tube rectifier for larger amp? (bass amp)

[lexington125]

Sunn used a tube rectifier with their famous 200s bass amp (with two KT88 output tubes). I remember reading about a few amp designs that used a pair of the 5ar4 tubes to deliver a more robust power supply. I have been unable to locate any of the sites that discussed the two tube rectifier design - can anyone point me toward any schematics or a discussion of this idea?

[pdf64]

See the Selmer PA100 http://bmamps.com/Schematics/Selmer/selmer_pa100.pdf
I remember reading that it was preferable to link the plate of an eg GZ34 and dedicate a tube to each B+ phase, and the Selmer seems to conform with that.

[Leo_Gnardo]

Sunn used a tube rectifier with their famous 200s bass amp (with two KT88 output tubes). I remember reading about a few amp designs that used a pair of the 5ar4 tubes to deliver a more robust power supply. I have been unable to locate any of the sites that discussed the two tube rectifier design - can anyone point me toward any schematics or a discussion of this idea?

FWIW there is a super size tube rectifier but I can't recall the model number. It's octal base like the others but draws 4 amps filament current IIRC. Whether you find this or double up on GZ34/5AR4, make sure your 5V winding will supply the extra current.

[R.G.]

Seriously folks, consider using 1200V/12A FREDs and series resistors/thermistors. Yes, I know that there are people who violently disagree with this, and who will contend loudly that they can hear the difference if any of the electrons in their tube amp EVER flowed through a semiconductor, but you really ought to try it and see how much silico-pollution is really there. There is a huge savings in parts, time, and power if your ears and sensibilities are not offended.

[deci belle]

Seriously folks, consider using 1200V/12A FREDs and series resistors/thermistors.

My sensibilities would be much less offended if you would provide the real-world part-numbers, RG.

Kevin O'Connor doesn't stoop to doing that for us mere mortals either.

I can figure out what jack to use from Randall Smith's hand-scrawled "Boogie" schematic, but Kevin's hand-scrawled schematic isn't helping me chose the jack for his "best effects loop copied by everybody". He told me this morning to study it some more. I just laughed.

When I can grasp your ideas, RG, I most certainly implement them, but if you are satisfied with leaving people in the dust (who are cursed with the penchant to learn by doing, (aren't electrical engineers) and who do not have their fingers on the pulse of cool parts and their sources as you are), that's not so helpful enough.

I'm an adept and mentor myself in other fields— but not in this one.

[R.G.]

Sorry about stepping on your sensibilities. I try to make it a practice to help mere mortals because I'm one of those too.

I first ran into FREDs a decade or so ago, when they were about $10 each. I had to go look up the current state of affairs, and was pleasantly surprised. You can now get a 1200V/10A FRED (Fast Recovery Epitaxial Diode) for $1.67 at Mouser, and they are in fact in stock. $3.50 for some very manly rectifiers is a good deal.

Here is one link:
DSEP8-12A FREDS at Mouser.
I selected this one for "enough" and "cheap". Since they first came out, the field of fast and ultrafast, soft-recovery diodes has exploded. There are probably hundreds of part numbers, so I don't keep the part numbers current, I look them up at Mouser as I need them.

[loudthud]

See the Selmer PA100 http://bmamps.com/Schematics/Selmer/selmer_pa100.pdf
I remember reading that it was preferable to link the plate of an eg GZ34 and dedicate a tube to each B+ phase, and the Selmer seems to conform with that.

I remember seeing something that warned against using one tube for each phase. The idea was that although the average current was the same, the peak current was too high and it stressed the cathode. Early versions of the Sunn 2000S (also 1000S and 1200S) used two GZ34s in parallel. After about 1971 silicon diodes were used.

Link: http://el34world.com/charts/Schemati...Sunn_2000s.pdf

[Jazz P Bass]

Since they first came out, the field of fast and ultrafast, soft-recovery diodes has exploded.

I guess you can thank the extensive use/ design of switch mode power supplies for this.

[deci belle]

Haha~ Yay! Thank you, R.G.!

I've used the axial FREDs for Fender-style rectification so far, but they might not have the same performance and values as the ones you spec'd.

I like when I've made amps with two rectifier tubes (w/ optional switching to SS). Two 5U4GBs or GZ34s are cool!!

[stevehoover]

RG...

Are the DSEP8's any "quieter" than U5408's that I used on my last build?

Besides the TO 220 mounting, any thing else that I should be aware of?

Thx, Steve

[R.G.]

Haha~ Yay! Thank you, R.G.!

I've used the axial FREDs for Fender-style rectification so far, but they might not have the same performance and values as the ones you spec'd.

Well, don't go reading too much into it just because I found a part number. That choice was made on (1) they are FREDs and (2) they're high enough voltage and current and (3) after that, pick cheapest.

I looked at the data sheets and the recovery numbers look about right, and they do maunder on about "soft recovery characteristics", but none of these diodes will specify "slam-off noise" which is really what causes the problem.

So my choice is just an uneducated pick from the right (as I understand it) batch of parts. I don't know that they are any better than the axial FREDs you have already used.

The real answer here is probably to use FREDs now that they're not exorbitant to use, and then to do a listening test. Even ordinary "slam-off" diodes can be quieted by doing a good job of snubbing them, but this is tough because the snubbing has to be matched not to the diodes, but to the diodes' resonance with the parasistic inductance and capacitances of the wires connected to them.

The actual noise mechanism, as I understand it, is the abrupt current shutoff when the recovery carriers are finally swept out of the diodes' active region acting like a pulse excitation to the L-C parasitics. These ring and make a squark of RF noise that then gets back into the circuit by radiating to the active parts. The soft recovery diodes reduce this by not hitting the resonances with as much broadband noise from an abrupt change in current.

Snubbing actually does much the same thing but by eating some of the turn-off energy before it hits the LC parasitics. It's like hitting a bell with a chopstick instead of a sledge hammer.

Are the DSEP8's any "quieter" than U5408's that I used on my last build?

Besides the TO 220 mounting, any thing else that I should be aware of?

Can't say, for the same reasons. I have not gone off and hooked up some diodes to see how they emit, just grabbed what ought to be better than ordinary rectifiers from a family of quieter diodes. The general wisdom is that fast, soft recovery diodes are in general better for this use. None will be perfect, but all of the fast, soft ones will be quieter than ordinary-duty diodes like 1N4007s, or the old originals in SS rectified Fenders and Marshalls.

Other than TO-220 mounting and the usual cautions that this stuff is flying around at hundreds of volts, no. The TO-220 package can usually get rid of about 2W in free air, so I suspect that you won't need heat sinks, but put them where you won't accidentally touch the tabs on the diodes. Nasty shocks would result.

[trobbins]

RDH4 makes the comment of preferably paralleling the two diodes in a dual diode valve and using that for one side of a full-wave circuit. But there is no justification provided. I can't see any discussion on preferences in more detailed tech docs, except for RDH4's comment on using added series resistance with each diode, so as to reduce the opportunity for current sharing imbalance, especially for low resistance diodes such as mercury arc diodes.

I can't rationalise any benefit one way or the other except that it is common for a full-wave secondary power transformer to have slightly different DC resistance in each half-secondary, and as such it may be of some benefit to mix/match the valves to balance the net resistance of each rectifier arm, and hence attenuate 50Hz ripple across the filter cap - which would be easy to do when paralleling the two diodes in each valve package.

At least the Selmer circuit has a secondary CT fuse, but the preference is to fuse both secondary winding arms and not the CT when also generating a fixed bias off the HT secondary.

[loudthud]

I'm not sure where I read what I said above, but I found this:

From the RCA Receiving Tube Manual RC-26 Page 18

Parallel operation of rectifier tubes furnishes an output current greater than that obtainable with the use of one tube. For example, when two full wave rectifier tubes are connected in parallel, the plates of each tube together and each tube acts as a half-wave rectifier. The permissible voltage and load conditions per tube are the same as for full-wave service but the total load handling capability of the complete rectifier is approximentaly doubled.

When mercury-vapor rectifier tubes are connected in parallel, a stabilizing resistor of 50 to 100 ohms shoule be connected in series with each plate lead in order that each tube will carry an equal share of the load. The value of the resistor to be used will depend on the amount of plate current that passes through the rectifier. Low plate current requires a high value; high plate current a low value. When the plates of mercury-vapor rectifier tubes are connected in parallel, the corresponding filament leads should be similarily connected. Otherwise, the tube drops will be considerable unbalanced and larger stabilizing resistors will be required.
　
Two or more vacuum rectifier tubes can also be connected in parallel to give correspondingly higher output current and, as a result of paralleling their internal resistances, give somewhat increased voltage output. With vacuum types, stabilizing resistors may or may not be necessary depending on the tube type and the circuit.

[R.G.]

The problem with paralleling vacuum rectifiers - well, semiconductor rectifiers, for that matter - is that the lowest resistance/lowest forward drop one hogs all the current. With vacuum devices, this exceeds the current density the cathode can provide in good order and starts degrading the cathodes. When one dies, the other is left with all the current and it starts dying. Resistance in series with the diodes forces them to share current to some degree, and there are methods to calculate the degree of sharing.

Using one two-section rectifier tube gets two made-at-the-same-time cathodes running in parallel, analogous to using two monolithic semiconductor junctions that are well matched, and helps minimize the voltage and resistance differences.

Frankly, the guys in the Golden Age would have gone to semiconductor diodes in a heartbeat if they had them.

Given that semiconductor diodes with forward-drop-faking resistors are anecdotally so good at faking the results of a vacuum rectifier, why not...

Ooops, sorry. I'm ranting again.

[trobbins]

It would be interesting to see a simulation of a full-wave CT rectifier HT circuit where the DC secondary winding resistances were mismatched, as normally occurs in practice. Of the PT's I've measured, the DCR mismatch has been below 10%.

The effect of that PT DCR imbalance would be worse when using semiconductor diodes, although I haven't come across anyone taking measurements of the ripple at the fundamental frequency, and whether some padding resistance in one diode arm can be used to null out that fundamental.

Of course PT DCR is not an issue for doubler or bridge rectifier circuits where the PT secondary is a single winding. Doubler circuits were prolific for larger powered vintage PA amps, so on-resistance variation between the two valves used could similarly cause imbalanced ripple.