Ultralinear Miller capacitance

[rdf]

Hi all, searched and couldn't find an answer on this one. Given the same OPT and tube, in this case a Hammond 1628se and an EL84, how does input grid capacitance vary between triode, ultralinear and pentode connections? I know triode is higher than pentode but can't find an answer to where ultralinear falls.

The question comes from playing around with a direct coupled EF86-EL84 SE. Converting the output stage from pentode to ultralinear appears to have ameliorated the funkiness of the Hammond around 13 kHz but also results in an impressive rolloff at 15 kHz. I'm trying to figure out if the latter is a result of reduced gain available for global feedback or the EF86 choking on the new load. Thx in advance.

 

[Sch3mat1c]

As with all the other parameters of UL, it also sits on the fence between triode and pentode. To find it, first find the gain of the stage, then divide by the turns ratio of the UL tap (43% tapping would be 43% less gain than seen at the plate), and finally add 1 and multiply this by Cgs.

Tim

 

[rdf]

Thanks kindly!

 

[PRR]

Tim is correct. But for quick thinking, ignoring "+1"s and "43%"s: Miller of a UL will be much higher than a Pentode, about half that of a triode.

> Converting the output stage from pentode to ultralinear appears to have ameliorated the funkiness of the Hammond around 13 kHz but also results in an impressive rolloff at 15 kHz.

That's not (mostly) Miller.

Transformer is very sensitive to source impedance. The source impedance of a pentode is about 10X the useful load impedance, "very high". The source impedance of a triode is lower than the useful load impedance, "low". UL splits the diff, "medium".

The transformer has capactitance and leakage inductance. Looking at your curves elsewhere, more than one set of L and C inside the audio band (I wonder if your transformer is faulty? interleaved wrong, or even mechanical resonances?). Obviously an L and a C is a Tuned Circuit, modified by any R laying around (load and plate resistances). I admit I just spent time trying to "prove" it in SPICE, and did not get anything worth showing.