Level Shift in CF stage (after LFO stage)

[tubeswell]

I posted this problem over at the MEF but everyone is too preoccupied with xmas /NY over there and amp building just can't wait

There's a lot of thorny questions - sorry for the rambling post

This amp (I built a little while back) is going through trem tubes at a fantastic rate of knots. I've been through 3 trem tubes in the last 3 months. Each time the symptoms are the same, it starts with a popping noise when I switch the trem (footswitch) off, and then eventually (quickly) builds up to a staticky thump that's in time with the speed control. And it is cured when I pop a new trem tube in, which lasts for okay about 20 hours, and then the same thing happens over again.

Rather than elevating the heaters some more and possibly risking endangering other tubes, I thought implementing the level-shift in the CF part of the LFO circuit would get the cathode voltage down to a 'safe' level. I have no idea how much you can level-shift a DC coupled LFO and CF and still have it working reasonably well, but I assume that as long as one engineers it to keep the CF voltage swing potentially bigger than what you need to affect the output tube grid bias voltage, it will work. Is that right?

I had a word with Simcha Delft (who lives locally here in Wellington), who recommended I try the level shift. But its a matter of trial and error a bit. I have some fundamental appreciation of concepts and Merlin's book is helpful.

At present I have the 6G9B bias-vary trem circuit - the CF is direct coupled to the LFO stage and the CF cathode is about idling at about 220VDC (with an HT of 426V and a cathode resistor of 470k), and the heaters are elevated to 43VDC.

By my reckoning with the existing DC coupled pair (using Merlin's cook book as a guide):

Quiescent plate current of LFO stage = (426-213)/470000 = .00045A (.45mA)

Current through LFO Rk = 2.03/4700 = .00043A (.43mA)

.02mA of current is 'missing', and the bias on CF stage is:

220-213 = -7V,

so the 'missing' .02mA of current (big deal) is being added to the current flowing through the (existing) 470k Rk of the CF.

No doubt this changes when you put a level shift voltage divider in between the two stages.

I was wondering what the whole effect would be on the characteristic form of the LFO wave compared to what you would get out of a 'normal' DC coupled pair. With the latter, you would get a more asymmetric wave - because the CF Rk steals current from the LFO plate one one side of the signal swing.

However with this level-shift idea, it is introducing another route for the current to take to get to the LFO plate.

I do wonder about how a level shift will affect the characteristic bias balancing act that otherwise takes place in a gain stage-CF DC-coupled pair. The introduction of a voltage divider in between the two stages has to have an effect on how much grid current will be 'stolen' from the first stage's plate resistor by the CF stage's cathode resistor (which takes place in a typical DC coupled pair), and it will no doubt affect the CF bias.

The components in particular I am thinking of changing/introducing are shaded in green and I have put questions marks after the component values and question marks after the prospective voltages.

I haven't done a proper analysis of the cathode resistor value required to get approx 85VDC. But my reckoning is that with the LFO plate at 213VDC, I need to knock the grid of the CF stage down to about 77VDC - which is 36.5% of 213 - and estimating that the cathode would need to be biased at about 8V above that, I reckon I'd need to aim for 85V on the cathode, but I didn't yet go to the step of calculating the cathode resistor needed to get that with an HT/CF plate of 426VDC, (I think that it should be around 180k to get 85V at the CF cathode to get a quiescent bias voltage of around -6, -7 v. with the grid voltage at 77V)

Also I'm not sure about the build-out resistor - so I tentatively put that at 100k. The slope/build-out resistor is currently 220k, and I'm guessing it may need to go down to 100k with the level shift implemented (because I'm anticipating there will be less signal swing)

The impedance bridging is probably inadequate? given there is 470k plate resistor on the LFO stage, and the ra there is about 62k(? - well I'm not sure about that ra value - since the stage is biased a little on the cool side and there is a 470k load line). (If I opted for an AC coupled CF stage instead, I suppose bootstrapping the CF stage's grid would increase the input impedance there, so that is always another resort).

The grid leak/lower part of the voltage divider in the DC coupled level shift in the schematic is 270k, because that works with the 470k upper resistor to give the CF grid about 36% of the voltage from what is at the LFO plate. (I think this gives a total of 740k - in parallel with the 470k plate resistor for the LFO stage - as far as AC is concerned. Somebody please correct me if I'm wrong - because I intuitively suspect the bypass cap in parallel with the 470k would alter this to some extent).

If this 270k needs to be something like 470k, then I would need to make the upper resistor 820k to get the same voltage division (which I think would then make 1.3M in parallel with the 470k LFO plate resistor - as far as AC is concerned - when I go to determine the output impedance of the LFO stage. Again somebody please correct me if I'm wrong). I guess that if the total resistance of the level-shift voltage divider is increased from 740k to 1290k, that would slightly affect/reduce(?) the overall output impedance of the LFO stage.

I guess I could use something like 4M7 and 2M7 in the interstage voltage divider to get the level shift mentioned in 1) above (i.e.; if the impedance bridging with 470k/270k divider isn't enough)? Please correct me if I'm wrong.

Also I haven't calculated the proper freq for the bypass cap in the voltage divider yet - .1uF is a stab in the dark. Any opinions welcomed.

I was thinking another option might be to take the HT for the whole LFO/CF circuit from a separate 10k supply resistor (that would be in parallel with the 10k supply resistor after the screen node) and use another 20uF to decouple that. That would at least lower the HT to around 360 for that circuit. Anything has got to be better than continually forking out for trem tubes

The attached schematic also shows a couple of AC coupled variants - fixed bias vs cathode biased. I was also wondering about the effect of using AC-coupled CF stages after the LFO.

Can anyone please comment on the the effect on the trem wave-form from using AC coupled CF (as opposed to DC coupled)?

Sorry if I've repeated myself several times in a rambling kind of way

Cheers

[Jana]

If it was me, I would try dropping the B+ feeding the entire LFO circuit. That 426 volts. Put a 3 or 5 watt resistor in and also bypass it with a cap. Not sure of the value, maybe start with a 10K or 20K and see how much it drops. I would drop it as much as possible and still have the LFO working. My initial guess would be to drop that 426 down to about 300 or so and see what happens.

[Merlinb]

The only serious effect level shifting should have is to reduce the amplitude (depth) of the trem signal, because you're decreasing the gain of the oscillator a bit. Shouldn't be a big deal though.

Before you do that, though, I would try adding a grid-cathode protection diode to the CF, as grid-cathode failure is not uncommon in DC coupled CFs. (What voltage have you elevated your heaters to?)

Here's a picture of what I mean. Because this is only a trem section, you don't need the 47k resistor, just the diode.[img:200:235]http://www.freewebs.com/valvewizard/Cat ... lower4.jpg[/img]

[tubeswell]

Thanks Merlin

The heaters are elevated to about 40V. (It doesn't show this on the schematic). I was wary of going higher because of possible problems with other tubes, and I only have 100V 47uF caps available where I am (unless I go to 450V caps).

I will try that trick with a diode thanks. (That didn't occur to me even tho' I read it in your book - Duh )

[Merlinb]

I will try that trick with a diode thanks. (That didn't occur to me even tho' I read it in your book - Duh )

Also, I forgot to say that if you still have problems then it will probably be better to convert the CF to an AC coupled version, rather than use level shifting. This would still allow you to run the CF with a somewhat lower cathode voltage, by choosing an appropriate cathode resistor, but it would retain a very high input impedance.

[tubeswell]

Thanks (again) for that Merlin.

I have a couple of AC-coupled options shown on that scheme. I guess the cathode-biased, bootstrapped version is superior as far as both the biasing consistency and the input impedance is concerned. Are those resistor values appropriate do you think?

I was also wondering about whether there would be any problem with excessive grid current/blocking distortion if I used an AC-coupled CF. (Or is that concern somewhat mollified by the really low freq we are talking about? - i.e.: so that the cap discharge time is not really an issue??)

[Merlinb]

Are those resistor values appropriate do you think?

they look fine to me.

I was also wondering about whether there would be any problem with excessive grid current/blocking distortion if I used an AC-coupled CF.

Blocking is a transient phenomenon, you can forget about it for a sine-wave trem!

[tubeswell]

FWIW I changed the trem CF stage to an AC-coupled variant (See schematic)

Curiously the idle voltage with the 180k rk on the CF stage was something like 230V, and the trem sounded mighty fine like that (with a 100k 'slope' resistor).

However that wasn't achieving my objective (of lowering the h-k voltage). So I ended up changing the rk to 100k and that brought the CT voltage down to 172. With the heaters at 40V, this makes the h-k voltage about 130, so we'll see how that goes. The trem is obviously not as intense this way, but maybe the 'slope' resistor can go down quite a bit further yet (47k at the mo').

(Incidentally at Merlin's suggestion, I tried adding a 1N4007 from grid to cathode before I changed the CF to an AC coupled version, and the trem didn't work. Don't know what I did wrong. Merlin said to install it without any resistor in series, so that's what I did. Never mind - undone that mod now)