Err...wow. Are these things even worth the money?

[Retrovert]

My experience bypassing has not been a positive one, bypassing Sprague Atom electrolytics in the power supply of an SET amp with high quality film caps made for very noticeable smearing in both positions of the power supply (to power tubes and to front end). Cleaned up right away when I took them back out and that's how it's stayed.

Hmmm. I am surprised by that. This was on C1 or C2 in the power supply?

All I can think of offhand is that the amplifier has a bit of high-frequency oscillation and the film capacitor has such low ESR that it coupled the noise harmonics into the ground plane.

Can you throw a scope on that supply and look for oscillation? It will be in the AM band or higher. Tubes have amazing bandwidth and many amplifiers oscillate, totally unbeknownst to the owner.

 

[GordonW]

.

Yes. Also because large capacitors are very poor at filtering higher-frequency ripple current. Electrolytics are even worse. This is why we use decoupling capacitors all over the place with digital logic. The fast switching speed creates very high-frequency noise.

I actually did this decoupling-cap technique in the CJ preamp I referred to above- I put small value film caps (.1uf) directly across the power supply terminals, where the power supply leads went into the individual circuit boards for the phono stage and line stage. I don't know if they are having much additional effect to the bypassing I did on the PS board- but it can't hurt, when you are dealing with as much gain as a phono stage has...

Regards,
Gordon.

 

[GordonW]

Hmmm. I am surprised by that. This was on C1 or C2 in the power supply?

All I can think of offhand is that the amplifier has a bit of high-frequency oscillation and the film capacitor has such low ESR that it coupled the noise harmonics into the ground plane.

Can you throw a scope on that supply and look for oscillation? It will be in the AM band or higher. Tubes have amazing bandwidth and many amplifiers oscillate, totally unbeknownst to the owner.

You may be onto something here.

The ESR of the electrolytic cap, may have been itself damping oscillation in the output stage. Removing that ESR may have actually increased the effective Q of the circuit, to the point where it started to oscillate...

Regards,
Gordon.

 

[toxcrusadr]

Nice post !!! Agree about bypassing. I am NOT an EE like you, or technically as versed as you are. Dramatically is a good word !! I do what you suggest, typically I find it takes not two values, but about 5 to 7 different bypass values, in any one C location, selected by ear. Each uF value plays a different part of the spectrum. Some helpful values might include 0.68 , 0.22, 0.068, 0.033, 0.022 and 0.01 as determined when listening for drop-outs.

Those very small caps have very high impedance even at the upper registers of human hearing. 0.01 uf is 795 ohms at 20 kHz, and 1600 ohms at 10 kHz. 0.03 uf is 265 and 500 ohms respectively. If they sound good to you, who am I to argue, as I'm not an EE either. But functionally, how does a cap like that contribute anything to the final product? Very little current would pass through it.

Don't take this as dismissing the entire concept of bypassing. It just seems that it would require larger caps.

 

[Retrovert]

Those very small caps have very high impedance even at the upper registers of human hearing. 0.01 uf is 795 ohms at 20 kHz, and 1600 ohms at 10 kHz. 0.03 uf is 265 and 500 ohms respectively. If they sound good to you, who am I to argue, as I'm not an EE either. But functionally, how does a cap like that contribute anything to the final product? Very little current would pass through it.

Don't take this as dismissing the entire concept of bypassing. It just seems that it would require larger caps.

Ahhh, but this is not about reactance. The reactance, btw, is a function of total capacitance, not individual capacitance. Large capacitors behave as a number of parallel capacitors, even though it is a contiguous surface.

The smaller capacitors have lower time for dipole alignment which happens every cycle, and this matters at audio frequencies.

 

[BillWojo]

Retrovert, for years DIY folks have been using the Russian K40Y-9 and FT-3 caps in there builds with great success. With your description of Russian manufacturing I would think that there would be a lot of reported failures but I'm not seeing that reported anywhere, just good results.
As a matter of fact the supply of the K40Y-9 caps seems to be drying up, less and less sellers on Ebay have them and the prices are getting higher. You can still find them and the Teflon FT caps for well under 10 bucks a pop but they are not as cheap as they used to be.
Most reports are very favorable as to the sonics when using these, especially the FT-3 type, they are often compared to the high end boutique Teflon caps but cost a small fraction of the price.
I often wonder about some of the very costly caps and who is making them. Are they rolling them on the kitchen table? Very few get sold so there is not a large user base as to there quality or durability compared to the Russian caps. And most folks who spend a small fortune on some miracle cap will tell themselves that it does sound better. Just human nature.
Please site some reports on failure rates and quality problems with the above caps. Without that I call BS. I mean no disrespect as I read a lot of your post and agree with a lot that you have to say but I feel there is a bias in your opinion.

BillWojo

 

[arts]

if I had Jed Clampett's or Thurston Howell's money

Those references are dating us......

 

[BillWojo]

If Gilligan had the money he'd be in for sure.....

BillWojo

 

[Selmerdave]

Hmmm. I am surprised by that. This was on C1 or C2 in the power supply?

All I can think of offhand is that the amplifier has a bit of high-frequency oscillation and the film capacitor has such low ESR that it coupled the noise harmonics into the ground plane.

Can you throw a scope on that supply and look for oscillation? It will be in the AM band or higher. Tubes have amazing bandwidth and many amplifiers oscillate, totally unbeknownst to the owner.

It's a zero-NFB design, and a commercial product (Decware) so I'd be surprised if it had an issue with oscillation. As I said I tried in both positions in the power supply with similar results.

 

[dowto1000]

On a hobby? People spend more than that on dinner. Or a bottle of wine.

I've heard opinions from all over the map from amp designers. Some are big believers in cap upgrades. I believe Conrad Johnson offers $2000-$3000 cap upgrades to some of their amps and swear that it makes a significant difference. Others say it makes little difference.

Well, I am designing and building a prototype 6AQ5 SEP DC UL amp, and I just today ordered 22 caps to act as BYPASS caps. All main caps, in each location, are film, I don't use electrolytics ( use WIMA DC LINKs with 4 leads ).

Using industrially priced WIMA caps, of the right type and value, my total bypass cap cost, at eleven different 'C" positions, was $73.26 plus shipping.

If I would have used my favorite for sound, DynamiCaps, it would have taken 33 caps, not 22 pieces, for bypassing these eleven main Cs, and the bypass cap cost would have been $966.55 plus shipping.

These ordered WIMA bypasses will be nice - enough sounding, I would estimate within 80 % of what the DynamiCaps sonic performance would be. The DynamiCap bypassing would be UNFORGETABLE to hear.

But to get 80% of the DynamiCap performance, for 7.58% of the DynamiCap cost, is what it all boiled down to.

Conrad Johnson is reflecting the SAME trade-offs, except they are not DIY, and must put in a mark-up, on all they offer !! Nice that they offer that.

I just finalized this cap bypass order ......over the last two hours !!

Dowto10000

 

[Retrovert]

It's a zero-NFB design, and a commercial product (Decware) so I'd be surprised if it had an issue with oscillation. As I said I tried in both positions in the power supply with similar results.

Ok. The oscillation may have begun when you increased the Q of the power supply. It's not a feedback issue, it is a ringing issue.

Remember ESR provides damping. It's just like adding a small resistor between the capacitor and ground. So the ripple current (or ringing current) is converted into heat and dissipated. This, of course, stresses the capacitor when a great deal of ripple current is present, which is why electrolytics do not well handle ripple current and why the resulting ohmic heating shortens the lifespan.

Does it have a solid-state rectifier? Diode Qrr can stimulate transformer ringing. With a low impedance supply the ringing can stimulate oscillation. Even without oscillation ringing will degrade the sound quality. The narrower the conduction angle the more the current pulse looks like a square-wave, and the more it stimulates the tank circuit in the transformer causing ringing. That, plus an underdamped supply, is going to generate standing waves and may cause oscillation effects if the ringing is additive.

But transformers can ring even in tube supplies. So that issue is always a problem.

If the ESR goes to negligible amounts the supply becomes seriously underdamped and oscillation can start. So you unintentionally took a stable power supply because of the ESR in the filter capacitors and destabilized it, causing any ringing in the transformer to no longer be damped.

Damping comes in three varieties.

Under Damped (0 ≤ ζ < 1). Oscillates above and below zero at its natural frequency (w × d) where (w = 2 × Pi × f). Think saloon doors which swing back and forth until they gradually stop.

Critically Damped (ζ = 1). Most rapidly converges to zero any with no oscillation. Think of a door closer which perfectly closes door at the optimal rate so the door closes without a sound.

Over Damped (ζ > 1). Converges to zero any without oscillation, but takes longer than a critically damped system. Think of a door closer sized too small so door slams shut, but slower than if nothing were present.​

 

[Selmerdave]

It has a 5U4 tube rectifier.

 

[Retrovert]

It has a 5U4 tube rectifier.

Ok. What is C1 on that? Or is this one of the choke input units?

Most important, have you asked Steve Deckert about this?

I expect he has a ready explanation for the phenomenon and, if not, can reproduce it himself and tell you what is happening. Maybe someone on the Decware forums?

It has to be that a high-Q supply filter no longer damps out oscillation or ringing originating somewhere in the amplifier. Adding impedance is always the trick to lower Q (i.e. increase losses) and thereby stop ringing/oscillation. The grid leak can sometimes do this, but the grid stopper certainly does.

Remember, the capacitor cannot be adding sonic signature. What it can be doing is interfering with damping. The parasitics would create tanks in the hundreds of MHz, far above what would be amplified. Maybe a tube goes that high, but doubtful. I think there's something no longer being damped.

I suggest asking Decker and then report back. It would be interesting. Knowledge would be gained.

 

[Selmerdave]

Ok. What is C1 on that? Or is this one of the choke input units?

Most important, have you asked Steve Deckert about this?

I expect he has a ready explanation for the phenomenon and, if not, can reproduce it himself and tell you what is happening. Maybe someone on the Decware forums?

It has to be that a high-Q supply filter no longer damps out oscillation or ringing originating somewhere in the amplifier. Adding impedance is always the trick to lower Q (i.e. increase losses) and thereby stop ringing/oscillation. The grid leak can sometimes do this, but the grid stopper certainly does.

Remember, the capacitor cannot be adding sonic signature. What it can be doing is interfering with damping. The parasitics would create tanks in the hundreds of MHz, far above what would be amplified. Maybe a tube goes that high, but doubtful. I think there's something no longer being damped.

I suggest asking Decker and then report back. It would be interesting. Knowledge would be gained.

It's 5u4->40uF->1k->40uF->6.8k->2x20uF in series. I don't recall at the time if I reported this or not, it was probably a dozen years ago. I do seem to recall Deckert recommending against it in order to maximize simplicity, but the audible results were all I needed to make my decision. I believe it was C2 and C3 that I bypassed, together and individually, but it's possible I did C1 as well though not as sure about that.

 

[Catmanboo]

I'd rather spend the money on beer, motorcycle go-fast parts, ammunition, pizza & cat food, since I doubt I'd be able to hear any subtle sonic improvement to justify that cost.
My luck, that'd be the first part to fail anyway
But that's just my practical side talking!.

 

[dowto1000]

It's 5u4->40uF->1k->40uF->6.8k->2x20uF in series. I don't recall at the time if I reported this or not, it was probably a dozen years ago. I do seem to recall Deckert recommending against it in order to maximize simplicity, but the audible results were all I needed to make my decision. I believe it was C2 and C3 that I bypassed, together and individually, but it's possible I did C1 as well though not as sure about that.

Off topic.

It would be really nice to PSUD that power supply circuit, and see if the 1 K Ohms of series resistance added, prior to the Final tube, can be easily reduced to 20 Ohms, or fifty-fold. This gets better amplifier dynamic response, from the supply.

I am thinking two small ( 2 pound ) sub-critical chokes, of about 10 Ohms each, and maybe from the rectifier tube into a L, not into a C.

If you are able to document VACs, VDCs, DCRs, from the Power Transformer forward to C2, I would be happy to PSUD analyze that for you.

If possible to do, it will be heard as greater dynamic response, and generally, more overall listening fun. Series resistances in the B+ filter, to the Finals, is a very important thing to minimize.

There might be under-chassis room to do this, but one needs to fully document the stock supply, as suggested above, first.

Dowto1000

 

[Selmerdave]

Off topic.

It would be really nice to PSUD that power supply circuit, and see if the 1 K Ohms of series resistance added, prior to the Final tube, can be easily reduced to 20 Ohms, or fifty-fold. This gets better amplifier dynamic response, from the supply.

I am thinking two small ( 2 pound ) sub-critical chokes, of about 10 Ohms each, and maybe from the rectifier tube into a L, not into a C.

If you are able to document VACs, VDCs, DCRs, from the Power Transformer forward to C2, I would be happy to PSUD analyze that for you.

If possible to do, it will be heard as greater dynamic response, and generally, more overall listening fun. Series resistances in the B+ filter, to the Finals, is a very important thing to minimize.

There might be under-chassis room to do this, but one needs to fully document the stock supply, as suggested above, first.

Dowto1000

The power transformer is a Hammond 272FX, so I would think those specs are available. What you're proposing would raise the B+ a fair bit, and it's at about 350V as is which is already high for a triode-connected EL84 so I think I'd pass just on reliability grounds. Plus I realize you think everybody must share your preferences but I took this amp out of service in favor of PP monoblocks with feedback some years ago (I also triode-connected those finals and removed feedback but prefer them with), on balance they suit my preferences better even though the Decware sounded excellent. I still have it boxed up but it hasn't seen use in a bit.

 

[Retrovert]

It's 5u4->40uF->1k->40uF->6.8k->2x20uF in series. I don't recall at the time if I reported this or not, it was probably a dozen years ago. I do seem to recall Deckert recommending against it in order to maximize simplicity, but the audible results were all I needed to make my decision. I believe it was C2 and C3 that I bypassed, together and individually, but it's possible I did C1 as well though not as sure about that.

Yeah, this suggests ringing which was damped out by the ESR of the existing filter capacitors. It doesn't take much, and 1 or 2 Ω would be enough. Add an impedance path which is so low it cancels out the impedance and destabilizes the supply.

Willing to do an experiment?

Take the bypass capacitor and add a series resistor big enough to ruin the Q. The impedance of the bypass at HF (ringing) frequencies should therefore be sufficient to damp the ringing and lower the Q. If that doesn't sound bad, the issue is ringing.

An RC snubber across the transformer would likely be the best way to fix this. Before the rectifier. You want to pass AM-band frequencies, say above 100k Hz. This will prevent any tank circuit from self-stimulating because the losses are too high.

I think you've got a power supply just on the cusp of being unable to damp transformer ringing, and the ESR is just barely enough to damp it.

 

[Retrovert]

I'd rather spend the money on beer, motorcycle go-fast parts, ammunition, pizza & cat food, since I doubt I'd be able to hear any subtle sonic improvement to justify that cost.

Pretty much. (Sauteed chicken livers for kittehs is not expensive, though.)

I greatly doubt the difference between boutique and properly bypassed PP is audible. Between boutique and electrolytic, yeah, fer sure. Between boutique and ceramic, oh, mos def. Between boutique and tantalum? Yeah, no doubt. But between boutique and PP film? Eh, doubtful.

I offered to set up that experiment some years ago—double-blind with amplifier behind a curtain, two witnessed, and different coupling capacitors switched in and out for each trial using a randomized pattern, 30 trials, must do better than 2/3 to win—but nobody wanted to try. Even with money at stake at good odds.

Reminds me of the test JJ Johnston did where many purported audiophiles couldn't tell the difference between tube amplifier and a transistor amplifier.

 

[dowto1000]

The power transformer is a Hammond 272FX, so I would think those specs are available. What you're proposing would raise the B+ a fair bit, and it's at about 350V as is which is already high for a triode-connected EL84 so I think I'd pass just on reliability grounds. Plus I realize you think everybody must share your preferences but I took this amp out of service in favor of PP monoblocks with feedback some years ago (I also triode-connected those finals and removed feedback but prefer them with), on balance they suit my preferences better even though the Decware sounded excellent. I still have it boxed up but it hasn't seen use in a bit.

Hello , hello !!

Something doesn't add up here with 350 VDC you claim, and the Hammond PT published specs applied in PSUD.

Unloaded, the PT, rated at 175 mA., is speced 354 VAC each half. I assumed about a 93 mA. stock load, and it dropped to half its 175 mA. / 300 VAC rating, or, I used 327 VAC in PSUD, as the secondary VAC of the PT.

This only gets 294 VDC at C2 in the stock supply when simulated, not your 350 VDC claimed. If I plug in the UNloaded VAC 354 VAC, its still only about 330 VDC at C2.

Realize, Hammond PTs in this series took a production - running change on their primaries, in about 2009 or so. They went from a single 115 VAC to a dual tapped 115 and 125 VAC. Maybe, if you have a dual tapped PT, you are connnecting your line on the INSIDE taps, and with possibly your "high" home LINE ( up to 125 VAC is possible) MAY be connected to the PTs 115 VAC winding. That will get you a too high B+ easily. Easy to check and fix, also.

IF you could dig the unit out of storage, and tell us about the primary ( tapped, and if so, connected on what primary tap ) and the MEASURED VAC of the house's line, the high voltage secondary at idle, and the VDC at C1 and C1, and their respective VACs, I could do some much better evaluations in your behalf.


My thought was to eliminate the 40 uF after the 5U4G, go to a 5U4GB, and use a non-critical low DCR L filter, which, depending on many variables, MAY possibly reduce the VDC at C3 to a reasonable point. Since you are not using the unit, will you give it a decent second look !!

In PSUDING, the stock CRC filter with 1.000 Ohms dropped 19.37 VDC on a 15% step, whereas the L/CL/C dropped 8.9 VDC on the same sized step.

From my experience, the 1,000 Ohms kills any amp - dead cold. I need more data to try to assist here.

Dowto1000