Dumbleator Frequency Response

[martin manning]

New technology in the house: DSO with integral waveform generator and automated frequency response plotting. A Dumbleator is right there on the bench, why not try it out on that? Easily enough, a gain and phase vs. frequency plot was produced for the whole thing, buffer stage connected directly to the recover stage, controls all the way up.

Ok, so how close can you get to that result with a circuit simulator? A basic SPICE model initially produced something similar, but the high frequency roll-off (above 10 kHz) was missing, and the treble cut shelf was about a dB higher wrt the max gain at 100-200 Hz.

The high end roll-off is most likely due to capacitance in the shielded cables running front-to-back inside the box. Hanging some parasitic capacitances representing the cables in the model fixed that, and the required amount seems consistent with the known cable properties. The difference in the treble shelf could be in the Cg-a built into the 12AX7 tube model vs. the actual tube in the Dumbleator, or just capacitance between the socket pins and leads. Adding 1pF from grid to anode on the recovery amp was enough to bring it in, and I believe this is more likely than the 220k grid stopper resistor or even the parallel 470p cap being far off spec. With those tweaks, the shape from the model is very similar to the measured response. Overall gain is still almost a dB off, but that is probably the tube model vs. the hardware.

As always: All models are wrong, but some are useful.

[xtian]

That's cool! What DSO?

[martin manning]

Keysight DSOX1204G. The exercise above is a good example of how much things that are not on the schematic matter.

[Colossal]

Keysight DSOX1204G

So much want.

Do you have that at home or work?

[martin manning]

New home scope. The ol' Tek 2246 still works fine, and I like having an analog scope around, but it was time to get digital. I usually stay away from multi-function tools, but I'm liking this Keysight a lot. Of course it has to control the generator to do the frequency sweeps, and it's a more capable generator than any of the ones I have. When do you not use a generator with a scope anyway?

[martin manning]

There has been some question about the LNFB connection as shown on the D'lator schematic, i.e. is it before or after the 220k//500p feeding the grid. A couple of people have confirmed that it is before the R//C on actual units (see Dumble files section https://ampgarage.com/forum/viewtopic.p ... 69#p140269). That configuration produces the response curve shown above, with the 2-3 dB treble cut.

The thing is, if that's true the LNFB has virtually no effect, and you have to wonder why it is there. If the FB is moved to the grid side of the 220k//500p, the low end drops ~4 dB, resulting in a treble boost as shown below, which would seem to be more useful for counteracting the effects of cable capacitance. If the 220k//500p is removed entirely, the response becomes flat, and again the LNFB has virtually no effect (lower curve below).

It's hard to say how this circuit evolved, and we'll probably never know. To see how much this matters, the easiest thing to try is shorting the 220k//500p, which is what I think I'll do when I get a chance.

[erwin_ve]

Thanks Martin,

It is at least worth a shot. Somwehere in the next 2 weeks I'll try in my dumblelator.

Erwin

[gktamps]

Keysight DSOX1204G. The exercise above is a good example of how much things that are not on the schematic matter.

Yes! Measurements are GOOD.

Particularly interesting exercise for me at this moment as I'm experimenting with a couple of different Dubleator type builds, including one a subminiature tube effects loop board.

Thanks, Martin!

[martin manning]

Sounds like fun, interested to hear about your findings.

I'll amend my plan to short the 220k//500p... Better to parallel an 8-12k to maintain the grid stopper function