Reservoir capacitor upgrade...

[QSilver]

I'm going through an SX-838 I got recently in great condition. I've gone through it all and I'm planning on migrate some of the regulators to heatsinks as theres plenty of room in the case.

I was wondering about changing the main filter capacitors for the power amp and I just so happen to have a pair of 15,000uF capacitors i was going to use on an SA-9100 a while back but they were too tall.

Now my question is, is that too much for the rectifier diodes to cope with as the original capacitors are 8800uF IIRC.

Now, I looked, and the SA-9100 uses the same rectifier (4x SR3AM) to feed two 15,000uF capacitors. The SX-838 has the same SR3AM diodes so is it safe to say it'll be happy with a 15,000uF upgrade.

My only reason for suggesting such a large upgrade is that I have them spare and they'll fit perfectly.

 

[Retrovert]

Eh. Don't drink the capacitor-bank Flavorade. At least not without adding some ice and a dash of vodka.

Adding more capacitance as the first bank is suboptimal. You'll stimulate transformer ringing and stress both the transformer and diode.

I've explained this elsewhere, but, in brief, the issue is rectifier conduction angle. A capacitor only charges when the voltage is higher than what is stored. So by maintaining a larger store of current you narrow the charging region, which then modulates the rectifiers and transformer during charging as a square wave with a low duty cycle. (On for only a brief period.) Not only does this square wave cause the transformer to ring and stress all of the components, but the massive current draw in a short timeframe may cause a voltage drop, which is self-defeating. Remember, you're pull all of the current in a tiny portion of the AC cycle so if I gets big, then R better be very small to avoid IR voltage drop.

Also, the initial charge when the capacitor has no store is going to appear as a dead short to the transformer and rectifier, so an inrush limiter is needed.

This is why a tube rectifier has such low capacitance limits; it's all about instantaneous current load.

If you want a bigger bank, add a resistor and then put in the bank after that. Look up CRC filtering.

Don't forget to add bleeder resistors. Otherwise you're supplying current when the amplifier is off, and that can cause all sorts of issues because the bias can be in unexpected states. Plus there's the issue of safety for you when working on it.

The proponents of huge capacitor banks don't understand conduction angle which makes me wonder what else they're missing.

 

[QSilver]

Thanks for the information. It does already have bleed resistors fitted.

My question is though, that if the SA-9100 has a simple bridge rectifier of 4x SR3AM and two 15,000uF capacitor banks, why won't the SX-838 be happy with the 15,000uF capacitors too as its rectifier is the same 4x SR3AM. Surely the inrush and current demands will be the same on power up?

 

[Retrovert]

My question is though, that if the SA-9100 has a simple bridge rectifier of 4x SR3AM and two 15,000uF capacitor banks, why won't the SX-838 be happy with the 15,000uF capacitors too as its rectifier is the same 4x SR3AM. Surely the inrush and current demands will be the same on power up?

Is the amount of charge stored in both capacitors identical? If not, then the current demands will be different.

Beyond that, you aren't considering the load during normal operation when the capacitor is charged, just not to the same voltage at the rectifier's peak output. Bigger capacitor means the voltage drops less, so the conduction angle is smaller. All of the problems I described kick in.

As I wrote, CRC is the solution. Well, CLC is better, of course, but that may not be possible.

 

[Hipocrates]

Thanks for the information. It does already have bleed resistors fitted.

My question is though, that if the SA-9100 has a simple bridge rectifier of 4x SR3AM and two 15,000uF capacitor banks, why won't the SX-838 be happy with the 15,000uF capacitors too as its rectifier is the same 4x SR3AM. Surely the inrush and current demands will be the same on power up?

Maybe in the pioneer forum someone already has done this. The inrush current can be a problem on the power switches too, there's some thread about how to deal with that using a triac.
I would test them.,,, just remember to adjust BIAS and Offset after the replacement.

 

[Retrovert]

The inrush current can be a problem on the power switches too, there's some thread about how to deal with that using a triac.

The TRIAC avoid micro arcs on the switch during closure and opening which gradually erode it and cause failure. Different problem. (Switch and relay contacts are not perfect, and open and close from bouncing when being closed.) A TRIAC will not address the inrush current at the transformer or ringing. For that a current limiter is needed.

Beyond the inrush issue is the operational problem because of the even more narrow conduction angle.

 

[Hipocrates]

The TRIAC avoid micro arcs on the switch during closure and opening which gradually erode it and cause failure. Different problem. (Switch and relay contacts are not perfect, and open and close from bouncing when being closed.) A TRIAC will not address the inrush current at the transformer or ringing. For that a current limiter is needed.

Beyond the inrush issue is the operational problem because of the even more narrow conduction angle.

Okie dokie, thanks for the clarification

My english sometime is no good and did not mention that, just protect the switch, not the input current, for that you will need something like a soft start circuit right?

 

[QSilver]

Good point about the power switch - I've done the Triac conversion on a few units before which saved the switches.

Going back to the capacitor problem. I'm looking at the two schematics. Both have identical rectifier and voltages - +/- 45VDC and when idle, both have a draw of 60mA on those lines. They are used exclusively for the output transistors.

Now, the SA-9100 is rated at 60WPC @ 8 Ohms whereas the SX-838 is rated at 50WPC @ 8 Ohms. So its not a big difference in power to warrant a large energy reserve with respect to the SA-9100.

Maybe I'm being dumb, and if this isn't a good idea, then I'll just replace like for like. But I want to understand why I shouldn't do this if indeed, i shouldn't.

 

[Retrovert]

just protect the switch, not the input current, for that you will need something like a soft start circuit right?

To clarify. The TRIAC protects the switch contacts because they handle no current, just whatever it takes to turn the TRIAC on, which is very little at low voltage. The inrush current limiters are devices like NTC thermistors whose resistance declines as they warm up. So the current is limited to a small value which can then increase to full value as the device warms up.

The big issue here is that an inrush limiter can't fix the current peak every single AC half cycle when the capacitor is recharging. The bigger the capacitor, the smaller the conduction angle, and the greater the charging current spike.

 

[QSilver]

The big issue here is that an inrush limiter can't fix the current peak every single AC half cycle when the capacitor is recharging. The bigger the capacitor, the smaller the conduction angle, and the greater the charging current spike.

Ok, but why would the capacitor loose charge more rapidly as its capacity increases?

As for the inrush, the rectifier etc can handle the inrush on the SA-9100, so why wont it handle it on the SX-838, if the rectifier and DC voltages are the same?

 

[Retrovert]

Going back to the capacitor problem. I'm looking at the two schematics. Both have identical rectifier and voltages - +/- 45VDC and when idle, both have a draw of 60mA on those lines. They are used exclusively for the output transistors.

I explained this. Those capacitors only charge when the applied voltage exceeds their stored voltage. That creates a short-region charging time and a charging spike.

Look up CRC supplies. That will explain the fix in further detail.

 

[Retrovert]

Ok, but why would the capacitor loose charge more rapidly as its capacity increases?

It doesn't.

What happens is that it better maintains its voltage so it charges during a smaller region of the AC cycle.

 

[Retrovert]

As for the inrush, the rectifier etc can handle the inrush on the SA-9100, so why wont it handle it on the SX-838, if the rectifier and DC voltages are the same?

It isn't just inrush current. That only happens at startup.

This is about conduction angle which happens every AC half cycle. The diode must supply all of the charging current during a narrow window. This increases the peak current on the diode.

 

[QSilver]

Right, so as its charging for a shorter period of the AC cycle, that means the conduction angle is smaller? So this is about the constant current required being passed over a shorter window of time, which would stress the rectifier?

Sorry if I seem slow...

 

[Retrovert]

You got it.

One quibble: the current isn't constant because the amplifier isn't constant. But, yes, it's the charging current per cycle which is being compressed into a smaller window so the current has to increase during that window to move the same number of electrons into the capacitor. Remember, Amperes is Coulombs per Second. Shorten the time and you increase the amperes for that peak, but not overall. Same number of electrons move every half cycle, they just get crammed in a LOT faster.

This is the electrical engineering that isn't taught in university (I certainly never saw it mentioned, and my professors didn't know how to solder, and looked down upon it, and this was decades ago) and the popular books don't discuss it. But it's what really happens and real-world power supply designers worry about it.

You might also look up Mark Johnson's work over at diyAudio on the Quasimodo and Cheapomodo. That explains transformer ringing. Also read up on diode Qrr. And conduction angle, of course.

 

[Retrovert]

Again, this is why tube rectifiers have very small limits for the first capacitor. The cathode will be rapidly damaged by the over-current. The current demand is so high the rectifier will arc and that blows apart the cathode in short order.

The Ole Ones could have said, "do not exceed this much peak current during the charging cycle" but nobody would have understood what that meant, and rectifiers would be exploding like popcorn. So they distilled the peak current into maximum input capacitance which is the same thing, but expressed in a simpler way that everyone could understand and then willfully violate nonetheless.

 

[QSilver]

Ah, that makes a lot more sense, thank you!

I had wondered why valve amplifiers could stand having smaller reserves but I had initial put it down to the fact they are much higher voltage and so current delivery was perhaps faster because of that, or rather higher voltage, less current.

If you can stand me asking more questions... so the conduction angle is smaller if I fit the larger capacitors, and this means shorter harder bursts for the rectifier? So, if in the SA-9100, the same rectifier can cope with these shorter bursts, is it safe to say that the same rectifier in them SX-838 will be able to take the shorter bursts too? Or does that now shift the problem over to the transformer etc?

Just reading up... so short current bursts required by the transformer will cause it to "ring" potentially.

Maybe I should've mentioned, I'm coming at this from a Solid State angle

 

[Retrovert]

Ah, that makes a lot more sense, thank you!

You're welcome.

I had wondered why valve amplifiers could stand having smaller reserves but I had initial put it down to the fact they are much higher voltage and so current delivery was perhaps faster because of that, or rather higher voltage, less current. [...] Maybe I should've mentioned, I'm coming at this from a Solid State angle

Yes, I do understand that. I used the tube rectifier as an example because it is a very common limit but one that people do not understand. And it helps to explain about conduction angle and peak current. Different technology, but same principles involved.

If you can stand me asking more questions... so the conduction angle is smaller if I fit the larger capacitors, and this means shorter harder bursts for the rectifier? So, if in the SA-9100, the same rectifier can cope with these shorter bursts, is it safe to say that the same rectifier in them SX-838 will be able to take the shorter bursts too? Or does that now shift the problem over to the transformer etc?

Yes. Exactly so on the shorter-harder bursts.

As far as the rectifiers, I would switch to better ones for lower Qrr (see below). I don't know about your particular amplifier's power supply. It isn't just the diode current it's the transformer heating. Plus the fact that the square wave (it is a very narrow slice of the sinusoid half-cycle so we can call it a square wave) whacks the transformer causing it to ring.

Remember, the shorter that burst the higher the frequency. Now you've got another problem: you have a carrier wave that is going to be modulated by the mains harmonics. RF transmitting noise around in the amplifier and being received on every lead. Oh, the joy.

I would look into replacing all of the rectifiers with low Qrr diodes and then add snubbers to them. That will clean up the diode noise, too. Inexpensive.

Just reading up... so short current bursts required by the transformer will cause it to "ring" potentially.

Yes. But not just "potentially". In reality. Yes, it will ring like a bell. Johnson's Quasimodo and Cheapomodo will both let you resolve that issue without using math. Damping the transformer prevents the ringing, even when whacked by a squarewave. The damping lowers the Q of the tank circuit formed by the transformer's parasitic inductance and capacitance, so it can't oscillate. (High Q means low losses, low Q means high losses.)

 

[restorer-john]

..Adding more capacitance as the first bank is suboptimal. You'll stimulate transformer ringing and stress both the transformer and diode...

This is not a given. Some transformers could be affected and some bridge rectifiers that were marginal could fail. They are in the absolute minority. This is utter scaremongering.

Filter capacitor choices in the 1970s were entirely budget and space consideration driven. There are many instances where budgets resulted in the original designs being stripped of PSU filtering to hit a budget- even on TOTL gear. Through a product's life, PSU caps were often downgraded as the YEN went up.

The SA-9100 used 15,000uF 50v caps that were made in the early 70s when capacitors were always well above their rated capacitance, many were within a +100/-20% tolerance. Often, I pull out PSU caps that are nearly twice the rated capacitance- even 40 years on. These days, capacitors are more often UNDER their rated uF than over. This alone shoots a nice hole in your argument- you'd be making the amp perform worse than it did before.

It's all very well to rant on about conduction angle current bursts on the sine, but the larger value caps will be able to deliver more current than the smaller value caps for longer before their voltage drops. The upside for moderate increases in PSU filtering, is often a huge improvement in ripple, especially at high power and the consequent improvement in distortion and performance into lower impedances.

 

[QSilver]

This is not a given. Some transformers could be affected and some bridge rectifiers that were marginal could fail. They are in the absolute minority. This is utter scaremongering.

Filter capacitor choices in the 1970s were entirely budget and space consideration driven. There are many instances where budgets resulted in the original designs being stripped of PSU filtering to hit a budget- even on TOTL gear. Through a product's life, PSU caps were often downgraded as the YEN went up.

The SA-9100 used 15,000uF 50v caps that were made in the early 70s when capacitors were always well above their rated capacitance, many were within a +100/-20% tolerance. Often, I pull out PSU caps that are nearly twice the rated capacitance- even 40 years on. These days, capacitors are more often UNDER their rated uF than over. This alone shoots a nice hole in your argument- you'd be making the amp perform worse than it did before.

I had heard a little about this too, more about the older 70's tolerances.

What would be your take on trying out the 15,000uF caps on the SX-838?