Well, those weren't the problems I was thinking of - the angle I was always coming from, what I was concerned with, were the factors that affected subjective sound qualities - reliability, heat production was not being considered at all. Obviously if one was going to thoroughly re-engineer the unit the type of modifications you mention would be on the table - but my chase was for the elusive factors that delivered the sound I was trying to reliably extract.
Since the unit in raw form managed the trick once, and did it repeatedly many times following, I needed to be careful in the process of tweaking - don't want to throw out the baby with ...
I'm not sorry I went about it the way I did; this was purely hobby for me, nothing in my work area related to audio engineering - I was feeling my way forward, based on the best information to hand at the time.
Since the unit in raw form managed the trick once, and did it repeatedly many times following, I needed to be careful in the process of tweaking - don't want to throw out the baby with ...
I'm not sorry I went about it the way I did; this was purely hobby for me, nothing in my work area related to audio engineering - I was feeling my way forward, based on the best information to hand at the time.
It is a 50-50 guess, only if one had no idea of any difference which to choose and is guessing.
I am guessing only about Y being without R6 from this very modest reproduction system here on my desk. What I am not guessing about is that there is a difference heard between the two. X is more compressed as well.
My guess is somewhat based on the apparently clearer sound (assuming volume levels were the same... (which they might not be... havent checked) of Y and that might be from lower distortion - thus higher Z is known to produce lower distortion..... so, sort of an educated guess.
I have heard some argue that you cannot hear differences when the system distortion is much higher - it will mask the lower level distortion and this system must be pretty high in distortion (speakers in particular). Doesnt seem to apply here. So, whats up with that, now?
-RM
Last edited:
Eeh, but that's also what others did, Frank.
For example by going to an audio show, pick up a Perreaux brochure, peep inside and feel how hot the heatsinks were.
Then calculate at home how high the dissipation of that amp must be, how hot the dies of the Hitachi's had to be, how heavy the transformer was loaded and how much the rail voltages must have sagged.
Without a schematic, just a heatsink catalogue and a calculator.
Means I've got a Perreaux SM/PMF folder that's closing in on 30 years.
(helps if one had thermodynamics classes at university, at a somewhat higher level than the average EE)
Jacco, obviously you get great pleasure from having a deep knowledge of audio amplifier technology and history, and that as an interest in itself is as worthwhile as anything else - but that's not where I'm coming from. The unit is merely a means to an end for me, grist to the mill - I'm not interested in producing a finely engineered amplifier, I'm interested in achieving convincing sound; something competently engineered in every possible area obviously helps in the task, but it's not my first priority.
My amp is running at 42degC (I do not care about current, I set my workshop temperature to 20degC and then adjust the bias for 42degC) there are 8 power devices and the heatsink is SK479/200/SA. The question is: what is the idle current if the rail voltage is +-72Vdc?
Was that one of the questions answered in your theses?
P.s. My amp realy runs at 42degC
see all of GEDLEE's papers. This one gives more detailed facts.... perception of distortion:
View attachment The Perception of Distortion.pdf
-RNM
P.s. My amp realy runs at 42degC
Higher level thermodynamics
That heat sink is 0.28K/W (from manuf. catalog).
Thus 22K temp rise needs a feed of 78W
78W/(2*72V)=0.54A
0.54A/8=68mA per device
This is a fast and rough calculation.
Due to other thermal resistances and parasitic thermal flows, the actual current per device will be a bit higher.
George
True, I'm a toilet paper scribbler, worse than you can imagine.
But it does show quickly that the Perreaux designer was a bit of an amateur, at least up to the early 1990s, and what the primary flaws of the PMF amps were.
Much more efficient than guesstimating, or getting in touch with your Shamanism (@ Jason V. Serenasius)
I'm using the -120db distortion class A amp Mr Macura has referred to a couple of times, best I've lived with yet.
I don't give a rats azz whether it's discrete or opamp, SS or valve, global NFB or non-NFB, or carved in solid gold. (well, tacky does cross a boundary)
When I accidently drop a knife, I often drop it again to re-experience that sound. I easily loose half an hour each day, due to listening to real sounds.
Getting nearer to that sound is what it's about.
You must confuse me with an EE (not enough pay)
(for one heatsink and 8 devices, quicky toilet paper suggests around 100mA, depends on the thermal coupling to the chassis, Fransman. The Perreaux wipe said that Hitachi devices in those amps ran too hot, apart from crossing the acceptable Vds border)
There are many conditions that I did not consider. The 68mA is the minimum possible. Your 100mA is very probable.
?
I assumed pure classA operation at idle.
Feel free to use your red pen
George
Jacco may be be making very valid points about design aspects of the Perreaux, but once I had realised what it was capable of delivering in SQ - if the stars were in alignment - I became intensely curious whether other, "ambitious" amplifiers showed similar potential. I made the rounds of every high end store I was aware of, with that Status Quo CD clutched in my hot little hands, to see how their merchandise shaped up ... and the news was all bad: irrespective of brand and status they all fell short, sometimes by very large margins - something that surprised me greatly at the time ...
So, I got lucky when I bought the unit - enough of the right potential was there, close enough to the surface ... just a little bit of scratching brought it into the light of day ...
So, I got lucky when I bought the unit - enough of the right potential was there, close enough to the surface ... just a little bit of scratching brought it into the light of day ...
There are many conditions that I did not consider. The 68mA is the minimum possible. Your 100mA is very probable.
?
I assumed pure classA operation at idle.
Feel free to use your red pen
George
So close, Jacco was right, the total idle current is 400mA, the amplifier power devices consume 75mA each, the 15V-shunt for the input buffer, servo, rectifier for the VU-meter and some other house keeping uses the other 100mA's
dW/dT is not constant for each ΔT above ambient.
(didn't mean to imply that EE's are less bright, but thermo for an EE curriculum doesn't have to go as deep as e.g. a mechatronics engineer as EUVL had)
Last edited:
John, i'm on your side on those.
Working on slew rate and "simple is beautiful" since my first steps in audio design (sometime around the 70th ;-)
I believe there is two schools (i tried both).
One consist to use complex circuitry, in order to minimize distortions. With the adds of CCS, Current mirors, Diamonds and super pairs, various kind of Cascodes everywhere.
Working on paper and simulations.
Oh, i believe it produces very good distortion numbers, very good for instrumentation, but don't produce always so much good listening results. Loose of dynamic little details, harshness etc...
Notice that, all this complexity works at the detriment of the high slew rate and stability, witch is, for me, the most important thing, as long as we are in a feedback loop configuration. Poles and their phases rock'n'roll are our enemies.
The other school consist to chose the most simple, but clever solution we can, and to optimize-it at his best. Correlating each advancement with listening experience. At the end, it can look trivial, looking at the schematic. Oh, all those hours spend on the bench table, to reach this result.
Sometimes, i don't agree with you, like silver oriented cables, that shock my sens of logic, or some strange "Bybee wire purifier" promotion looking like snake oil to me.
And I teased you repeatedly on those points.
But for most of the else, i am in accordance with your points of view, based on my modest past experiences. And i believe your apparent "Auto satisfaction" hide, in fact, a permanent doubt ;-)
(Yes, John, you can sometimes be maddening;-)
I don't like the "John Curl bashing" witch seems the favorite game in this forum. I know that you cannot have your references and so brilliant success in audio design without solid scientific knowledge and expertise. Kind a jealousy ?
Anyway, such a long and successful career, years of passion in its exercise deserve, at least, some respect. You have mine.
I believe we all have to learn from others experiences, and keep both curiosity and an open mind to other ways to consider the things. And indulgence for defects of character of the others, we all have our part. (Oh, why don't I like flies ?)
On your side, you probably have to change 180° your deep instinctive disgusting for OPAs (i know where it comes from ;-). Things, happily, had changed. It just brings shorter paths, fast temperature exchanges, good pairing to the same king of schematics you can build in discrete. Can save a lot of efforts and money. Oh, those slew rates of current feedback OPAs ;-)
They are like girls. A whole package. Chose the best one, and love her for who she is. (Oh, why am I divorced ?)
Last edited:
OK folks, I think all those that wanted to try this listening test have done so by now. So here is the key......
Project X = LM4562 + 100 ohm
Project Y = LM4562
Results.
Esperado. First choices were correct, but then the doubts set in and I think you found them very similar on a casual listen.
Fas42. just listened (thanks Frank) and found X to be superior by a considerable margin. Rich harmonics and true to life.
Max Headroom. Felt X was the loaded one. Correct.
Rayma. Preferred Y.
RNMarsh. Quite definite. Y was without. Correct. Did you prefer it sonically as well ?
Vacuphile ? ? thought they sounded very close... darned close
Waly. Correctly identified X as with R6. Did you listen or analyse though
So there we go. A big thanks to all that took part. Looking at the results it seems that many were able to identify the loaded one.
The big question after this though is, which would you prefer if you didn't know what you were listening too. I shall leave that one with you.
Project X = LM4562 + 100 ohm
Project Y = LM4562
Results.
Esperado. First choices were correct, but then the doubts set in and I think you found them very similar on a casual listen.
Fas42. just listened (thanks Frank) and found X to be superior by a considerable margin. Rich harmonics and true to life.
Max Headroom. Felt X was the loaded one. Correct.
Rayma. Preferred Y.
RNMarsh. Quite definite. Y was without. Correct. Did you prefer it sonically as well ?
Vacuphile ? ? thought they sounded very close... darned close
Waly. Correctly identified X as with R6. Did you listen or analyse though
So there we go. A big thanks to all that took part. Looking at the results it seems that many were able to identify the loaded one.
The big question after this though is, which would you prefer if you didn't know what you were listening too. I shall leave that one with you.
First, congratulation for your friendly and clever resume.
One remark about some preferences we can have. The samples were both distorted and bad recorded. Piano saturated during attacks, out of tune with a lot of awful dancing harmonics of the other chords. And the sax recorded too close, with a lot too much breath in it.
So, if the X killed a little those harmonics and transients, some can prefer this sample, while he can have exactly the opposite preference, with a very good recording, with percussion, drums etc... ?
One example of the listening difficulties, witch need several different samples to get the full landscape.
I notice that, at this game of identifying witch one was 'loaded', not so much mistakes from all of us ?
Oh, and it made-me remembered how painful, tiring and difficult are those listening comparizons when it is not so obvious.
Last edited:
Waly. Correctly identified X as with R6. Did you listen or analyse though
Listening, the sax low bass compression is IMHO pretty obvious.
From the mouth of Sir Fix-alot...
I like big amps, and I cannot lie..
Single ended doesn't catch my eye...
Scott, a question..
I had a vendor deliver some equipment, with an odd setup.
A smart widgit with serial comms (BISS protocol), 100 feet of cable, three twisted pairs. 5v, gnd, a pair of clock wires, and a pair of comms wires. Clock and comm are differential, and all wires were 22awg equiv. They realized supply drop was too much (.75v), so decided to pull the rails out of the 2 by 3 pair cable, and into a 12awg by 2 twisted pairs to power two devices at the far end. Because the cable was mis-identified by the vendor, I dropped it into the cable tray partition being used for power and noisy cables.
During operation of one device type, a solenoid, the smart widgit momentarily goes belly up. The solenoid cable pulses magfield into the ground loop, so the datacomm common mode voltage goes nuts. It flips well past both 5v and gnd during energization. So much so, that if the inputs are junction isolated from the substrate, the common mode is sufficient to forward bias the tub isolation. The vendor put mov's at the source relay, with no diode clamp at the far end, so they also forced turnoff clamp currents to be within the entire cable.
They've reported no power cycle is required to bring the devices back up, so it looks like it's not latching. However, I suspect that the tub junction forward capacitance (which is far larger than the reverse bias capacitance) is causing one input to come out slower than the other, which causes a large differential mode signal.
So my question: For devices which use junction isolation, is scr action at inputs an issue anymore? And, do designers consider the tub junction profile to tailor or speed up the junction's return to reverse bias? TFR, as it were.
I asked here because an input pair can suffer this as well if the input pair is overwhelmed by slew, so I figgered it was kinda on topic.
jn
I should have elaborated and said 'loaded down' sounding.
I have encountered this 'loaded down' effect previously...loss of attack, sustains not right, dynamics not right, and in this case sounding 'wrongly' out of tune, ie the intended discords are not sounding as they should.
For these reasons I would reject a system sounding like X, even in the absence of a Y comparator.
In this test, an AB/ABX testing regime is not required, X is self evidently 'wrong'.
It is these kinds of characteristic sound faults that make sighted testing (during development) perfectly valid for the likes of JC, Esperado, myself, and others.
In my experience, a mod/tweak/new construction may sound 'ok' to start with, BUT after minutes/hours/days playing sufficient number of tracks of differing genres, minor wronguns become evident, and further become glaringly obvious to the point of major annoyance.
Short term ABX testing is not the right tool to discriminate minor differences reliably, except in the hands of the very experienced listener.
I find it interesting that two of us here were able to correctly identify the 'loaded down' sound on minuscule laptop PC loudspeakers.
IOW, Franks 'famous' PC speakers are actually up to the task at hand.
Dan.
Last edited:
- Status
- Not open for further replies.