Tube stage HF roll-off problem

[Ilya]

I'm having the following problem. Here's the input mic preamp stage (this is the part of the V76 circuit verbatim):

For some reason I get the HF roll-off that starts around 6-8K, gets to the -5dB point at 10k, and -10dB at 20k. I'm sure this happens around the very first tube since I measured the output at the marked point and the roll-off is there.
I'm running this circuit in a simulator and I cant get any roll-off no matter what parasitics I include in the sim.
Interestingly, when I go into the tube directly (omitting the input transformer), the HF response is flat. This leads me to the thought that something is funky in the impedance department between the input transformer and the tube input impedance. I tried various sources, but could not get rid of this roll-off. This roll-off gets worse if I connect the input windings in parallel.
What could be causing this issue?

 

[emrr]

No change with gain setting?

 

[Ilya]

Gain setting doesn’t affect the roll-off.
I must correct the previous statement. The -10dB at 20k is the worst case and happens with primaries in parallel. When in series, the roll-off is approximately -5dB at 20k, but it’s still perceptible.

 

[RuudNL]

What happens with the signal level on the grid of the first tube?

 

[emrr]

I’d first swap transformer secondary leads, see if better/worse/fixed. If not then i’d put the input transformer into a DI with 1M+ input Z and confirm it’s passing treble correctly there. Eliminate the transformer as the problem definitively.

There’s a very wide window of Z acceptability, anything from 10k to 100k should be fine.

Your test generator source can also be the problem. If the generator and input pri aren’t loading correctly it may roll off treble in a test result.

 

[Ilya]

What happens with the signal level on the grid of the first tube?

Input -20dBV
-13dB at the first tube grid
-13dB output at 25n cap (part 17), minimum gain setting,

I'm measuring with my sound card input, so it probably loads the stage.

I’d first swap transformer secondary leads, see if better/worse/fixed. If not then i’d put the input transformer into a DI with 1M+ input Z and confirm it’s passing treble correctly there. Eliminate the transformer as the problem definitively.

You mean swap hot and cold? Will do that.
The transformer measures fine. I get a flat output from the transformer alone (although I measured with the sound card which probably has 10k input impedance).

I suspect that interwinding capacitance and miller effect has something to do with this issue. At least when I added those capacitances in the sim I started seeing the same looking results.

 

[ruffrecords]

You mean swap hot and cold? Will do that.
The transformer measures fine. I get a flat output from the transformer alone (although I measured with the sound card which probably has 10k input impedance).

I think he means swap the secondary winding. Some high ratio transformers do have a different response gepending on secondary polarity.

I suspect that interwinding capacitance and miller effect has something to do with this issue

The first stage is a pentode so Miller effect should be virtually non-existent.

Cheers

Ian

 

[abbey road d enfer]

What could be causing this issue?

Excessive leakage inductance of the transformer. What are you using?
What ratio did you measure when you tested it on its own?

 

[Ilya]

Changed the secondary polarity and nothing changed.

Excessive leakage inductance of the transformer. What are you using?

Sowter 1004 (VINTAGE RECORDING AMPLIFIER TRANSFORMERS J to Z)
I can measure leakage inductance. Need to look up how to set the measurements correctly

 

[emrr]

You mean swap hot and cold? Will do that.
The transformer measures fine. I get a flat output from the transformer alone (although I measured with the sound card which probably has 10k input impedance).

Swap the secondary leads. Sometimes there are very large treble differences due to capacitance/ground orientation.

Measuring with a 10K transformer secondary load forces better response on the transformer, so long as the source is capable of driving it.

Consider the layout of the secondary leads to the tube, if long, or running through DI paths or relays it may be degraded.

 

[Ilya]

Here're interwinding capacitance measurements.
P1 to S1 - 108pf
P1 to S2 - 120pf
S2 to S3 - 78pf

With S1 and S2 in series or parallel capacitance is 180pf

Leakage inductance with primary shorted - 118uH
With secondary shorted - 103mH (that's millihenry, not a typo)

 

[Ilya]

Swap the secondary leads. Sometimes there are very large treble differences due to capacitance/ground orientation.

Measuring with a 10K transformer secondary load forces better response on the transformer, so long as the source is capable of driving it.

Consider the layout of the secondary leads to the tube, if long, or running through DI paths or relays it may be degraded.

The leads including PCB traces are not that long. I suppose that when John Hinson mentioned that he couldn't find any good 1:30 transformer he meant exactly what I'm encountering now.
Maybe I need to try a 1:10 Cinemag, but I don't have any good trafos handy.

 

[RuudNL]

But is there a frequency dependence when measuring the level on the grid of the first tube?
(In that case, the transformer would be the culprit.)
What happens if you inject a signal directly on the grid?

 

[emrr]

Wow 1:30, just processing that info. That's tough. The Altec 458A/459A use 1:31.6, almost nothing else has such a ratio, and that's push-pull, which in my experience always delivers better transformer response than single ended, comparing the same transformer in both circuit types. Completely unnecessary for modern gain needs, 30 dB in the transformer alone!

Input -20dBV
-13dB at the first tube grid
-13dB output at 25n cap (part 17), minimum gain setting,

In that light, it appears gain is not as it should be at all, and likely in full overload, should be +10dBV roughly at the grid, way too much. If that's on the right track at all, try -50 or -60 input.

 

[Ilya]

But is there a frequency dependence when measuring the level on the grid of the first tube?
What happens if you inject a signal directly on the grid?

I'm not sure what you mean in the first sentence.
If I inject the signal into the grid directly, the frequency response is flat.

In that light, it appears gain is not as it should be at all, and likely in full overload, should be +10dBV roughly at the grid, way too much. If that's on the right track at all, try -50 or -60 input.

With minimum gain the output of the whole chain (including the following stages) is approximately 30dB which correlates to what I'm getting at the output.
I'll try to get better probes and measure voltages with the DMM which shouldn't load the circuit as much as a sound card input does.

 

[emrr]

You don't need to measure a different way, you just need to quickly confirm that if you lower the input level by 30-40 dB it still behaves the same way, 30-40 dB lower. Eliminate overload possibility by shifting the entire level structure.

If the input is indeed -20dBV, it's already way too hot for a 1:30 input transformer and any following tube circuit.

 

[abbey road d enfer]

Leakage inductance with primary shorted - 118uH
With secondary shorted - 103mH (that's millihenry, not a typo)

Then there is something wrong. The primary leakage L should be about 1/1000th of the secondary one.
With primary shorted, you measure secondary leakage, right?
It looks like you have crossed your figures. 118uH pri and 103mH for sec seems about right.

You say that you see this low-pass at the grid of the 2nd tube. This is not correct, because you're in the loop. The response there is not necessarily flat. An absurd example would be to measure the signal at the Vas of an opamp.
The only valid measurements are at the grid of the 1st tube and at the output of the second. Due to the parafeed inductor, teh resposne of teh 2nd stage is somekind of a high shelf, so with teh NFP loop closed, teh 1st stage will compensate.

 

[Ilya]

you just need to quickly confirm that if you lower the input level by 30-40 dB it still behaves the same way, 30-40 dB lower.

Yes, it behaves the same. I lowered the test signal to -48dBm. Now at the output of the transformer it's -23dBm. -38dBm at the first tube output and -19dBm at the second stage output.

Then there is something wrong. The primary leakage L should be about 1/1000th of the secondary one.

Doh, of course that should be vice versa. I mixed up secondary and primary.

The only valid measurements are at the grid of the 1st tube and at the output of the second.

Ok. I'm not seeing the roll-off at the grid of the first tube. And the rol-off is there at the output of the second tube

 

[Ilya]

Ok. I've simulated the parallel connection of the primaries and included the 50pf interwinding capacitance. Voila, the LPF is there, and is very much like what I am seeing in the real world.
I guess there's nothing that can be done about this...

 

[emrr]

OK, 25dB transformer gain seems more reasonable. On paper it should be closer to 30, so that would seem to point to a problem, it really shouldn't have such losses if the ratio is accurate. The gain is more like a 150:50K type.