>>> Too many caps in the power supply line
>>How can one tell there are too many ?
>Usually, but not always, you see several capacitors in parallel...
So even for a power amp, rather less total capacitance then to have caps in parallel ?
Any reason why (e.g. ringing between caps) ?
How does that manifest itself sonically ?
Thx,
Patrick
>>How can one tell there are too many ?
>Usually, but not always, you see several capacitors in parallel...
So even for a power amp, rather less total capacitance then to have caps in parallel ?
Any reason why (e.g. ringing between caps) ?
How does that manifest itself sonically ?
Thx,
Patrick
This is a no-brainer, John, since you've told it before. 😉
The most important was feedback resistor, then all those multiple bypass capacitors.
If permitted, I would pay attention to J3/J4 and try to avoid that connector altogether.
Best,
Hi John,
I find the bias regulation circuit interesting. If I see it correctly, you look at the output level, and after some processing use that info to adjust the bias level. You said before that it is low-high bias but I guess it is not a two-settings bias but can vary in continuous levels.
I should probably sim that part but can you tell us, what are the sort of time constants here? There also must be an appreciable delay between changes in output level and changes in bias, which can easily account for audible differences.
jan
I find the bias regulation circuit interesting. If I see it correctly, you look at the output level, and after some processing use that info to adjust the bias level. You said before that it is low-high bias but I guess it is not a two-settings bias but can vary in continuous levels.
I should probably sim that part but can you tell us, what are the sort of time constants here? There also must be an appreciable delay between changes in output level and changes in bias, which can easily account for audible differences.
jan
Last edited:
With decoupling caps, its as much about the ground return path as it is about the rail decoupling point. Get it wrong, and you can actually be injecting rail noise into the ground.
I am quite interested as to why you used mosfet cascodes on the front end John, and not bipolar?
I am quite interested as to why you used mosfet cascodes on the front end John, and not bipolar?
Yes, got that. Comparator U4D. Was looking at the epak rectifier around D22, D23. C66 is charged quickly and discharged much slower. So with high input levels, the bias is quickly reduced, but when the signal level drops, it takes some time for C66 to discharge and to restore 'high bias'. Did I get that right?
jan
It is a low-hi bias, depending on whether there is input signal or not. There should be nothing in-between.
I like fets, so I use them when I can
This is only the first part of the story, I have not discussed the second part, yet. We had to do more than change the schematic.
What I am attempting to show, (over and over to some degree) is what it takes to make a SUCCESSFUL audio product that sounds first rate.
I like fets, so I use them when I can
This is only the first part of the story, I have not discussed the second part, yet. We had to do more than change the schematic.
What I am attempting to show, (over and over to some degree) is what it takes to make a SUCCESSFUL audio product that sounds first rate.
Now what would it take John Curl to make it a SUCCESSFUL audio product that sounds first rate.? Start here:
Brgds
I m afraid that there s more than this list...
The power amp part is somewhat archaic and the high power
output capability will change nothing about it.
OK, so what I see is that with high signal level the bias is low and with low signal level the bias is high? Did I get that right?
Edit no I think it is the opposite, with more signal the optocoupler gets more drive, shutting off Q1 and Q30 which increases the bias? Anyone can help me out here?
jan
Last edited:
Jan
Other way around. (with signal higher bias).
p.s. Looking forward to receiving linear audio v3, thanks for your efforts.
-Antonio
Other way around. (with signal higher bias).
p.s. Looking forward to receiving linear audio v3, thanks for your efforts.
-Antonio
Hi,
You can use other methods to have multiple capacitors that do not share the problem of multiple parallel capacitors.
Multiple capacitors plus parasitics create interesting resonances.
Some will call it "glare" or "harshness".
Ciao T
You can use other methods to have multiple capacitors that do not share the problem of multiple parallel capacitors.
Multiple capacitors plus parasitics create interesting resonances.
Some will call it "glare" or "harshness".
Ciao T
Isnt PS by-passing a design implementation to achieve a minimum required impendance over the desired range of frequency? I dont see how this rules out parallel capacitances if done properly.
I see the bigger issue (or at least more subjective) determinining the minimum required impendance over frequency.
-Antonio
I see the bigger issue (or at least more subjective) determinining the minimum required impendance over frequency.
-Antonio
Hi,
IF (and that big IF) it is designed accordingly, it may do so indeed.
Please re-read my original post:
"Usually, but not always, you see several capacitors in parallel..."
I do not rule it out, however in most cases a lot of parallel resonant circuits means a lot problems.
I can see sense if for example we see several smaller value film cap's paralleled to generate a low ESR that matches that of the big electrolytic cap and VERY LOW ESL and we see in parallel a much larger value capacitor with medium ESR, which essentially acts as snubber, but to be honest, I have yet to see such a use published.
Ciao T
IF (and that big IF) it is designed accordingly, it may do so indeed.
Please re-read my original post:
"Usually, but not always, you see several capacitors in parallel..."
I do not rule it out, however in most cases a lot of parallel resonant circuits means a lot problems.
I can see sense if for example we see several smaller value film cap's paralleled to generate a low ESR that matches that of the big electrolytic cap and VERY LOW ESL and we see in parallel a much larger value capacitor with medium ESR, which essentially acts as snubber, but to be honest, I have yet to see such a use published.
Ciao T
The biggest issue is always to dampen transients that result from switching of rectifier diodes. Almost no one cares, though it is well measurable. If not damped, very short spikes get through the amplifier and occur at speaker terminals. On can find them by a fast digital scope in a single shot trigger setting.
Hi,
And if someone has carelessly "bypassed" the power supply, instead of removing these transients, they set off extensive tails of resonant energy.
If only instead the designer had elected to soft switching rectifiers (or schottky ones) place a smidgen of inductance and resistance between multiple cap's of the same type, snub the mains transformer and had carefully considered and snubbed the bypass network...
Shame I almost never see this kind of design.
Ciao T
And if someone has carelessly "bypassed" the power supply, instead of removing these transients, they set off extensive tails of resonant energy.
If only instead the designer had elected to soft switching rectifiers (or schottky ones) place a smidgen of inductance and resistance between multiple cap's of the same type, snub the mains transformer and had carefully considered and snubbed the bypass network...
Shame I almost never see this kind of design.
Ciao T
All my 'A' rated products have high speed, soft recovery diodes throughout.
AS DOES the modified HCA 3500 under discussion, once changed.
Subtlety in adding snubbers and RFI proofing are not completely worked out in this example.
The best reason that I can determine for using a minimum of paralleled power supply capacitors of similar value, is that the capacitors are poorly placed relative to what they are actually supposed to do, and essentially generally do nothing, but create secondary resonances that can easily be simulated with a transient analysis of a similar network with the inductance of the caps and wiring added. Of course, you need a low impedance at the circuit itself, and there I usually use a REALLY GOOD polystyrene or polypropylene cap, that I now to be well internally damped, and generally well behaved.
While this may be obvious to some, is was NOT obvious to me for about 5 years, when I allowed these extra caps to be added to the design. When we removed them, the sound got better for the HCA-3500. End of story.
When we get finished with these details, you might be interested in what else we did.
AS DOES the modified HCA 3500 under discussion, once changed.
Subtlety in adding snubbers and RFI proofing are not completely worked out in this example.
The best reason that I can determine for using a minimum of paralleled power supply capacitors of similar value, is that the capacitors are poorly placed relative to what they are actually supposed to do, and essentially generally do nothing, but create secondary resonances that can easily be simulated with a transient analysis of a similar network with the inductance of the caps and wiring added. Of course, you need a low impedance at the circuit itself, and there I usually use a REALLY GOOD polystyrene or polypropylene cap, that I now to be well internally damped, and generally well behaved.
While this may be obvious to some, is was NOT obvious to me for about 5 years, when I allowed these extra caps to be added to the design. When we removed them, the sound got better for the HCA-3500. End of story.
When we get finished with these details, you might be interested in what else we did.
Yes, a single cap may have a specific resonance peak (in fact all caps have that).
The purpose of the additional parallel cap is to lower the first resonance peak, and replace it with a lower one (at higher freq). Then you use another parallel cap to lower THAT resonance and replace it with a lower and higher freq one etc.
But you must know what you do, you must know the parasitic resistance and inductance of the various cap types of course.
If you are interested in this, there a set of very nice articles with graphs on my website by Kendall Castor-Perry (aka the Filter Wizzard).
jan
Last edited:
Without the hi speed diodes, we are now in 'B' territory, and I am reasonably sure that if just the circuit board mods were implemented, we would have a pretty good product, much like the HCA-2200, but more powerful, that would satisfy most customers.
Now, to go for the 'gold'! How can we make this into an 'A' rated product?
With the hi speed diodes we get a step closer, but there is more to do.
First, every input and output wire has to be of a certain quality, usually subjectively determined, and usually not from the hardware store, yet not very expensive, like some wires tend to be. Connectors, both input and output, have to be noted VERY CAREFULLY as to what material they are made of, despite their external appearance. We normally use a magnet to test for steel, etc.
At this level, we might be an 'A-' or a 'B+' . Close, but not near perfect. more later
Now, to go for the 'gold'! How can we make this into an 'A' rated product?
With the hi speed diodes we get a step closer, but there is more to do.
First, every input and output wire has to be of a certain quality, usually subjectively determined, and usually not from the hardware store, yet not very expensive, like some wires tend to be. Connectors, both input and output, have to be noted VERY CAREFULLY as to what material they are made of, despite their external appearance. We normally use a magnet to test for steel, etc.
At this level, we might be an 'A-' or a 'B+' . Close, but not near perfect. more later
- Status
- Not open for further replies.