EQP1A clone: +11dB @18kHz (even in bypass). Tweak Zobel? (help!)

[soapfoot]

Hello!

Just finished up an EQP-1A. It came out nicely! Decided to run some sweeps today and noticed an anomalous behavior.

Circuit is identical to schematic. Triad HS-66 input, Cinemag CM-HS29 interstage, Cinemag CM-S217D output.

All performs well except there's a curious rising frequency response peaking at about 18.2kHz before rapidly falling off.

It seems as though perhaps one of the Zobel networks needs to be tweaked (to compensate for slightly different interstage and/or output transformers?)

All values are currently as per stock schematic (270pF and 43k on interstage secondary; 620R and 3000pF on output secondary).

I'm more of a recipe-follower than a design engineer, so I'm not sure which of these networks I should begin experimenting with, nor which direction my experimentation should take (increase/decrease cap? increase/decrease resistor?), or even whether I'm on the right track entirely.

The behavior is present whether the EQ is engaged or bypassed (below sweep was taken with the unit in bypass).

Appreciate any insight/hunches/methodologies!

 

[ruffrecords]

Have you tried it with a 600 ohm load at the output/

Cheers

Ian

 

[soapfoot]

Will try that now

 

[soapfoot]

Have you tried it with a 600 ohm load at the output/

Cheers

Ian

Started by hanging a 619R resistor across output XLR. Bandwidth started to fall away on both ends.

Experimented with a few other values, 1k5 seemed to work the best (pictured below). Now I'm -0.6dB at 20 Hz and -0.4dB at both 8k and 17.5k.

A little spike to +0.5dB at 18.2k before falling away rapidly.

Does this seem like a solution, or should I keep looking to try and improve?

Adding capacitance to the output Zobel seemed to smooth that peak out a touch, but I'm not sure whether I should keep picking the scab.

(Scale on below images is quite a bit finer resolution than the above images, since we're so much closer).

 

[Rogy]

Hi,

If your measurement setup is soundcard-based, try another soundcard or put an active buffer between the EQP-1A and the soundcard.

Good luck!

Rogy

 

[soapfoot]

Hi,

If your measurement setup is soundcard-based, try another soundcard or put an active buffer between the EQP-1A and the soundcard.

Good luck!

Rogy

that’s a good thought

because it was easy to do so, I patched in my LA-2A instead (no gain reduction), and it behaved completely normally.

then I patched in the EQP-1A before the LA-2A, and it still exhibited the rising response, and the 1k5 across the output still fixed it.

so it does seem like it really does want to be terminated somehow—it’s just curious to me why it performs so much better with 1k5 than with 620R.

for what it’s worth, I looked up input Z of the interface I’m using to test, and it’s stated as 10k for the line input

 

[soapfoot]

Thus far, adding a 1k8 resistor across output seems to be the best. I'm +0.75dB at 14kHz and -0.5dB at 20kHz.

 

[Moby]

I presume that there is the oscillation. S217D is really hard to replicate to work correctly. I know it from my personal experience, but I made it to work as expected at the end.
When I had a problems with the early prototypes I found out that lowering the feedback centar tap resistor from 1k to 560 or 680ohms helped me with oscillations. I believe that this can help in your case.

 

[soapfoot]

I presume that there is the oscillation. S217D is really hard to replicate to work correctly. I know it from my personal experience, but I made it to work as expected at the end.
When I had a problems with the early prototypes I found out that lowering the feedback centar tap resistor from 1k to 560 or 680ohms helped me with oscillations. I believe that this can help in your case.

Good tip; seems easy enough to try

 

[Moby]

Good tip; seems easy enough to try

Let us know if it helps. If yes then the problem is inside the output transformer.

 

[soapfoot]

Let us know if it helps. If yes then the problem is inside the output transformer.

Will do!

Also, I twisted the two leads of the start/finish of the feedback winding together when routing them to the pair of 360R resistors. You don't think this would contribute to instability, do you? It doesn't seem like it would, but I thought I'd mention it.

Reducing the 1k from center tap to ground would effectively reduce the amount of feedback, correct?

Thanks again!

 

[soapfoot]

Let us know if it helps. If yes then the problem is inside the output transformer.

Okay, I removed my 1k8 load resistor from the output XLR and made some tests into a 10k impedance line input. I'm not quite sure how to interpret the results

First, with the stock 1k. You'll notice a sharp peak right around 20k, something like +12dB



Then with 4k7 added in parallel for (circa) 820R. You'll notice our peak hasn't changed much


Then with 3k3 in parallel for nominally 765R, you'll notice our peak begins to look shorter; "only" about 9dB now



Changing to 2k2 in parallel for nominally 680R gets us slightly further improvement. Peak is now more like +8dB

1k5 in parallel for 600R flattens us out by about another dB.



1k in parallel for 500R nets incremental further improvement. Now we're about +6dB at 20 Hz, but still far from flat.



I stopped my experimenting here.

Does this look like the dreaded oscillation?

 

[Moby]

All I can say is that your measurings looks almost the same as I started to play with the S217D transformer replica. Maybe you should contact the manufacturer of it and ask about possible tips or values changes they suggest?

 

[soapfoot]

Thanks, I've e-mailed them. Sometimes getting a return e-mail is a bit tricky so I might call early next week.

 

[soapfoot]

I heard back from David at Cinemag, who is awesome and very helpful. We tried a few things over the phone. No solution yet, but he says his S-217D performs great on his bench, and he hasn't heard any other reports of oscillation from other customers.

Also, we removed input signal and there was no oscillation present, so I presume I don't have an oscillation after all? But something still may not be quite right. It does still look "normal-ish" with 1k8 across the output XLR, so I'm holding that for now.