offered free upgrades for recent units and "at our cost" ones to earlier machines.
If you liked for me to propose any mechanism or physical process that would explain the outcome (including my own surety picking "my" design and preferring it whereas other clearly did not prefer it) I would have to admit to being clueless.
I really do not know and I did not expect this outcome (I was quite certain that my fairly elaborate PSU regulation and cleverly implemented extra filtering would swamp anything like the double capacitance or capacitor types "into inaudibility", but it did not do so.
I would not generalise beyond the actual individual amplifier design tested, but I suspect there are many lessens hidden in such a small and simple test.
BTW, in my experience differences between mains cables or interconnects of differing construction, R/L/C/G etc. are large than what we experienced in this test, while metallurgy for otherwise identical cables ranks lower.
Ciao T
Last edited:
dvv, I really appreciate your effort here. I want to read through this very carefully and ask some follow-up questions. You put in enough effort writing that I don't want to shoot from the hip but give it the careful examination it deserves.
OT: Texas is in the South but it's not very Southern. It's a very distinct and separate culture. Within that, there are even finer regional distinctions. Austin is a high tech/university town, and is well-known as one of the music centers in the US. I've lived in many other places, but never one with as much easy access to great live music as here.
My Serbian sentences involve your mother, your sister, and some rather explicit depilation practices. I was told by the RAAL guys that at least one of the sentences was quite creative- I credit a brilliant Serbian scientist who used to work for me.
OT: Texas is in the South but it's not very Southern. It's a very distinct and separate culture. Within that, there are even finer regional distinctions. Austin is a high tech/university town, and is well-known as one of the music centers in the US. I've lived in many other places, but never one with as much easy access to great live music as here.
My Serbian sentences involve your mother, your sister, and some rather explicit depilation practices. I was told by the RAAL guys that at least one of the sentences was quite creative- I credit a brilliant Serbian scientist who used to work for me.
OK, first question- what was the test format? You talk about A and B, but it's unclear what's being "selected" or "guessed."
edit: What do you mean by "To complicate things, switching was done with a randomly chosen number of A/B switches, up to 7 switches per one go and was logged accordingly"?
edit: What do you mean by "To complicate things, switching was done with a randomly chosen number of A/B switches, up to 7 switches per one go and was logged accordingly"?
How to lie with specifications:
Just browsing around "respectable" manufactures web sites. Noticed a preamp .008%THD at 20Hz. Well, I sure as well hope so. What is it at 20K? 5%? I would hope it is a misprint.
Looking over my old Hafler 110, the input is loaded buy an un-buffered volume control, So I suspect the distortion as published was with the control at max. I may do some simple greening of it too. Simple stuff like emitter degen in the input pairs and maybe ccs. More than that I should start with a much better unit.
Besides these numbers not being what makes a clean unit, you can't believe them anyway.
Just browsing around "respectable" manufactures web sites. Noticed a preamp .008%THD at 20Hz. Well, I sure as well hope so. What is it at 20K? 5%? I would hope it is a misprint.
Looking over my old Hafler 110, the input is loaded buy an un-buffered volume control, So I suspect the distortion as published was with the control at max. I may do some simple greening of it too. Simple stuff like emitter degen in the input pairs and maybe ccs. More than that I should start with a much better unit.
Besides these numbers not being what makes a clean unit, you can't believe them anyway.
They measured a 4.4 uS rising time..
Perhaps an input low pass filter.
Anyway , this completly render useless Stereophile s listening tests
if regular versions have effectively different slew rate..
Goldmund Mimesis 8 power amplifier Measurements | Stereophile.com
Last edited:
The whole idea was to discover whether WE could hear the difference between the two cables at anything resembling better than random. 10/10 and 9/10 went WAY over our own expectations.
The bottom line, although not spoken of nor identified as such, was to select the wiring we though let more music through. Never mind which, I assure you that if the copper wiring turned out better, I would have used it. Frankly, i don't give a hoot which will come out better.
So we checked out for "better" and "worse", having no idea at the time of auditioning which one we were checking for. Only the man switching had a log of which was which, and it was something of a surprise to me when I learnt that I had consistently checked silver as "better". Appearently, my friend was not very surprised, having had extensive experience with silver cables before where I had none.
I mean, come on, 20/20 has just GOT to count for something!
"7 switches in between" means that upon each change, the man would switch ab, ab, ab etc several times so that nobody could follow the "clicks", as we wouldn't know, even if we had tried to count, was that an a->b switch, or a B-B, an A-A or a B->A switch. Absolutely no way to follow. We use very high quality APEM (France) mechanical switches precisely because they make the click, but mostly because these are their professional series toggle switches, as used in pro Studer, Revox, Telefunken, etc gear, with a life expectancy of 1 million switches and a DAMN heavy price attached. But I'd buy them all the same, and in fact I do install them in the current models.
The man doing the switching had to log in which unit was allowed to play a piece of music as per the settings, so only he knew the sequence of units used, until we all sat down and reviewed the results. Looks easy, but actually, it's very tiring work, as wintnessed by the number and rate of growth of the pile of empty beer cans afterwards.
Dvv, I hope that I am allowed to tell you what I think.
I have done comparative tests over the years, and I have heard differences in materials, even when the measurement similarity implies that it should not be there.
Over 30 years ago in an LTE in 'TAA', I wrote about this comparison in response to Dr. Lipshitz, one of the champions of ABX testing, comparing the very new (at the time) 5534 IC and one of my studio board quality discrete circuit boards. I heard the difference and reported it, and of course, as I had not followed ABX protocol, it was dismissed by Dr. L. However, IF I was really prejudiced in the comparison, I would have heard NO DIFFERENCE, because that would have been the cheaper and easier way to build future products. However, with some individuals, you can never get past the opinion that there are few important differences, and therefore you must be imagining things.
Secondly, when it came to differences between silver and copper hookup wire, I made a comparison test of that as well with the Blowtorch preamp.
You see, my business partner died, unexpectedly, leaving me to finish 5 units that were already paid for, for someone in Japan.
Unfortunately, somebody ran off with the spool of silver wire that he used as hookup wire for the previous 40 units, and it was almost impossible to replace, so I looked for the highest quality copper hookup wire that I could find, and we built a unit with it, instead of the silver wire that was traditionally used. Then I had to listen carefully to the final result, and there WAS a difference, but it was small enough to get away with to individual customers, so long as they did not have an original to directly compare to. In fact, it was easier to listen to the copper wire version than the silver wire equivalent. We proceeded to make the next 5 units in this way, and now they are in Japan.
Still, it always nags me that the difference was there. So be it.
In reality, you can never prove your position to someone who simply does not believe that differences of this kind exist. Tests will be demanded, parsing of protocol procedure will ensue, and nobody will be satisfied.
I have done comparative tests over the years, and I have heard differences in materials, even when the measurement similarity implies that it should not be there.
Over 30 years ago in an LTE in 'TAA', I wrote about this comparison in response to Dr. Lipshitz, one of the champions of ABX testing, comparing the very new (at the time) 5534 IC and one of my studio board quality discrete circuit boards. I heard the difference and reported it, and of course, as I had not followed ABX protocol, it was dismissed by Dr. L. However, IF I was really prejudiced in the comparison, I would have heard NO DIFFERENCE, because that would have been the cheaper and easier way to build future products. However, with some individuals, you can never get past the opinion that there are few important differences, and therefore you must be imagining things.
Secondly, when it came to differences between silver and copper hookup wire, I made a comparison test of that as well with the Blowtorch preamp.
You see, my business partner died, unexpectedly, leaving me to finish 5 units that were already paid for, for someone in Japan.
Unfortunately, somebody ran off with the spool of silver wire that he used as hookup wire for the previous 40 units, and it was almost impossible to replace, so I looked for the highest quality copper hookup wire that I could find, and we built a unit with it, instead of the silver wire that was traditionally used. Then I had to listen carefully to the final result, and there WAS a difference, but it was small enough to get away with to individual customers, so long as they did not have an original to directly compare to. In fact, it was easier to listen to the copper wire version than the silver wire equivalent. We proceeded to make the next 5 units in this way, and now they are in Japan.
Still, it always nags me that the difference was there. So be it.
In reality, you can never prove your position to someone who simply does not believe that differences of this kind exist. Tests will be demanded, parsing of protocol procedure will ensue, and nobody will be satisfied.
Last edited:
RAAL, huh? So you know Aleksandar Radosavljevic, aka "Aca", aka "Aca Ribon"?
Hell of a guy, and he's from the eastern part of the country, where they truly excel at explicit language.
Just for the hell of it, ask him how was his trip to Croatia without a passport? He discovered that he had left his passport at home right at the border between Serbia and Croatia, so Solaja (http://www.solajaaudio.com ) and I had to continue without him, and I got his drivers to finalize a sale he had in Croatia.
Like I said, helluva guy, great just to hang out with, and he does make some damn good ribbon speakers.As for my sister and mom, you'd be out of luck, Sy, both are deceased.
And, much unlike my countrymen, I don't insult too easily. Must be the 25% Austrian blood in me.
Last edited:
Indeed, but whether or not that it's because of the silver versus copper is the question. You have to understand that 20 out of 20 is remarkably large in any sensory comparison. Even if I were testing something I knew to be audible, a score like that would have me reviewing my procedures- not just with an engineer, but also with someone who has experience in setting up sensory tests. The biasing can be completely unconscious- the Linear Audio article I referred to gives several examples and may suggest better test setups for you.
There's a lot of uncontrolled variables there, and if you're interested in actually finding out if you and/or your buddy can hear the difference between copper and silver (which is an extraordinary claim), you have almost everything you need to do it. You just need a better test protocol from the sensory aspect- I'm sure your electrical stuff was done with great care.
There are 2 big differences: to evaluate thoroughly if change of a part of a wire can be used by a customer to claim that the device sold is not as as advertised, and the real difference when it is clearly audible that it is a breakthrough, and it sounds like a real band in your room. When water running on record caused my subconscious reaction of jumping and turning around; when my guest said that he wants too a house near a creek because he likes how frogs sing; when my father in law insisted tat somebody played piano in our house; when my guests were scared by roaring lion; all this and other things happened when I did not use special cables, capacitors, triple-more-blind tests. All bits added to reproduction quality are valid, but the end result is always the same: the better is the system, the less it is heard itself.
However, we can use thermometer, X-rays, to check what is broken in human body. Like, designing equipment I always measure how big are things that are heard as unnatural additions, in order to minimize their effects. But such measurements are useless if we want to check how well the person is prepared to perform on Olympic games. The only measurement is, the real competition.
The same with sound system: if it is sick, we must measure certain things, sometimes working as a detective to find what is wrong. But we can't use the same measurements to tell how inaudible will be the whole system.
However, we can use thermometer, X-rays, to check what is broken in human body. Like, designing equipment I always measure how big are things that are heard as unnatural additions, in order to minimize their effects. But such measurements are useless if we want to check how well the person is prepared to perform on Olympic games. The only measurement is, the real competition.
The same with sound system: if it is sick, we must measure certain things, sometimes working as a detective to find what is wrong. But we can't use the same measurements to tell how inaudible will be the whole system.
First, let's be done with the protocol - John, you are welcome to offer your views, whatever they may be, to me at any time. Agree or disagree as you see fit, because even if you disagree, and say why, I am still learning - rest assured that I will at least seriously review my own opinion. I have followed your work for over 35 years, since the early days of Mark Levinson, on to the Parasound, which remains one of my (as yet) unfulfilled wishes. As far as I am concerned, you have "made your bones" several times over and that means I can only benefit from your knowledge. I am Old School, a definitely dying breed in these fast strike, fast buck days of whizz kids, and have been taught to respect solid work.
As for your above text, this is EXACTLY the same situation, just the background is different, and I am one up on you here, because I keep the silver wire under lock and key.
As you said, the difference is small, but it's there. If we used the much cheaper but still excellent copper wire, nobody but us would even know, until some DIYer changed the wire and reported the difference, or some such.
But I would know, and it would nag me to hell and back because I would always think I had not quite done the job, when I so easily could have. And that's something I don't want to live with.
I quite agree with your view that some things you cannot prove even if you yourself are aware that they are there, but even so, it's healthy to get one's *** kicked here and there by even some total sceptic because I think we all tend to drift away a bit, sort of shut ourselves into our own views more than is good for us. Sometimes, this can lead us into very much dead end streets without us even noticing it.
Yes, he is guilty on both counts. A man well worth knowing.
You should meet his dad, whom he takes after - just as impressive. Strong willed, opinionated, overt, but you get the feeling that if he's got your back, you're safe as houses.
If you liked Aca, I suspect you'd love Milan Karan (Karan Acoustics | Manufacturer of High-End Audio ). All 6'6'', 260 lbs of him, if there was a real life Gentle Giant, that would be him. He makes some awesome audio products.
Agreed.
Only in this case, we were not dealing with anything as big as a revelation or some such, simply an improvement. And small as it may be, it gets the device nearer to disappearing.
Anyway, I don't believe in revelations, perhaps because I've never witnessed one and because I don't write for magazines. I believe in slow, steady work, and improvement in small steps, I believe that a working project is never finished, because I think everything can be improved, even if by only a little.
random observations - not specific criticisms
even 0.5% resistor tolerance may not be enough - the worst "worse case" for 2 gain setting networks with 0.5% R is a 2% level mismatch which can be "clearly audible" as a difference but not percieved as a loudness change
level, frequency response matching to better than Clark's ABX threshold curves are considered an absolute necessary condition for anyone from the psychoacoustic science camp to credit subjective discrimination as possibly "interesting"
ABX Amplitude vs. Frequency Matching Criteria
with only 20 dB negative feedback the open loop gain match needs to be better than 10% to even get to the 1%, 0.1 dB level - people seldom match semis for all gain limiting "parasitics" like output conductance
the level, frequency response matching should be verified at the final, post switch electro-acoustic transducer terminals – I assume many won’t have the mics and room/couplers to do actual listening position SPL measurements with good repeatability
better consumer/prosumer soundcards have the resolution, repeatability, if not absolute accuracy or channel matching
it could be illuminating to see if the difference is robust enough to survive hi res digital encoding - maybe Audio DiffMaker can find the difference in the amp’s outputs
I would certainly try it to characterize, eliminate any measurable differences before even considering subjective evaluation
even 0.5% resistor tolerance may not be enough - the worst "worse case" for 2 gain setting networks with 0.5% R is a 2% level mismatch which can be "clearly audible" as a difference but not percieved as a loudness change
level, frequency response matching to better than Clark's ABX threshold curves are considered an absolute necessary condition for anyone from the psychoacoustic science camp to credit subjective discrimination as possibly "interesting"
ABX Amplitude vs. Frequency Matching Criteria
with only 20 dB negative feedback the open loop gain match needs to be better than 10% to even get to the 1%, 0.1 dB level - people seldom match semis for all gain limiting "parasitics" like output conductance
the level, frequency response matching should be verified at the final, post switch electro-acoustic transducer terminals – I assume many won’t have the mics and room/couplers to do actual listening position SPL measurements with good repeatability
better consumer/prosumer soundcards have the resolution, repeatability, if not absolute accuracy or channel matching
it could be illuminating to see if the difference is robust enough to survive hi res digital encoding - maybe Audio DiffMaker can find the difference in the amp’s outputs
I would certainly try it to characterize, eliminate any measurable differences before even considering subjective evaluation
Last edited:
Odd that you should say the above.
For decades, I maintained that gain inbalance in a stereo amp will null an awful lot of spatial information in the signal (assuming it's there, of course). I've been called names not to be mentioned in polite company. Some people are just born naturally stupid, I guess, and there is no cure.
Once, for the hell of it, I installed trimmers as pots on the input of a lowly, some would say Zero Fi, Technics integrated amp, one of those using chip amps. After some fiddling to adjust the output level to within 1 mV of the target signal (thank God it was a multiturn trimmer!), which was 17.9 Vrms into 8 Ohms, or 40 Watts, we (the owner and I) did hear more detail in the same system, under the same conditions as before "the mod", with the "Tone defeat" switch ON, which incidentally bypasses the balance pot as well.
I think a designer is faced with a choice here. One, he can use 0.1% metal film resistors, which are kinda hard to come by and are by no means cheap, but the worst is that in my admittedly limited experience, they deviate from nominal values by a lot more than 0.1%, I've seen even 0.3%. Two, he can just take his chances and use standard 1% resistors, perhaps checking them out by measurement before installatio. And three, he can install a 25 turn trimmer and go for precision adjumstment.
Logically, the third option is the best, but it does have the one obvious caveat - it puts that trimmer slap bang in the middle of the signal path, and multiturn trimmers are not known for their sound quality. But, if you're on a tight budget, I'd go for that option as the most cost-conscious.
Regarding semis, you're quite right in stating people do not match them for anything else but sheer gain. I suppose that's better than nothing, but on the other hand, real transistor analyzers are rather expensive and are simply not readily available to most of us. In other words, I can't afford one, much as I'd love to have it.
If I'm wrong, please, PLEASE let me know.
Obviously, why bother with a mod which does nothing worth mentioning otherwise?
But then, none of us knows how many small steps it will take for the equipemtn to sonically disappear, do we? So we work on, step by step, hoping to reach it relatively soon.
But admit it - it's FUN!
Every little step brings its own rewards.
How much factor can you give that if a change makes an audible difference by reasonable test methods, and you have invested considerable intellectual capital, that the change will be presumed positive when it may not be?
I would think that conformation of an audible difference is the invitation to find an objective measure of the difference. Without objective measurements, we are reduced to art. Not knocking art, but I like engineering too.
I would think that conformation of an audible difference is the invitation to find an objective measure of the difference. Without objective measurements, we are reduced to art. Not knocking art, but I like engineering too.
- Status
- Not open for further replies.