Selecting valves for a 12 valve PP output stage

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

Bjorn Zetterlund

Well-known member
Joined
Jun 3, 2004
Messages
123
Location
London, UK
I just bought a Mesa Boogie 400+ Bass amp, and it needs all new valves. The output stage is 12 5881/6L6 valves in a push pull configuration. This amp (and all Mesa Boogie amps) has no bias adjustments, and it needs well matched Mesa Boogie valves to achieve best performance...no bias adjustment and properly selected valves is one of Mesa Boogie's main principles, if you like. There is a whole essay about it on www.mesaboogie.com The output stage is extremely simple; -63 V bias to grid via 2k7 (5W) resistor, with cathode directly to ground and the plate directly to the primary of the output transformer. The centre tap of the output transformer is at 540 V. The control grid input is via a 2k2 resistor for each valve.

I would like to learn more about how to select these output valves as one matched pair in the UK costs 50 pounds...with 6 pairs, that becomes a major expense!

My first query is; in this type of amplifier with many valves in parallel, how much of a benefit is there in having all valves matched as opposed to having six matched pairs, or simply having the idle current match overall? Specifically in my case, how well behaved (or not) would this output stage be if I simply got 12 valves and mixed and matched to get equal 'push' and 'pull' currents at idle?

Secondly, on a more practical level, what is the best way of measuring the idle current of any individual valve? I was thinking I could make a short 'valve socket extender' cable to enable me to measure the current at the cathode with a normal multimeter. Another possible solution would be to add precision current sensing resistors on the cathodes. You could even have a 12-way switch with a test point at the back to enable quick measurement of individual valves...would be nice to have to quickly diagnose a faulty valve.

Any and all thoughts on this are most welcome and much appreciated!

Thanks,

Bjorn
 
Wow. 12 tubes. How do you know they need replacing?

If you don't need all that power, you could change your speaker impedance to reflect half the load to the output primary and then use 6 tubes instead.

540 volts is scary. I do not know if all 6l6's will tolerate that. Unless there is a big chunk eaten by the copper loss in the output.

You could mod the power supply to drop the B+ a bit. And add a bias adjustment. What kind of bass style do you play? A huge cathode resistor or two might be a neat mod. You would need a giant bypass cap for bass.

No wait, this is push pull. Does degeneration cancel?

Thinking..........
 
I have been thinking about this for quite awhile now, tube matching not 12 tube outputs. I did a few checks on a brand new pair a matched Svetlana (real svetlana not EH) 6L6s, running in a silvertone 1484, I checked the bias of the two tubes seperately ever so often, over a couple months, after about 2 months the tubes had drifted quite abit from being matched, my ears could not tell the difference but my meter could. I do not recall how far off they were as this was awhile back. I assume all tubes drift, and this is why some places burn in their tubes before matching them, but does that really help. How much should a tube drift and are matched pairs just hype? I am about to put a NOS matched pair of RCA 6K6GTs into an amp I built, I think I will keep an eye on them and see how well the old tubes stay matched over time and use, and maybe experiment with how far off tubes can be before cross-over distortion is actualy an audible issue.

adam
 
I think tube life is another reason for matching. For some reason, they burn out quicker if one tube is doing more work than the other.

With the 12 tube thing, I would think that mis match erros might tend to cancel. Kind of like the 8 triode blocks in the Fairchild 670.
 
Why would the other tube work harder, it is not being told that it and its matched friend is going to make 50 watts of power, the tube is just being told that it has 48mA, 400 volts and 4000 ohms to dump all that plate swing into. Am I missing something? Push-pull amps are somewhat new territory for me when it comes to the theory, but I really do not see why one tube is going to work any harder then how it is set. Sure if there is a massive mismatch and you bias it according to the tube that wants 50mA and the other tube only can take 25mA, but I can not see many times that would be an issue, in my somewhat limited experiance new tubes especialy are all pretty close, with the exception of the occcasional suicidal tubes. They seem close enough that you could just bias to the median of the two tubes needs and you will be good to go.

adam
 
Hi CJ and Adam,

According to the schem, the actual voltage at the plate is 500 V suggesting that the primary drops 40 V. I'm not sure if that's right as the schem is a bit difficult to read. Have to add also that the schem I have is for the 400, not the 400+, and it shows only six output valves. One of the selling points in the manual is that each valve handles comparatively small currents, enhancing tube life. Gotta run to work, but will check in tonight...thanks for the replies!

Bjorn
 
Despite my recent questioning of the operation of push-pull output stages there are a few things i can tell you. A 1 ohm resistor from the cathode to ground on each tube is a great way to see the current, ohms law tells us that voltage drop through a 1 ohm resistor is the same as current through the resistor. I can also tell you that needing to buy "Special matched Mesa Boogie" tubes is are marketing scam. I reccomend Svetlanas 6L6s for current production tubes. I am not sure about the UK, but here in the US if you want real Svetlana tubes you need to buy what we call Winged C tubes, New Sensor (sovtek/EH) got the US trademark on theSvetlana name and all tubes sold as Svetlanas are actualy sovtek, So if you get them make sure on the names over there. Also you do not need 12 matched tubes, 6 matched pairs will do fine, cross over distortion only happens in the actual pair and not between pairs. You just need to look in the amp and make sure that you put the pairs in the right sockets.

adam
 
I think CJ is correct, but remember I am a mere novice. A 12 tube push pull output is 6 parallel tubes (group A) in series with 6 more parallel tubes (group B). All the tubes in group A are running at the same output voltage & all the tubes in group B are running at the same output voltage. Theoretically these 2 group voltages should be the same. Total current for group A = total current for group B. Consider what happens if a couple of tubes in group A get weak. The remaining tubes from that group are now working harder trying to match the series current of all strong tubes in group B.

Is this right?
 
But what would make the tubes in group B relise that group A is not working up to par and that they now need to work harder. I am trying to figure out why this would be either way, but its just not happening on paper. Once I get my amp finished I can experiment with this abit, but I can not finish it untill some parts show up


adam
 
Sounds like hype to me....
If you were going for a pristine bass sound (I am struggling to think of a situation you would use this for.... if you were after that you would DI the thing) then I could understand...
But Mesa Boogies !!
Turn up the gain and plug in a bass ...As long as the 6L6s weren't microphonic it shouldn't be a problem...
Or am I missing something
 
Mesa "select" their tubes i.e. they buy bulk and measure plate current at the intended operating point. Then they grade them by magnitude of bias voltage needed to obtain proper plate current corresponding to ca 60-70% of max plate power dissipation. That's all there's to it. Contrary to Mesa's recommendation a lot of people are getting their amps converted to adjustable bias. A not too complicated mod that an experienced amp tech shoul be able to implement. This will ofcourse void any warranty that you may still have on the amp.

Alex
 
Have you tried Watford Valves shop?

They are great - some of the matched 6L6's are only£9 too.

http://www.watfordvalves.com/mainindex.htm
 
Actually, after thinking about this, tube matching is a good idea. The reason is that when you have that much power, and the tubes are mis-matched, the amp will tend to self destruct. The reason is that if one tube is off, it will be passing a lot more current than it should be. And there is no way to tell which tube it is unless you have those 1 ohm resistors in every cathode. And the 1 ohm figure shoud be exact.

The Trainwreck guy, (forgot his name) recomends taking your volt meter to the shop and going thru a bin of resistors to get a good match. If the voltage across the resistor is off 0.01 volts,this corresponds to 10 ma difference thru the tube, which is lot, considering that each tube will only be pulling about 30 to 40ma. You need a good meter with enough digits also.


This tube mis-match problem is why a lot of Ampeg SVT's tend to self distruct after a fairly short time.

It is also a good idea to balance the phase inverter so that each side is being driven equally. This could be done with tube swapping, or possibly plate resistor tweaking. Someone out there probably has a better method.
 
> I just bought a Mesa Boogie 400+ Bass amp,... 6 pairs, that becomes a major expense!

You can't drive a Ferrari, or a Mack truck, with a Toyota attitude to money.

Any way you slice it, a 400 watt tube amp needs a LOT of GOOD tubes.

> it needs all new valves.

Why do you think so?

I made a living on some Bogen 300 amps. They pushed the tubes hard. I got the amps from a video facility where they had been powered-up all the time and working full-power several hours a day, for over 10 years. One caught fire (a bad solder joint finally let go) so they were replaced. Of the 20 tubes, 18 were original Bogen and 16 worked perfectly well (and I did a LOT of testing to upgrade the amp sound, so I woulda spotted a goofy tube).

I did recently find a gassy 6L6G but it was from 1939.

> -63 V bias

What is the screen voltage?

But any way I slice it, that looks like near-cutoff. It is going to have crossover distortion. Problem is, near-cutoff is the most variable part of a tube's curves: one tube may give 10mA and another give 40mA for the same grid bias.

> how to select these output valves

Buy a couple hundred, test, sort-out 12 very-similar, get rid of the other 188 tubes. Matching only makes sense if you have LARGE quantities of tubes available and different needs for the different characteristics.

Put in the 1 ohm resistors. I would bus 6 resistors, then run another 1 ohm 2-watt from bus to ground. That way you can measure each tube and the whole bank of 6.

If you must use wire-wound: do not let the plate leads get anywhere near the cathode resistors. Makes a lovely radio transmitter, except so mis-biased that the tube melts.

I would not putz around with 0.01% resistors. Tube work is never that accurate. Use 5% resistors, buy 30, test, use the best-matched 14 of them. (I would just use 5%, but some folks love to play with ohm meters.)

Measure the tubes you have. Check at idle. If you can rig a 400 watt(!!!) dummy load and tone generator, test again at 200W-400W output and see how different the DC currents are.

If it sounds like crap, you may have a couple sick tubes. I have one sicko in my quad-8417 Dynaco. Put a 16 ohmn speaker on the 4 ohm tap, run two tubes at a time (one each side of the output winding). After a few swaps you'll have found the stinkers.
 
And the reason for no bias adjust is that the SVT's used to fry the bias pots all the time. When a tube shorts out, it takes the bias pot with it.

You can put in "fuses" in the form of low wattage resistors on the plates if you want to keep playing after a tube lets loose.
 
Thanks for all the replies!

Firstly, my reasons for suspecting the health of the valves were initially purely for mechanical reasons; on nearly all the output valves, the glass envelope had become detached from the plastic/bakelite (?) valve base, and were basically free to 'flop about'...I didn't think that was something worth fixing with glue. I tried it with a tone generator and scope with a 4 Ohm dummy load whilst observing the 600 Ohm balanced out, which is a feed that is connected to the speaker output via a pad. I don't have a pad for the scope, and it was just faster for me to measure that way. In any case, the manual states that this output should give 2 v peak to peak as the output approaches clipping, but in the case of this amp one side of the waveform started flattening out at only 1 v p-p. The clipping was extremely unbalanced, and I assume that a balanced push-pull should clip symmetrically? I haven't checked the balance of the phase splitter yet, mostly since due to the mechanical problems with the output valves, I had already made up my mind that it needed a re-valve. The amp also gives out some hum, and the odd bit of crackle and spurious noises, which I'm sure could equally well be from the small signal stages. One of the output valves is very microphonic as well; seems to resonate at about 200 Hz or so when you tap it. Before I forget, the screen sits at -62 and the other grid is at -52...the schematic is a bit blurry and these are not measured values.

I have no problem spending the money on good matched and tested valves as long as it is money well spent and not just extra money on the mesa boogie name brand, but this place is great for finding out what the real deal is. Also, I wondered about what the basic criteria were for an output stage with so many valves, and if like CJ suggested, the errors tend to cancel out. In the solid state world, the LM394 supermatched transistor (as used in the 990 op-amp input stage), two groups of 25 transistors are wired in parallel to create two very well matched devices...but obviously a valve power amp is a very different beast.

I think my approach will be to re-test this amp with known good valves. A guitar-player friend of mine has 5 sets of matched mesa boogie 6L6's, and I'm sure he'll lend them to me for testing purposes. If it behaves properly I will most likely add the cathode resistors as suggested, and then measure the current of the known good valves and compare these to the old ones and take things from there.

Anyway, thanks for all the help with this. Quick responses to various questions; I play reggae to heavy rock mostly, with loud drummers and sometimes even louder guitarists! The output stage is completely devoid of feedback. Thanks for the watford valves link; I used billington export last time I bought valves, and they are very good too.

Always a pleasure to talk things over with the lab! :sam: :sam: :sam:

Bjorn
 
Don't worry about the tubes being seperated from the bases. This is very normal, along with the locater pins being broken off.. The vaccum will not be affected as the wire leads coming out of the glass beneath the base are brought thru the glass and sealed.

Minus 62 on the screens? Hmmmm. Maybe they grounded them to the cathode to run this thing in triode. I think the supressor is connected internally on the 6L6GC, maybe you measured the grid by mistake. Pin 4 is the screen.

Be careful, right next door is the plate at pin 3.

Be careful with the scope if you venture into high voltage land.

I would use the speakers for the scope test.
Dummy loads are fine for power testing, but you would probably be better off with speakers for a bias setup.
You need that reactive load on there.

The dummy load will look good at one freq, then be off on the next freq.

Here is the Jack Darr link again, just in case....

http://www.diyguitarist.com/Misc/J_Darr.htm
 
> the screen sits at -62 and the other grid is at -52

Not at all reasonable. G2 "Screen" has to be over 250V to get that kind of power out of 6L6. G3 is usually not pinned on power tubes; in any case it straps to the cathode (unless you are building a TV set and it has snivets).

G2 could be tied direct to +540V, but that's a bit high. It could be tapped down a voltage divider resistor string, around 400V. It could be tied to the plate for triode duty, or an OPT tap for UltraLinear, but not at 66 Watts/pair.
 
Pardon my sloppyness with the terminology and pentodes in general...

Having had a second look at the schem: the plate supply at the primary centre tap is 540V. G2 is at 535V (after a filter choke on the supply; the 540V is sourced at the main smoothing caps). The grid (as in the one with the audio input, pin 5 on the valve) is at -62. No other connections are show apart from the plate to the transformer primary, and the cathode to ground...sorry for the confusion :oops:

Bjorn
 

Latest posts

Back
Top