hacking kit JTM45 bias circuit

[mozz]

"Sorry but it doesn´t work that way."

But the rectifier tube is before the filter caps. Chances are your bias voltage will be there first even if you used 1000uf. Whether the rule of thumb is correct is moot. Anyway, best to be true to the original schematic if you want the JTM45 sound. If you have a 8uf or 100uf on your bias supply, you aren't going to hear any difference. Only thing with the JTM45 is the voltage on the presence pot, but there is a update/mod for that and sometimes they squeal with the presence and treble all the way up.

[trobbins]

My concern with RC time constants is more related to any fusing of the primary or secondary side of the power transformer. Although the output stage valve plates could at worst take a short pounding, it should be short in terms of their thermal time-constant.

Fusing could be affected as the initial surge current could exceed the typical fuse capability for surge amplitude and time duration for a hot turn on event.

The RC time constants worthy of note are for the bias supply, the B+ supply, the L/R output stage valve path, and the coupling caps to the output stage. Both the charge and discharge time-constants are implicated for a hot turn-on event of circa 1 second in the off-state, as that could substantially discharge the B+ supply and the bias voltage, but be short enough to cause both output stage valves to conduct according to their B+ and bias levels without any cathode thermal delay.

B+ levels usually rise in just a few mains cycles (ie. circa 100-200ms). Hi-fi output transformers and output stage coupling cap values can cause long time-constants, but are likely to be << 1 sec for an instrument amp. That bias circuit with 180k and 8uF stands out as the dominant time-constant, as it is well over 1 sec, and along with the coupling cap charge-up, could keep both output stage valves in the grid-conduction region for a few hundred ms (which is long for fuse timing).

[Dave H]

Your bias cap only has to charge up to like 50v to be enough to move on, but it is doing that through a 180k resistor. I would have to think that resistor rather than the relative cap size is what slows it down. But then one has to wonder how many seconds does it take the bias to charge up to tube protecting levels, compared to the 10-20 seconds it takes the tube heaters to allow current to flow.

It takes a few seconds. The plot below is for 180k and 2 x 22u.

[Chuck H]

What I'm thinking is that there are many bias supplies that use a single cap. In that case you'd be hard pressed to mitigate ripple on the half wave rectified bias supply even with a 100u cap. With the two cap bias you can do better. The point here is that a bigger single cap probably isn't going filter enough ripple to make it worth while. And besides, we're counting on the output section balance to cancel most of any hum anyway. And Mike's intent here is to achieve good balance which will only improve that. So the cap value really only needs to be big enough to filter gross ripple and decouple LF below audio to avoid any LF signal interaction via the grid loads where they meet atop the bias supply IMPEDANCE. 8u is plenty for that. There was a guy here for a while (SN rhymes with moundsurugan) who insisted again and again that increasing the bias cap value improved low end. Hogwash. I think there are a couple of Mesa designs where this may marginally be true, but not at all for 90+% of guitar amps.

I guess my point is that a single 100u cap won't filter the ripple from a typical bias supply anyway and two would stretch the charge time to where it may need to be considered. 8uf is big enough to decouple. I say leave the supply design alone. It works well like it is.

[Enzo]

I think we have two concerns going here. One is worry that the bias voltage will not be present fast enough to protect the power tubes from excess current. I think that is a non-issue, as the bias voltage comes up in a second or two, and the power tubes take longer than that for the heaters to come up. And even so, a second of double current wouldn't hurt the tubes anyway.

The other concern is sufficient filtration to prevent hum. One suspects the original dual 8uf design was sufficient in this regard.

[bob p]

I guess it depends on how you define "sufficient," and whether your objective is to repair an amp, build a faithful clone of a classic amp, or design a better amp.

The world is full of noisy amps. I hate that in an amp, and when I'm building I'll do all sorts of little things to try to make noise go away. I'm even willing to do some of those diminishing return things, because I just don't like noisy amps.

The JTM45 schem snippet linked above shows a pi filter, not a single cap filter. A pi filter by it's nature is going to provide far better ripple reduction than a single cap filter, and an 8uF+8uF pi filter is going to provide good ripple reduction. But a pi filter using a pair of 100uF caps rather than a pair of 8 uF caps will have markedly less ripple on it. To see this, just fire up your favorite modeller. the improvement is significant.

Given that caps are so cheap today, I don't see a big downside to going with a larger value. In a market that's not geared to producing low value caps for obsolete tube amps the 100/100 was cheaper from my supplier. That's a win/win scenario for the builder who doesn't feel bound to building to a 50 year old design, just to have authenticity and all of the problems that come with it. I've got nothing against building faithful clones, but at the same time I don't think it's heresy to tweak a classic circuit to improve it.

[Enzo]

I completely agree that larger caps likely will reduce hum. The original design passed the Marshall muster, whatever their standards were.

[bob p]

I would have expected the standards to improve somewhat over the 50 year span where we've seen the prices of caps reduced so dramatically. But I guess the standards haven't changed as much as I had thought whey would.

looking at the JTM45 we see that 50 years ago Marshall used a pi filter with 8 uF caps. Today I'm looking at a DSL40C schematic and they're using a 22uF and and 8 uF cap in the bias supply pi filter. Contrast that with Fender, who is currently using a single 100uF cap in the DRRI. I know it's not a perfect comparison, but it does show that the big manufacturers are thinking like Chuck -- a little bit is enough, and if that helps to contain cost then that's good enough for the MI market.

getting back to the OP, this thread is about hacking the circuit to improve it, so why not make it more quiet, as long as we don't blow up his amp?

[Chuck H]

Ok... On Bob's suggestion I simulated some stuff.

Using a circuit similar to what's been proposed above (120k series, two cap filters divided by a 15k, 47k load) I get:

Pair of 8uf get within two volts of final bias in two seconds and has 30mV of ripple

Pair of 22uf get to within two volts of final bias in 5.5 seconds and has 4mV of ripple

Pair of 100uf take 24 seconds to get within two volts of final bias (and still climbing slowly beyond a full minute!) and, of course, has no ripple worth reporting.

I'd go with 22uf

[J M Fahey]

"Sorry but it doesn´t work that way."

But the rectifier tube is before the filter caps. Chances are your bias voltage will be there first even if you used 1000uf. Whether the rule of thumb is correct is moot. Anyway, best to be true to the original schematic if you want the JTM45 sound. If you have a 8uf or 100uf on your bias supply, you aren't going to hear any difference. Only thing with the JTM45 is the voltage on the presence pot, but there is a update/mod for that and sometimes they squeal with the presence and treble all the way up.

Oh, I was being a Devil´s advocate and taking the "worstest" case, using instant-ON silicon rectifiers; of course using hot cathode rectifiers which add a thermal time constant (still shorter that main tube cathodes´ because they are usually direct heated) the situation is even safer for power tubes.

[J M Fahey]

I think we have two concerns going here. One is worry that the bias voltage will not be present fast enough to protect the power tubes from excess current. I think that is a non-issue, as the bias voltage comes up in a second or two, and the power tubes take longer than that for the heaters to come up. And even so, a second of double current wouldn't hurt the tubes anyway.

The other concern is sufficient filtration to prevent hum. One suspects the original dual 8uf design was sufficient in this regard.

And even if barely so, bias ripple is applied equally to both halves of a balanced amplifier (any push pull configuration) so it mostly self cancels.

[trobbins]

For those with just PSUD2 software - it can also simulate this type of bias circuit and filtering (and very easily too) - but with the proviso that it will only allow a 'positive voltage' bias supply circuit!

[Chuck H]

For those with just PSUD2 software - it can also simulate this type of bias circuit and filtering (and very easily too) - but with the proviso that it will only allow a 'positive voltage' bias supply circuit!

That's what I used. So... There is also the "proviso" that PSUD2 won't allow you to plug a RC filter in front of the C filter for a half wave rectifier. So you have to plug the series resistance (120k for my example) in as the transformer winding resistance. Otherwise it's ltspice. Which works fine, but isn't quite as fast and sometimes I have to dick around because I can't remember how to get it to sample and show the specific results I'm after.

[trobbins]

There is also the "proviso" that PSUD2 won't allow you to plug a RC filter in front of the C filter for a half wave rectifier.

I was able to select an RC stage after a half-wave ss stage - I have V2.0.4 - are you able to check again, it may just be the version.

[Chuck H]

I was able to select an RC stage after a half-wave ss stage - I have V2.0.4 - are you able to check again, it may just be the version.

I did try twice. Shut it down and started over. I was surprised because it does let me do that with a full wave.?.

EDIT: Ok, I just tried it again with the same result. So I pulled a trick on it. I started with a full wave recto, used the RC filter in front and then changed the recto to a half wave and it took it. My results were the same as above. And I discovered a way around the glitch with whatever "my" downloaded version of this excellent program is