Tube stage HF roll-off problem

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
I'm doing measurements once again.
Test signal unloaded output is -45dBm.
When connected to the transformer it reads -48dBm
Transformer output is -20dBm unloaded and -23dBm when connected to the tube.
 
Interesting, from memory 3dB loading of a transformer secondary meant for unloaded operation represents matching Z, is this tube grid really 120K-150K? Would imply then that the primary side is actual 150ish ohms. Does the test signal still measure -48dBm with secondary connected to the tube? Sorry, curiosity tangent.....
 
I'm doing measurements once again.
Test signal unloaded output is -45dBm.
When connected to the transformer it reads -48dBm
You probably mean dBu... :)
Transformer output is -20dBm unloaded and -23dBm when connected to the tube.
That is not normal. Are you sure the G2 is properly by-passed?
 
What is the signal generator? Is it a line output on your audio interface/soundcard? Looking at the Sowter specs for that trafo, it has 5Rdc for the primaries. I know we have a HF problem here but at least for LF this would make me question the inductance at say 100Hz, even if it does have a mumetal core. 1:30 is a tall ask for an input trafo that is supposed to accept a 200 ohm source, and I cannot see how a preamp with this Sowter trafo will work well with some mics. Out of curiosity and assuming you have an LCR meter at hand, what is the inductance of pri and sec at 100Hz (at 1V preferably)?

If I were you I would also connect all secondary wires together and measure the capacitance between that bunch of wires and the screen wire that goes to GND. This C is also important for HF roll-off. Your 103mH leakage L works out to just 6500Ω at 10KHz, or 13KΩ at 20KHz. That in series with the secondary windings should not cause the HF roll-off that you are seeing.

Are you using EF804 tubes? And have you grounded the screen pins? Just checking
 
Does the test signal still measure -48dBm with secondary connected to the tube?
No, it doesn't. It measures at -50dBm

You probably mean dBu
Probably. This is how my HP34401 shows.
Out of curiosity and assuming you have an LCR meter at hand, what is the inductance of pri and sec at 100Hz (at 1V preferably)?
Primaries are 370 mH each, Secondary is 1970H. I don't know the test voltage, sorry.
Capacitance between shorted secondaries and electrostatic shield is 320pf. I took a second transformer to measure this, but it's the same batch, so should be pretty close.
Are you using EF804 tubes? And have you grounded the screen pins?
I'm using 806s and I did ground screens.
 
No, it doesn't. It measures at -50dBm

OK, the generator has significant loading effects. May or may not indicate an error.

I would use a 20dB mic pad (could also use a 40-50dB line pad) between the generator and the primary so as to raise the load Z on the generator, remove any loading effects. Raise generator output accordingly.
 
What is the signal generator? Is it a line output on your audio interface/soundcard?
Sorry, missed this one. Yes, it's a sound card.
I would use a 20dB mic pad (could also use a 40-50dB line pad) between the generator and the primary
Actually, I do have the -20dB pad built into the front end (it's 2x866R series resistors + 200R shunt) and -30dB pad (2x3.16K resistors + 200R shunt). HF roll-off is even worse with either of these pads engaged.
 
Sorry, missed this one. Yes, it's a sound card.

OK, what make and model of soundcard is it?
Actually, I do have the -20dB pad built into the front end (it's 2x866R series resistors + 200R shunt) and -30dB pad (2x3.16K resistors + 200R shunt). HF roll-off is even worse with either of these pads engaged.
Hmm ok, well this certainly tells me that adding series resistance (2x866R) to the load degrades the HF response further. And if the intention is to use this preamp with microphones and sometimes with these pads engaged, then a lot of mics will have issues in driving this setup.

OK, the generator has significant loading effects. May or may not indicate an error.

I would use a 20dB mic pad (could also use a 40-50dB line pad) between the generator and the primary so as to raise the load Z on the generator, remove any loading effects. Raise generator output accordingly.
If the issue is that the source impedance of the soundcard is not low enough then I don't think this will get away from the issue, and may actually make it worse, as Ilya seems to have already observed.
 
The sound card is RME FireFace 400.
I’m learning more and more to the idea that it’s the transformer issue because feeding the signal to the grid directly doesn’t cause any frequency response deterioration.
I will try to find another 1:7 or 1:10 transformer and check how it behaves. This is quite difficult to do currently though.
 
Sorry, missed this one. Yes, it's a sound card.

Actually, I do have the -20dB pad built into the front end (it's 2x866R series resistors + 200R shunt) and -30dB pad (2x3.16K resistors + 200R shunt). HF roll-off is even worse with either of these pads engaged.
OK. RME FireFace 400 states 75 ohm output. 6K pad resistance, hard to believe that would be a problem.

Most vintage transformers tend to show some treble rise when a pad is engaged, as compared to the same source without a pad. I've never seen a situation where a proper pad didn't improve response somewhat. I have seen some very high value pads like the Shure 50dB type show treble loss from capacitive coupling in the pad itself.

None of this necessarily incriminates the sound card, just that it's worth eliminating that possibility. Do you have any other amplifier with low Z drive capability that could be placed between the sound card and the preamp?

It does seem more likely a transformer problem.
 
I tried driving the input with the Analog Discovery 2 that has very low impedance output, and it didn’t make any difference.
I think I should have some old Triad mic input transformers somewhere. I’ll try them and see if anything changes.
 
No, it doesn't. It measures at -50dBm
Well, it shows that connecting the tube's grid to the secondary loads it significantly, which is not normal, since the low-frequency input impedance of a tube is supposed to be almost infinite.
I'm using 806s and I did ground screens.
Have you noticed that the EF804 has a different pin-out than the 806 (or 86)?
Putting and 806 in place of an 804 reverses G2 and G3, which results in G2 acting like a degenerated anode, and no screen in between the grid and it, so the Miller capacitance would be quite high.
 
Have you noticed that the EF804 has a different pin-out than the 806 (or 86)?
Yes, I've accounted for the pin-out difference.

I didn't have any mic input triads, but I have UTC O-12 that can be configured as 50:500 step-up. Better, than nothing.
With O12 the frequency response is nice and flat no matter what I do with the front end pads. The level is obviously quite low, so I'm now more confident in my theory that it's the Sowter that is acting funky.
 
Yes, I've accounted for the pin-out difference.

I didn't have any mic input triads, but I have UTC O-12 that can be configured as 50:500 step-up. Better, than nothing.
With O12 the frequency response is nice and flat no matter what I do with the front end pads. The level is obviously quite low, so I'm now more confident in my theory that it's the Sowter that is acting funky.

Ok interesting! Well that C to screen of 320pF works out to 50k ohm at 10khz. It is something, bit shouldn't be the culprit. You could try lifting that screen from GND and then test.

As for pad R values, I still need to prove it out for myself, but for me it makes sense that the shunt resistor must be less than the lowest input impedance presented by the preamp, otherwise that impedance (possibly frequency and gain-setting specific) is going to affect the voltage across the shunt resistor, therefore messing up your Freq response. Some preamps vary with gain setting, such as the RCA op6. The v76 might also, due to a similar variable neg feedback path
 
It sounds to me like the Sowter might actually be designed for 50 ohm microphones. My experience with high ratio input transformers is the opposite of EMRR's. I find when driven with a low impedance source they peak around 25KHz but become flat or have some HF loss when fed with from the specified impedance (usually 150 or 200 ohms). You could try making a pad with a couple of 3K3 resistors and a 47 ohm shunt. This will drop the level by 43dB.

Cheers

Ian
 
Yes it could well be Ian, but Sowter offer this as a replacement for a v76 input trafo, and the v76 has an input impedance of 200 ohms I think. Looking at Ilya's inductance measurements at 100Hz (370mH each pri/1970H sec) then with primaries in series this will theoretically be 1.48H total, which then not considering the load on the secondary, this L presents an impedance of 465 ohms at 50Hz, or 232.5 ohms at 25Hz. To me this is Sowter cutting it as close as they can for a 200 ohm input Z and minimal secondary loading. With such a high ratio maybe they needed to so that the secondary wasn't a million turns.
 
Some preamps vary with gain setting, such as the RCA op6. The v76 might also, due to a similar variable neg feedback path
I don't see how it would be the case here, since the open-loop input impedance is supposed to be infinite. NFB could not change it.
It seems though that the impedance is not infinite, for a reason I fail to grab. Maybe a roasted tube socket?
 
You could try lifting that screen from GND and then test
Will check this, but frankly, I doubt that this will change anything much.
You could try making a pad with a couple of 3K3 resistors and a 47 ohm shunt
I'll try that, Should be fairly simple to do - replace 200R with 50R and leave 3K16 in place. Should be good enough for tests
It seems though that the impedance is not infinite, for a reason I fail to grab. Maybe a roasted tube socket?
This is an all new build, tube sockets are new as well. I'm pretty confident regarding the build and the exact circuit I'm using. I've checked it probably hundred times (I know that there's still room for error, but the chance is rather small). Is there any way to test for this impedance problem? As far as I can tell now, it's the impedance mismatch between the reflected impedance to the secondary and input grid impedance.
 
Yes it could well be Ian, but Sowter offer this as a replacement for a v76 input trafo, and the v76 has an input impedance of 200 ohms I think. Looking at Ilya's inductance measurements at 100Hz (370mH each pri/1970H sec) then with primaries in series this will theoretically be 1.48H total, which then not considering the load on the secondary, this L presents an impedance of 465 ohms at 50Hz, or 232.5 ohms at 25Hz. To me this is Sowter cutting it as close as they can for a 200 ohm input Z and minimal secondary loading. With such a high ratio maybe they needed to so that the secondary wasn't a million turns.
It is obvious that Sowter did cut some corners. The V76 input xfmr is deemed impossible to reproduce, because of the lost art of multi-chamber winding, segmenting and sandwiching.
1.48H for a 200r mic input is about 1/4th of a similar Jensen xfmr.
You calculation, that I don't dispute, shows a -3dB LF point at about 22 Hz, which is not what is usually expected. Typical should be -0.5-1 dB at 20 Hz, which corresponds to a -3dB point of ca. 10Hz, that would require a primary inductance of 4H.
However, it must be noted that inductance tends to increase at very low frequencies, so they may be just right.
Anyway, the issue is not LF response here. The nominal inductance of the xfmr is not bound to alter the HF response.
 
Back
Top