Overfiltering?

[Loogie]

...the amps were exceptionally quiet, and had remarkable transient response.

How do you hear transient response? Or better, what do you hear? What are you listening for?

...most guitar amps have the filter nodes oriented in series, rather than coming off of the input filter in parallel.

Why is this do you suppose?

I've been experimenting with a homebrew. After reading this thread I think I'm over-filtered. Right now I'm trying 40uf plates, 60uf screens and 22uf for each section of the preamp I've decided to decouple. I've tried different designs including some series/parallel filtering.

The amp sounds very good to me, but occasionally there are artifacts in the bass range that I don't care for. Difficult to describe. Not flabby, but intrusive in an unpleasant way. Its easy enough to remove them by adjusting gain and/or eq, but you've got me thinking about trying smaller caps. Its quiet and except for that occasional bass annoyance it does everything I want sonically.

[Bruce / Mission Amps]

There are a small number of guitar amp designs out there that do use parallel B+ rails from the main filter group.
There is nothing wrong with that at all but it does take a little more series resistance in the B+ rail to get the voltages down for those first preamp tubes down into the B+ levels that sound good, warm and not harsh.
Most of what I say is my personal preferences but I do find that too high of B+ on the preamps does not sound good either.

[Rob Mercure]

One farad input filter with tube recto idea

Well, perhaps I'm exaggerating, I dunno. Anyhoo, a "terribly obvious" idea occured to me that some of you have probably already thought of - to allow "high" values of capacitance - say 200uf and up - to be used with a tube rectifier. Assuming a "conventional" FW rectifier setup with a PT CT'd "B" supply secondary: Use a standby switch that lifts the output tube cathodes (or crosslinks the output tube control grids - not important, etc.) in one section with a second SPDT section that switches between the tied cathodes of two SS rectifiers or the common cathode of a thermionic rectifier. The parameter that limits using a valve with high first filter capacitance is the initial inrush current required to charge the capacitor which may exceed the tube's designed capabilities. Using this scheme the SS diodes would charge the first filter and once the bottle recto has warmed up it would switched in to take the smaller burden of "topping off" the first filter's charge (while forcing the wee little siliconistas to sulk in the corner).
Of course to some extent you'll limit/eliminate the "sag" in the power supply that many valve recto users desire but that's a function of the greater capacitance (which returns us folks to the semi-hijacked thread theme).

My 3 AM contribution.

Rob

[soundmasterg]

I ended up using parallel filtering rails for my brother's Silvertone 1484 mainly due to needing a higher voltage supply for one channel than the other since I added an EF86 to that channel. It worked well and sounds good too. I plan to use the technique on more amps in the future too.

[bob p]

hijacking:

rob, i had similar ideas to yours when i had to fix a pair of W4-M Williamson monoblocks that were eating rectifier tubes. the guy who had "restored" them committed some acts of true sacrilege -- the most alarming of which was putting several thousands of uF of filtration in the Pi filter that was supplied by a 5R4 rectifier. powering up the amp was pretty stressful on the diodes -- you could see a burst of glow in them at power-on and they had to be replaced frequently. evidently the guy who redesigned the amp liked the idea of excessive filtration, but didn't think through the downsides of beating a tube rectifier to death every time that you powered it on.

[Steve Conner]

Hi guys, been off on holiday!

I hope this doesn't sound wrong, but my mental picture of how filtering works in guitar amps is that, with the regular values of caps used in old amps like the tweed Deluxe, when you crank the amp you get sag that sounds like a compressor. (Technically it IS a compressor/limiter, as the sagging screen and plate supplies modulate the gain of the power tubes as well as the maximum output voltage.)

Overfiltering with the much larger caps you can squeeze into an amp, in these days of etched foil electrolytics, more or less disables your compressor, and you lose the lively pumping, breathing effect that a good compressor gives. That, I argue, is why the amp sounds "stiff" and "dead".

If this theory were true, then you could still get that good saggy, squishy tone with huge filter caps, by adding an extra time constant to the screen node. But you wouldn't have trouble with hum and ghost notes.

The amp I'm building just now has a SS bridge rectifier with a Pi filter composed of a 50uF cap, a choke that weighs about a pound, and another 50uF. The plate and screen voltages for the power tubes come off this second cap.

I tried it against my old amp with the regulated supply, and the new one sounds great when cranked into a power soak. The old one sounds much the same cranked and attenuated as it does with the power soak removed and the MV turned down.

BTW, exceeding the datasheet maximum capacitance for tube rectifiers is NOT a good idea! It's not just about inrush either. Even in steady-state operation, too much capacitance leads to a small conduction angle and hence very high peak currents that use up the cathodes in no time.

[R.G.]

You're right, Steve. That's why tube rectifier data sheets list maximum values for the first filter cap. The bigger the cap, the smaller the time the rectifiers conduct at peaks of the AC wave, and the higher the current peaks must be to get enough DC in during the short conduction time. Eats the cathode coating. As I remember this, it's in the Tube Amp FAQ at GEO, since about 1994.

Some comments on power supply issues.

The tube rectifier limitation is a first filter cap only issue. For second through N filter caps, you may use as large a cap as you like, because it's only the first filter cap that takes the rectifier peak current beating. You can put any modest impedance between the first filter and a second filter cap and the second filter cap only takes the difference in DC voltages, much more modest and continuous than the first filter.

One scheme is to use a really modest first cap, 10uF to 20uF, then a choke that can also be modest, just to isolate the ripple voltage from the first cap. and a second cap as large as you like.

You can re-add squishiness by then introducing resistance or a tube rectifier as you like, to taste. The tube rectifier is limited in this case only by its peak average current and dissiption.

It also makes sense to star-wire the preamp tube power sections, subject to taste, as soon as you get the bulk B+ voltage down to what you need for them. Star wiring into a source-follower regulator can really cut ripple feed from the first filtering section.

[bob p]

BTW, exceeding the datasheet maximum capacitance for tube rectifiers is NOT a good idea! It's not just about inrush either. Even in steady-state operation, too much capacitance leads to a small conduction angle and hence very high peak currents that use up the cathodes in no time.

How's this for interesting: in those Heath monoblocks that I had mentioned earlier, the fellow who redesigned them was wise enough to overbuild the filter, using CDE 550C caps (!), which are expensive 105-C, ultra-ripple, long-life, inverter grade caps. What escapes me is why he failed to think about or recognize the reason for the premature heater failures in the rectifier tubes.

[bob p]

You're right, Steve. That's why tube rectifier data sheets list maximum values for the first filter cap. The bigger the cap, the smaller the time the rectifiers conduct at peaks of the AC wave, and the higher the current peaks must be to get enough DC in during the short conduction time. Eats the cathode coating. As I remember this, it's in the Tube Amp FAQ at GEO, since about 1994.

ah, cathode stripping. i remember being very surprised with disbelief when I read an old Ampage thread where the consensus of opinion was that the phenomenon wasn't an issue at the lower voltages used in guitar amps. (maybe Rob or Dai have an archive.)
.
Its usually pretty hard for me to remember exactly where I've read something, but as it turns out I read an e-book last night on a related topic. Some Ampagers might find interesting and helpful: Robert Tomer's 1960 publication, Getting the Most out of Vacuum Tubes. The book essentially discusses common failure modes for tubes and how to avoid them. It pays a worthwhile amount of attention to heater failures. Its more of a qualitative read than a quantitative one, and its only 150 pages long, so it can easily be read in an evening. Its available for download on Pete Millett's site as a PDF.

[bob p]

The tube rectifier limitation is a first filter cap only issue. For second through N filter caps, you may use as large a cap as you like, because it's only the first filter cap that takes the rectifier peak current beating. You can put any modest impedance between the first filter and a second filter cap and the second filter cap only takes the difference in DC voltages, much more modest and continuous than the first filter.

One scheme is to use a really modest first cap, 10uF to 20uF, then a choke that can also be modest, just to isolate the ripple voltage from the first cap. and a second cap as large as you like.

I remember reading about various input filters in one of my old texts (I wish I could remember which, but I've been reading a lot of old books from the Millett site lately), and one of the "tricks" that the old guys offered as a solution to the problem of "hummy" choke input filters was to place a tiny "hum reduction" cap in front of the choke input. This of course, converts the choke input filter into a Pi-type cap input filter. The difference, though, was that the "hum reduction" cap that they recommended at the input to the Pi was a really tiny little cap, so it didn't drop the filter's output voltage very much. Essentially, the authors called the first cap in the Pi filter a "hum reduction" cap, and the second filter cap a "smoothing" cap, for obvious reasons.

Reading between the lines, the obvious downside to their idea was that the small cap at the input to the filter took a horrible beating in terms of ripple current, so the authors recommended a non-lytic cap in that position, that was over spec'd for voltage, and mounted for heat dissipation. (Yep, there's no such thing as a free lunch).

another potential problem with the modest first filter cap approach is that the voltage regulation of the PSU isn't as good as it is with a symmetrical pi filter. of course, if you're not interested in HiFi, that might actually be desirable.

[Rob Mercure]

Hey Bob,

The cathode stripping discussion involved applying plate voltages to output tubes before they were warmed up - the specific was whether a standby switch was "necessary" for consumer power output tubes. As best I remember there was no mention of rectifiers in the original discussion. But Steve has put some of the old archives out now and they are easily searched if you want to tackle it.

Rob

[Steve Conner]

Wow, inverter-grade caps with tube rectifiers, that's kind of like putting gold-plated wire wheels on a Geo Metro.

That's one of the things I like about tube audio BTW. The limitations force you to be creative. In solid-state design, overkill is easy, but in tube design, when you play the overkill card your hi-fi amp ends up the size of a filing cabinet, needing 5 large roadies to move it, and costing about 50 grand

I have a lot of respect for the original designers of things like the Dynaco ST-70. They couldn't reach for MOSFETs and FREDs when they got stuck, they had to build to a price and make it manufacturable, and they still managed to make something that sounds sweet even by today's standards. That concept pleases me more than any amount of gold-plated high-end tube gear would.

The ST70 iron is still being cloned and manufactured today, but I doubt the chips they used in the Gainclone will still be getting made in another 50 years...

[Enzo]

Cathode stripping - I agree with Rob. When we poo poo cathode stripping, it is in regards to output tubes and not having standby. It didn't involve rectifiers.

And come to think of it, I doubt a standby switch would reduce the too-large-cap problem for a rectifier tube anyway, considering Steve's small conduction angle points.

[bob p]

And come to think of it, I doubt a standby switch would reduce the too-large-cap problem for a rectifier tube anyway, considering Steve's small conduction angle points.

i agree, even steady state operation would be a stress on the tube rectifier -- but steady-state operation with warm heaters is one problem and charging huge caps with cold heaters is another problem. eliminating one problem would be helpful and eliminating both problems would be optimal.

unfortunately servicing an amp isn't the same as designing an amp from scratch. when you're designing an amp from scratch, you have a clean slate and you can do whatever you like within the limitations imposed by the components. in contrast, when you're servicing an amp, you have to work within the limitations of the hardware that you've got to work with, owner budgetary constraints, and owner preferences. sometimes there's no option but to compromise, and that leads to the kinds of solutions that @Rob suggested. they're not always perfect solutions, but they're still helpful.

yes, designing a "proper" PSU from scratch would be a better approach, but sometimes people don't like your ideas about what is "right." some HiFi types won't let go of tube rectification, and some aren't technically adept enough to understand how a good sounding amp can be plagued by a bad design. making matters worse, cost conscious people might think you're nuts if you recommend throwing away expensive components that are working properly and have already been paid for -- especially when they have the option of replacing relatively inexpensive "disposable" components on an intermittent basis. sometimes you're forced to fix an amp with one arm tied behind your back, and you have no option other than to compromise. sometimes "fixing" an amp involves making a problem less bad instead of making a problem go away.

so what do you do when someone is adamant about keeping tube rectification and doesn't want to throw away the expensive inverter grade caps that have already been paid for?

in that scenario, even though you can't do much about the effects of the cap size on conduction angle, you can fix the problem of abusing the cold heater. fixing the avoidable cold-start problem is an inexpensive approach that should help the rectifiers to last longer, and that made the owner happy. he got to keep his tube rectifiers, he got to keep his expensive caps, and he might not have to buy rectifiers as often. it wasn't a perfect solution, but the owner liked it, and it was better than doing nothing.

@Steve and @Enzo, I have a question for you about this -- I designed a warmup circuit for the amp where the tube rectifier was taken out of the circuit and replaced by SS diodes for charging the empty caps. then, once the caps are charged and the tube rectifier is up to temp, the tube rectifier is added into the circuit. this effectively eliminated the "cold start" problem, but didn't do anything about the steady-state problem.

here's my question: how stressful would steady-state operation be on a warm tube rectifier that has SS rectifiers in series with it upstream, compared to a tube rectifier operating in steady-state all by itself?

i placed SS rectifiers in between the PT and the tube rectifier, as protection for the PT in the event that the tube rectifier failed. i'm wondering whether having the "heavy lifting" performed by SS diodes would help to reduce the stress that's placed on the tube rectifier by the narrow conduction angle. any insights you might have would be appreciated.

[R.G.]

ah, cathode stripping. i remember being very surprised with disbelief when I read an old Ampage thread where the consensus of opinion was that the phenomenon wasn't an issue at the lower voltages used in guitar amps.

It isn't. Cathode stripping only occurs in much higher voltages from what I've read. I've never seen it. But as Rob and Enzo say:

The cathode stripping discussion involved applying plate voltages to output tubes before they were warmed up - the specific was whether a standby switch was "necessary" for consumer power output tubes. As best I remember there was no mention of rectifiers in the original discussion.

Cathode stripping - I agree with Rob. When we poo poo cathode stripping, it is in regards to output tubes and not having standby. It didn't involve rectifiers.

The only way cathode stripping and rectifiers got linked was that there was a contingent of people who said that using solid state rectifiers was bad because the B+ came up instantly, before the heaters got hot, and that you ran the warmup time with cathode stripping going on each power-on. Near as I can find out, that contingent was wrong because of the low voltages in guitar amps (i.e. under 700+ V).

And come to think of it, I doubt a standby switch would reduce the too-large-cap problem for a rectifier tube anyway, considering Steve's small conduction angle points.

It won't. What does help is to use SS rectifiers into a first cap, then if you like the squishiness from a tube rectifier or the warmup delay, put a tube rectifier with the plates paralleled from the first cap to the second. The SS rectifiers and first cap take the ripple current beating, and the tube rectifier can be set up to have just the right amount of squishy by juggling the sizes of the caps.

I personally recommend to everyone who works on a tube amp with a tube rectifier in it to put two SS diodes in, one in series with each tube rectifier plate. Then if your tube rectifier shorts, you only notice that the B+ is up a bit, not that there is smoke and silence pouring out of your power transformer. That and a fused heater supply may be the two cheapest insurance investments you can make in an amp if you calculate loss prevention per dollar spent.