Transformer Impedance problem

[DaveP]

I have been working on the design of a signal amp for the EF85 compressor from scratch project.

I can have one tube/side in push-pull or say four/side and adjust the plate resistor to suit.  The configuration not only alters the overall tube parameters (mu stays the same, gm x 4, rp/4) but it alters the output impedance from the tubes that the output transformer sees.  This is especially relevent under gain reduction as the rp rises. see table.


I know that the primary inductance affects the low frequency response, but I'm not totally sure what causes the high frequency roll off, I'm guessing its the primary leakage inductance/capacitance but I would like some expert input here.
Edcor Data table below with my figures.


I have found that the lower the output Z, the better the frequency response range.  I have been following PRR's logic from here,
http://groupdiy.com/index.php?topic=30349.msg368001#msg368001. but a 10kTX as he suggests has too poor a response with one tube/side (triode wired), 4 tubes is ok.
Expert input please

Best
DaveP

 

[abbey road d enfer]

I have been working on the design of a signal amp for the EF85 compressor from scratch project.

I can have one tube/side in push-pull or say four/side and adjust the plate resistor to suit.  The configuration not only alters the overall tube parameters (mu stays the same, gm x 4, rp/4) but it alters the output impedance from the tubes that the output transformer sees.  This is especially relevent under gain reduction as the rp rises. see table.


I know that the primary inductance affects the low frequency response, but I'm not totally sure what causes the high frequency roll off, I'm guessing its the primary leakage inductance/capacitance but I would like some expert input here.
Edcor Data table below with my figures.


I have found that the lower the output Z, the better the frequency response range.  I have been following PRR's logic from here,
http://groupdiy.com/index.php?topic=30349.msg368001#msg368001. but a 10kTX as he suggests has too poor a response with one tube/side (triode wired), 4 tubes is ok.
Expert input please

Best
DaveP

That's a quite complex issue.
Consider you have a voltage divider, where the first (series) branch is constituted of the eakage inductance in series with source impedance , and the second  (shunt) branch is constituted of the load impedance in parallels with the stray capacitance. You could use some math to write the equations or you could use a simulator like LTspice, but anyway it's easy to see that lleak and Cstray being given parameters, increasing either Rsource or Rload will definitely reduce the HF response. OTOH reducing Rsource increases the level but reducing Rload reduces the level.
Put another way, frequency response increases as Rsource in parallels with Rload.(N1/N2)² while level decreases with the ratio of Rload to R source.
I hope I haven't made things muddier than before...
The issue there is that manufacturers generally don't publish the stray capacitance figure. You have to figure it out by measuring the primary impedance at HF, above audio frequencies, with a series resistor, and find the 1/C.omega asymptote.

 

[abbey road d enfer]

I have sim'ed the basic schemo (with a hypothetical transformer) for 4 conditions:
a)nominal-Z drive, nominal-Z load; that's the idle condition in a Vari-Mu comp
b)high-Z drive, nominal load, to simulate the condition under 20dB GR
c)high-Z drive, high-Z load; that's what happens under 20dB GR and bridging the output (loaded with 10x nominal Z, e.g. 6k instead of 600 r
d)nominal-Z drive, high-Z load; idle with bridging output
You can see that the FR, which is respectable at idle and under nominal load, becomes questionable under 20dB GR or 10:1 mismatch in output, and is plainly unacceptable under heavy compression AND load mismatch.

 

[DaveP]

Thank you Abbey for taking the time for such a full answer, I needed that confirmation of the situation.

I will see if I can measure the leakage capacitance, there are certainly some resonances going on around 100kHz.

Best
DaveP

 

[gyraf]

(Great thread. I posted some old-fashioned transformer theory in the tech.docs section - http://groupdiy.com/index.php?topic=62877.0  /Jakob E.)

 

[DaveP]

Thanks for those articles Gyraf, solid gold! ;D

This is why this forum works so well.

DaveP