Sound Quality Vs. Measurements

Status
Not open for further replies.
Last edited:
Frank, all your writing here begs a question, I'm just curious: have you noticed how Sony tends to do large PSUs?

They throw in two large value caps, like 15,000 uF and upwards, but they also add same type, same make say 4,700 uF in parallel. Appearently, they believe large caps filter better but are slow, while smaller value caps don't filter as well but are faster. They have been doing that for at least 25 years now.

Another, more involved, approach is what I saw in an Italian amp, by their well know local company Galactron, in cicrca 1979 or so. They had a central filtering reservoir consisting of two 10,000 uF caps un parallel for each line, and then added 4,700 uF caps about 5 mm at most away from the output devices themselves.

This intrigued me, so in the artwork of the amp I am doing now space and place was provided for 2,200 uF caps also next to each output device. I just had to try it.
 
A very simplistic viewpoint is to see the big value cap as providing the energy reserve to handle bass notes without the rail sagging too much, while the smaller unit provides the current transient on a "fast", high slew rate part of the waveform. Big caps have effectively relatively poor ESL and ESR, they're not happy being asked to "move fast".

Yes, the Galactron approach is spot on - as a subjective indication, that should give you relatively tight, "fast" bass ...
 
Last edited:
Actually, I think it's an honest attempt to get the best of both worlds. The big reservoir will filter well, the smaller caps will deliver about 2 Amps peak (using 2,200 uF caps) each almost instantaneously on demand.

Initially, I will try my default capacitor values and manufacturer, German made Fisher & Tausche, but I will also try others as well. The initial value for the central reservoir will be 2 10,000 uF caps in parallel for each power supply line, separate for each channel, so a total of 80,000 uF centrally, plus 6,600 uF per line locally (3 pairs of output trannies per channel, for a grand total of 53,200 uF per channel, or 106,400 uF for the whole stereo version.

I do expect good results, however I am also prepared in advance to have many a discussion with Nige, which might result in improvements over my initial setup. I certainly hope so and am very much open to suggestions. When the first working prototype is sitting on my table, I'll post the whole shebang here, so everyone can have a go.

And while I may appear to be firmly stuck to F&T caps, and while it true that I dearly love them, I am NOT vitally attached to them, nor am I a Japanese hater. My Philips HA 370 ("Black Tulip") power amp, rated at 60 WRMS per side, originally had two Nippon Chemcon caps rated at 6,800 uF/50V. They were 30 years old, and were replaced after some trial and error by two 12,000 uF/71V Nichicon Golden Tune caps. Thank God Philips made the rectifier way more powerful than really needed, but that's what most reputable companies did in 1981.

And Frank? That "very simplistic" viewpoint actually does work that way, I tried it before. Let the darn thing breathe and it will work better.
 
I have always suspected that beyond a certain point all we can reliably hear as amplifier upgrades are PSU improvements . As I said earlier it might be possible to contrive a PSU which is just about able to handle the ripple currents and has some surplus voltage to burn off . Suggestion 2 x 10 000 uF with 10 V over voltage to output dumpers ( class B ) . Smooth supply at - 10 V ( +/- ) for class A amp sections ( LTP VAS ) .

Here is the possible benefit apart from price . The amplifier and PSU could be combined . The track length almost zero at the 0V reference point . I have a hunch that making the dumpers first in line would be best . Then simple regulators to VAS etc . If someone says this is speculation , it is . We throw ideas around as money is the only other solution . Proving that speculation would be very expensive . People buying something is the only way ( Carver ) . Even then a victory of hype over reality is not unknown .

My belief is the big cap PSU's are merely giving cleaner VAS supplies . Somewhere in history the imperatives of the cheap amplifier market have become universal . That is CCRR will do miracles so no need to know about clean VAS supply .

An analogy I would like to put forward . I have often wanted an open formula for push rod engines in formula one . That is if you want 10 litre V12 you can have it . This would be to run against 2.4 litre V8 that will end this season . I know it would be a non starter as the 2.4 litre will outperform it . Weight , size and rev ability . The analogy is that mostly we are building 10 litre V12 's in audio . No harm in it and rather marvelous .

Unless genuinely needing 1000 W 1 ohm on peaks I doubt the energy calculations justify the capacitance often used ( me guilty also ) . A suggestion . Build an amp to my recipe ( over voltage dumpers , class B ) . Have VAS emitter at 0R 16 R 47 R . I suspect the VAS emitter resistance will have more effect on punch than adding 47 000 uF per rail . To repeat myself , I am convinced that the VAS base impedance is the slew sound of amps . The numbers far exceed any real requirements . One way to make an amp sound nicer is VAS emitter resistance . The better one is a high gain VAS transistor . For reasons I don't understand I feel a high gain VAS transistor rivals a Darlington . I am sure the Darlington VAS sounds opaque . Cascode seems better .

I beleive Citroen in the late 1920's said 330 cc was an optimum cylinder size ( He came from Voisin as an aero engineer ? ) . The French had a fiscal horse power tax which favoured long stroke engines ( bore = horse power ) . If my memory is correct the 330 cc was for square bore and stroke . Nearly 100 years later that rule is still about right . Amp PSU's are not as well understood . I come from two families of mechanical engineers . I am not one myself . I have worked with them and have had to run projects . The difference I see is mechanical engineers are use to component failure . Mechanical engineers must not have component failure . It simplifies the tasks . Sadly nothing holds the amplifier designer in check . The mechanical engineer is often struggling to meet a weight limit . Over engineering is a luxury that reality will not allow . What the mechanical engineer looks for is a safety margin . That's what an amplifier designer should be doing . They don't have to build it or sell it . They should know it .

A friend is using an amplifier I designed . The PSU is now 5 times larger . I got very upset when someone wanted to double the components in the amplifier . I think the reason was it would do no harm and allow him some design input ( ? ! ). The owner of the amp firmly said no . They put in some new current mirror designs I had suggested . The irony is , it is a FET amp . 5 times the PSU size will not do what converting to bipolar output stage might . That's what I would have done if wanting more punch . Audio Precission test suit available , still didn't let the obvious be seen . My test gear is mostly home made . It get the same test results . Dvv said I design amps by taking out parts until they stop working . Absolutely correct . However I use a double VAS .

Galactron were imported by Goodman's into the UK I think ? Goodman's were and are excellent engineers . The public image was dowdy . Somehow good engineering seldom is enough .
 
Last edited:
My experience says a Darlington VAS will sound warmer and somehow more powerful than a cascode VAS, but the cascode VAS tends to be somehow cleaner and better defined, with more detail.

Where a cascode VAS beats the pants off almost any other VAS configuration is the fact that it can be configured for exceptionally short settling times very easily, no particular pain, just 2 components (a fast diode and a resistor) over the standard fare. Those two components can shorten the settling time by a factor of 10:1, which is no joke. And if you wonder can we really hear that, the answer is yes, we can, the sound simply becomes better defined without turning into a shriek or some such. Sort of snaps into better focus.
 
Unless genuinely needing 1000 W 1 ohm on peaks I doubt the energy calculations justify the capacitance often used ( me guilty also ) . A suggestion . Build an amp to my recipe ( over voltage dumpers , class B ) . Have VAS emitter at 0R 16 R 47 R . I suspect the VAS emitter resistance will have more effect on punch than adding 47 000 uF per rail . To repeat myself , I am convinced that the VAS base impedance is the slew sound of amps . The numbers far exceed any real requirements . One way to make an amp sound nicer is VAS emitter resistance . The better one is a high gain VAS transistor . For reasons I don't understand I feel a high gain VAS transistor rivals a Darlington . I am sure the Darlington VAS sounds opaque . Cascode seems better .



.

Why are you guys always singling me out ....... :)
 
Why are you guys always singling me out ....... :)

No way am I . It is like having a 155 MPH car . It gives confidence . When forced to have 110 MPH make if feel like it would be 155 . Mine is like that . The engine is good for that . I have the base model and is limited . A simple remap and it will go from 90 to 168 BHP . Mine was higher . As I service it myself the computer defaults to standard mode if not having the computer reboot link . I still have the confidence it is there for the asking . My mate has a full mapped engine .

I am tempted to draw my minimal PSU concept just in case it wasn't understood . Looking at TO3 I would hook my 10 000 uF directly to the collectors using a conductive plate cum heat sink to maximize conductivity . The spare cap terminals to the ground plane directly ( GP ) . A single rectifier bridge to each collector on the heat sink . The spare terminal to the GP . This is to minimize losses ( one rectifier would be OK , two is easier ) . A dropper resistor to the next stage with capacitor and perhaps zener ( VAS ) . The TO3's 10 V higher than VAS . The 10 000 uf arranged at 90 degrees to minimize heat pick up . I would predict the amp would be a kicker and super sweet . The transformer wires would be usual length going almost directly to the TO3's . The TO3 terminal it's shape and extended for soldering . The 1 ohm stuff is possible if the caps can take it . 22 000 uF might still work . My speakers are 4 ohms on a good day . I often hear the current limiters working ( nothing nasty , just a gulp ) .

If the heat sink is internal and split it will be a conductor in both senses .

About ground planes . Is it true if of low resistance no need to do anything else ? I guess PSU , and speakers close to the rectifier end ? Signal in as far away as possible ? To stress it would have no star or bus-bar ground .
 
Last edited:
My experience says a Darlington VAS will sound warmer and somehow more powerful than a cascode VAS, but the cascode VAS tends to be somehow cleaner and better defined, with more detail.

Where a cascode VAS beats the pants off almost any other VAS configuration is the fact that it can be configured for exceptionally short settling times very easily, no particular pain, just 2 components (a fast diode and a resistor) over the standard fare. Those two components can shorten the settling time by a factor of 10:1, which is no joke. And if you wonder can we really hear that, the answer is yes, we can, the sound simply becomes better defined without turning into a shriek or some such. Sort of snaps into better focus.


Great point . Another is we get a dream transistor at low cost and using plentiful parts . I built a transistor valve cascode after loosing faith in all valve cascodes . It was incredible ( BC 550 C + ECC 82 I think ? ) . I didn't use it as the guys wouldn't like it . I used a pentode . The cascode said what I needed .
 
I was very surprised to find there is no practical use in going above 10 000 uF . If wanting a big amp I would use 6 x 10 000 uF offering 30 A ripple current .

This component seems optimum . 100 V devices fit my most interesting model . Often 100 V devices sound faster and less dark . The dialectic is better to resit flash-over .
ALS30A103KE100 - BHC COMPONENTS - CAPACITOR, 10000UF, 100V | Farnell United Kingdom
 
Status
Not open for further replies.