unsymetric phase splitter

i am still struggeling with the output tube stage of my passive eq.
the sound is close to an acceptable and finished state.
maybe now i have found a cause that wont be fixed with loading or compensation.

the two sides of the symetric amp doesnt match up in the extremes when looking at phase response. i first noticed this when checking phase across one of the output plates and the OT secondary, and then comparing it with the response of the other plate and flip the probe/groundclip on the secondary. one side goes predictably to 180º while the other side takes a turn around 45º, goes up to 0º and then turn again. like there is a resonance some where in one of the sides and not in the other...

today i could see that my interstage/phasesplitter transformer also shows a very unsymetric behaivior where one end has a 180º at 170khz while the other is around 350khz at 180º.

what say you?

the ll7903 isnt stated as having great symmetric properties like some other lundahl transformers. is it time to look for a better contender?

> 350khz at 180º.  what say you?

I say I can't hear 170KHz/350KHz, and generally can't care.

hmm.. i wonder if im looking at this in the wrong way since it is a symmetrical circuit and not a single ended one.

i just noticed that i can switch the oscilloscope to ch1+ch2. and invert one channel so i can switch between sum/diff. then i get a more pleasantly looking square.

here is a pic on both secondary ends of the interstage transformer. no load/zobel. 5khz square.

and here is with ch1+ch2 one channel inverted.

here is a pic with a zobel of two 1nf caps and a resistor between them and centertap, a T filter kind of.

notice that they are not super symetrical.

ch1+ch2

looking at the waveforms seperatly, i get confused.
looking at them summed, i calm down a bit.

a word on working with symetrical circuits and how to monitor them would be apriciated.

thanks!!

5v333 said:

hmm.. i wonder if im looking at this in the wrong way since it is a symmetrical circuit and not a single ended one.

Click to expand...

But the transformer is not symmetrical at 170+kHz; leakage inductance and stray capacitance cannot be made equal. You should limit your considerations to just an octave above 20kHz.

hi abbey road d enfer!

i dont understand if you mean - not to worry about whats happening in the extremes or
that the transformer is only good for circuits with 40khz bandwidth...

i am expecting a bandwidth at the output of around 60-85khz.

i opend the loop in my tube amp yesterday and checked margins again. both sides have the same gain/phase margin but at different frequncies (about one octave apart).
about 80º margin at unity and about 20dbs margin at 180º (about 18mhz).

all this seems like fine to me but theres still something sticky and edgy about the sound...

(forgot to check margins in the bass btw)

5v333 said:

hi abbey road d enfer!

i dont understand if you mean - not to worry about whats happening in the extremes or
that the transformer is only good for circuits with 40khz bandwidth...

Click to expand...

I mean that the xfmr manufacturer is concerned with audio frequencies; what happens outside is out of his responsibility.

i am expecting a bandwidth at the output of around 60-85khz.

Click to expand...

Then you may be expecting too much. What for? Are you dealing with ultrasonics?

i opend the loop in my tube amp yesterday and checked margins again. both sides have the same gain/phase margin but at different frequncies (about one octave apart).
about 80º margin at unity and about 20dbs margin at 180º (about 18mhz).

Click to expand...

How did you measure that? If you're concerned with stability margin, the input xfmr is out of the loop, then it's only the output xfmr, and since there is cross-coupling via the output xfmr, I doubt you can analyse separately the two halves.
Then it is quite possible the input xfmr secondaries have quite different resonant frequencies, due to construction.

the whole disscussion on high bandwidth -  i think i leave that to every one to decide for them self. i know what sounds i like. but if theres a matter of instability, then thats another matter.

lundahl arent the fastest transformers around all the time but it looks like this is what i get from sweeps and scope. maybe im doing this the wrong way though...

measuring margins: i keep the circuit as it is, except for disconnecting R11 R12 (feedback R´s) from the cathodes. tie thoose resistors to additional 1.5k resistors that goes to ground so the circuit is loaded like if it was closed loop.
one probe at the first grid, the other probe between the feedback resistor and the additional 1.5k resistor of the same half of the amp.

my new fluke scope has dual cursor operation that can be used to measure phase shift. sweep is done with my tektronix fg504 40mhz.

please tell me if this is the wrong way of doing it!

Then it is quite possible the input xfmr secondaries have quite different resonant frequencies, due to construction

Click to expand...

there are other lundahl transfomrers that are stated to have great symmetry for phase splitting while the ll7903 dont (i missed to read on this when i ordered the ll7903). im thinking about ordering one to try...

5v333 said:

measuring margins: i keep the circuit as it is, except for disconnecting R11 R12 (feedback R´s) from the cathodes. tie thoose resistors to additional 1.5k resistors that goes to ground so the circuit is loaded like if it was closed loop.
one probe at the first grid, the other probe between the feedback resistor and the additional 1.5k resistor of the same half of the amp.

Click to expand...

By measuring from grid to NFB return, you have somewhat eliminated the input xfmr. However you haven't eliminated cross-coupling via the OT. Proper analysis would involve disconnecting the plate on the side you're not measuring. It wouldn't be 100% accurate since the other half of the OT would not be properly loaded, but it would allow comparison between both sides.

hmmm..!

disconnecting one plate is what trobins told me in another thread i think. i prob wasnt paying attention. sorry trobins..!

i think trobins said something about dc loading the OT-half-not-in-use with a resistor or something...

or could we leave both plates in circuit but disconnect the ac-signal on one side? for ex disconnect one interstage cap..?

thanks abbey!!!

5v333 said:

hmmm..!

disconnecting one plate is what trobins told me in another thread i think. i prob wasnt paying attention. sorry trobins..!

i think trobins said something about dc loading the OT-half-not-in-use with a resistor or something...

or could we leave both plates in circuit but disconnect the ac-signal on one side? for ex disconnect one interstage cap..?

Click to expand...

That is also possible. However, it will not be 100% accurate, but enough to point out any quirk. That's typical of the difficuly in analysing a push-pull circuit, where both sides react with each other.

ok. have been doing some testing today. connecting the amp like in the picture. both feedback resistors goes to ground via 1.5k resistors. one interstage cap is unhooked in one end.

made some sweeps and checking phase/gain margins, then did the same with the other half of the amp with the other interstage cap unhooked.

gain/phase margin are similar to my recent tests. however i saw some ups and downs in the phase response between unitygain and 180º in one half of the amp. found out that some capacitance on the first cathode to ground or to the other cathode could smoothen this out.

but, when checking the other half this compensation just did everything worse. its like the two halfs dont like each other...

im gonna do some more experiment.

any thoughts?

5v333 said:

ok. have been doing some testing today. connecting the amp like in the picture. both feedback resistors goes to ground via 1.5k resistors. one interstage cap is unhooked in one end.

made some sweeps and checking phase/gain margins, then did the same with the other half of the amp with the other interstage cap unhooked.

gain/phase margin are similar to my recent tests. however i saw some ups and downs in the phase response between unitygain and 180º in one half of the amp. found out that some capacitance on the first cathode to ground or to the other cathode could smoothen this out.

but, when checking the other half this compensation just did everything worse. its like the two halfs dont like each other...

im gonna do some more experiment.

any thoughts?

Click to expand...

When you make these tests, how do you feed the signal?

through the interstage transformer via 470R resstor and the ground connected like in the picture.

5v333 said:

through the interstage transformer via 470R resstor and the ground connected like in the picture.

Click to expand...

So you're including the phase-shift of the input xfmr. You should disconnect teh input xfmr and inject the signal directly at the grid.

ok ill try that.

i guess you mean that this could show if the problem is in the transformer or further in the amp..?

5v333 said:

ok ill try that.

i guess you mean that this could show if the problem is in the transformer or further in the amp..?

Click to expand...

Correct; when something's fishy, try to break it down in a sum of smaller suspects.