Strange input loading effect.. or not?

[exclamationmark]

I'm currently testing out some solid state relays for their suitability in high impedance circuits, and they've actually performed pretty well. However, there is a little problem where I'm left scratching my head... I A/B tested two identical circuits, where one circuit had solid state relays in parallel across each resistor, with all of the relays being turned off. This has the effect of putting the transfer capacitance of the SSR in parallel with the resistor, which is only something like 20pf in my case.

I was initially disappointed with the high frequency response of the circuit, in that it starting rolling off at only 2kHz in comparison to the reference circuit. I suspected the relays in parallel with the plate resistor and grid leak (the grid leak after the coupling cap). So I removed those and nothing changed. Confused, I also removed the relay across the cathode resistor. Nothing changed. Finally, when I removed the relay across the grid stopper, the scope trace at the output suddenly shot up to be on par with the reference circuit. Tracing my way back to the grid of each triode, I found the voltage to be much lower on the side where the relay was attached. Again, when the relay was removed, the trace went up to meet the reference circuit. I stuck the other relays back in and they didn't seem to affect the frequency response at all until I got up to about 10kHz! I also swapped triode sections with each respective circuit but it was still doing the same thing.

The part that is leaving me puzzled, is why the frequency response drops when you place a capacitive device across the grid stopper. IIRC bogner used capacitors across the grid stopper to increase the HF response, not decrease it! I'm thinking it may be due to some loading effect and it's interaction with my signal generator, AKA, my headphone amplifier and PC, but I have to concoct some sort of buffer to test my theory out... Any suggestions as to why it's doing this would be enlightening



PS, in the schematic just pretend there are relays in parallel with the resistors in either circuit

[Alex/Tubewonder]

Which SSR exactly? Link to datasheet ?

[exclamationmark]

Which SSR exactly? Link to datasheet ?

http://www.vishay.com/docs/83823/lh1525at.pdf

[Steve Conner]

Puzzling to say the least . Did it do this even with no control wiring connected to the SSR?

As you point out, there is no way a shunt capacitance across the grid stopper can decrease the HF response. In fact there is no shunt impedance of any kind that could do that. I thought of capacitance between the MOSFET and the LED, but that should be only maybe 5pF. The only suggestion I can offer, is there a mistake or dry joint in your circuit somewhere?

[Steve Conner]

Puzzling to say the least . Did it do this even with no control wiring connected to the SSR?

As you point out, there is no way a shunt capacitance across the grid stopper can decrease the HF response. In fact there is no shunt impedance of any kind that could do that. I thought of capacitance between the MOSFET and the LED, but that should be only maybe 5pF. The only suggestion I can offer, is there a mistake or dry joint in your circuit somewhere?

[Tom Phillips]

I wonder if the SS relay is adding capacitance to ground. Does the problem occur if you swap out the relay with another of the same type?

[exclamationmark]

Embarrassingly enough, this circuit was actually built on a breadboard and only runs at 100v. I guess this allows the situation of having weird resonant circuits with leads flying everywhere, though. I tried connecting both the LED leads to ground to see if the floating inputs were doing anything strange, but it didn't have any affect. I substituted the relay with like 5 others, and it was doing the same thing. I think I'm going to have to rebuild the circuit for a sanity check

I don't know if this has anything to do with it, but the clipping gets slightly sharper looking when I stick the relay in parallel with the grid stopper, whereas without the relay it's more flat and rounded (though that might be an artefact of my ill compensated probes...). I should also mention once the frequency gets past a certain point (I think maybe 10kHz?) the amplitude response doesn't decrease anymore, in line with simulations. I don't even think this circuit has a -3dB point, at least one not measurable with my "signal generator". At about 5kHz the amplitude difference between the reference circuit and the relay circuit is greatest. I remember measuring the pk-pk voltage at the grids to be 1.25V and 1.13V respectively using the same probe, which is what is confusing me the most!

[J M Fahey]

Justb thought the same as Tom Phillips and add: the "transfer capacitance" in your datasheet is even smaller , .75 pF, and anyway "does nothing" across the grid stopper.
The 20pF or so quoted , although not specifically called so, must be capacitance to ground.
May be mistaken, of course, but so far found no other explanation.
I found your 470K grid stoppers quite high ... any particular reason for that?
PLEASE re post your drawing with the actual relays (SSR or otherwise) actually shown.
What we imagine *may* not be what you actually did.
I also suspect some other stray capacitance inadvertlenty added, but let's wait for the full schematic.
Include also the SSR driving circuit.

[exclamationmark]

Justb thought the same as Tom Phillips and add: the "transfer capacitance" in your datasheet is even smaller , .75 pF, and anyway "does nothing" across the grid stopper.
The 20pF or so quoted , although not specifically called so, must be capacitance to ground.
May be mistaken, of course, but so far found no other explanation.
I found your 470K grid stoppers quite high ... any particular reason for that?
PLEASE re post your drawing with the actual relays (SSR or otherwise) actually shown.
What we imagine *may* not be what you actually did.
I also suspect some other stray capacitance inadvertlenty added, but let's wait for the full schematic.
Include also the SSR driving circuit.

I think the .75pf stated is capacitance across terminals from just the physical separation between leads. I don't have an LCR meter to measure it though. As for the 470k grid stoppers... since the circuit was being driven from a low impedance source, I wanted to drive the tube into clipping and have pretty sharp clipping across top and bottom to discern any differences between the SSR and reference circuit (since the low impedance source would drive the tube into grid current operation without a grid stopper and I would see clipping on only one side of the wave). Serendipitously, it also happened to uncover this other little problem.

Right now it's actually my bed time, but I'll post the full circuit first thing tomorrow!

[exclamationmark]

I cooked up a little test rig which consisted of a 1Meg resistor to ground and the SSR in series to find the capacitance across the SSR in the off state (by measuring the bleed through signal) and found it to be in the range of 60-70pF on average. You wouldn't want to use these for channel switching without having a ground shunt! WAY above the quoted 25pF or so at 0v from the datasheet. I then proceeded to stick a 65pF capacitor in place of the SSR across my 470k grid stopper and hey presto, it's doing the exact same thing! I'm actually a little relieved that it isn't some evil solid state mojo stealing effect from the relay, but it's still perplexing to say the least.

[Steve Conner]

Oh well, thanks for taking one for the team! 65pF is a huge amount in tube circuit terms.

You might like to experiment with a negative DC bias to the unused "DC" terminal of the SSR. The capacitance decreases quite sharply with DC bias, however the tricky thing would be applying it without introducing clicks and pops to the signal path.

[J M Fahey]

Thanks a lot for actually experimenting and posting.
That's one more example that although simulation is a great tool, it does *not* replace actual breadboarding ... at all.
Simulation would have used the 22pF specified (or even worse, the 0.75pF "across terminals").

[exclamationmark]

I still have no idea how the shunt capacitance was decreasing the frequency response, but I did find one other tidbit of information concerning the rather large capacitance. Apparently MOS capacitors are frequency dependant, and most manufacturers state their capacitance values at 1MHz, probably in an effort to look impressive (the capacitance is lowest at high frequencies). The problem is, at audio frequencies the capacitance increases due to some mumbo jumbo about minority carrier mobility. So 30pF on a datasheet might turn out to be 100pF at audio frequencies!

[Steve Conner]

And no doubt the frequency-dependent capacitance will create some sort of cheesy distortion too. Bleah.

[J M Fahey]

Power MOS are even worse !!!
They may show up what's strictly not a gate capacitance but a voltage and frequency modulated admittance (ugh!!) which forces me to consider, when designing, a monster 2200 to 4700 pF capacitance which must be driven or else.
I repeat, not strictly a "capacitance" and a meter might show a much lower value (such as 500 pF) but if I consider this last value only, the Mos stage loses drive at higher frequencies and is either "dull" (if audio) or may even mysteriously burn (if switching) because it does not fully switch as expected. (Or it does but not as fast as needed)
Oh well.