I'm testing some line level center tapped transformers today for how evenly they phase split a signal. I've never done this before so maybe someone can explain what I'm seeing. I'm using this transformer:
EDCOR - XSM Series
10k to 10k
I hooked the function generator to the primary and set the level to 1 volt. I hooked two channels of the scope to the split secondary so the scope shows the split signal very cleanly in perfect antiphase. Leaving the level the same I try out different frequencies. 1K, 5K, 10K, 15K, 20K...
Here is my question. As I stay below 15K the amplitude of the two phases is pretty much equal within 1%. But as I move above 15K to 20K one of the phases grows in amplitude over the other phase by a greater percentage.
Is this caused by the fact that the transformer is reaching its rated frequency limit? Or is it caused by the fact that a higher frequency is exposing maybe that 1 turn difference between the secondary halves? Or is it caused by core behavior at the higher frequency?
In any case the antiphase remains perfect, its just that at higher frequencies one side of the secondary grows noticeably taller on the scope.
Using this as a PP phase splitter would cause an imbalance at higher frequencies? There is not much musical content above 15K. Am I being too picky? I was expecting to see both phases stay identical in amplitude as frequency increases.
The Edcor transformer above is actually much better than some others I tested today. But they all exhibit this rising frequency brings on amplitude imbalance behavior. In fact one cheap radio replacement transformer I have started to show this imbalance as I approached 8K to 10K! This one (its a stepper-upper so maybe that makes it even more difficult to keep balance).
https://www.tubesandmore.com/products/transformer-audio-interstage-10-ma
Do you suspect a step-down splitter would probably not exhibit this behavior as much? I have not gotten to testing a few of those yet. I suspect they will maintain balance better just on a hunch, if so I'd probably want to use the
at as a PP phase splitter instead, then get back the gain in the driver.
OK I just tested the Edcor 10K to 600R. It exhibits the same behavior but not as bad, I start to see it begin at 13K but the percentage difference between the phases in amplitude is not nearly as far off as a step up or a 1:1 transformer.
EDCOR - XSM Series
10k to 10k
I hooked the function generator to the primary and set the level to 1 volt. I hooked two channels of the scope to the split secondary so the scope shows the split signal very cleanly in perfect antiphase. Leaving the level the same I try out different frequencies. 1K, 5K, 10K, 15K, 20K...
Here is my question. As I stay below 15K the amplitude of the two phases is pretty much equal within 1%. But as I move above 15K to 20K one of the phases grows in amplitude over the other phase by a greater percentage.
Is this caused by the fact that the transformer is reaching its rated frequency limit? Or is it caused by the fact that a higher frequency is exposing maybe that 1 turn difference between the secondary halves? Or is it caused by core behavior at the higher frequency?
In any case the antiphase remains perfect, its just that at higher frequencies one side of the secondary grows noticeably taller on the scope.
Using this as a PP phase splitter would cause an imbalance at higher frequencies? There is not much musical content above 15K. Am I being too picky? I was expecting to see both phases stay identical in amplitude as frequency increases.
The Edcor transformer above is actually much better than some others I tested today. But they all exhibit this rising frequency brings on amplitude imbalance behavior. In fact one cheap radio replacement transformer I have started to show this imbalance as I approached 8K to 10K! This one (its a stepper-upper so maybe that makes it even more difficult to keep balance).
https://www.tubesandmore.com/products/transformer-audio-interstage-10-ma
Do you suspect a step-down splitter would probably not exhibit this behavior as much? I have not gotten to testing a few of those yet. I suspect they will maintain balance better just on a hunch, if so I'd probably want to use the
at as a PP phase splitter instead, then get back the gain in the driver.
OK I just tested the Edcor 10K to 600R. It exhibits the same behavior but not as bad, I start to see it begin at 13K but the percentage difference between the phases in amplitude is not nearly as far off as a step up or a 1:1 transformer.
Last edited:
I had struggled with this for a few years, and there are some posts maybe 10 years old which document some of my findings. Whether it matters/is audible or not is a debate that may not be fruitful, but I personally found what I measured is correlated to what I heard. So I pursued resolving it.
The issue of phase/amplitude imbalance between the two winding ends was most pronounced on transformer winding style, and I found to begin generally over 10k. Layered windings like Lundahl were all over the place, and could not be corrected. Unknown winding styles like Onetics were also highly imbalanced. Stray capacitance and its resonance with the transformer leakage and source impedance is ultimately the root cause.
Addition of load resistors or RC networks never corrected the issue that I was able to determine.
My two solutions:
Input transformers are less difficult to find a good splitter. The Jensen / Cinemag options are very nice. Nothing beat the Tribute nanocrystalline core- I could sweep that thing out to 50 kHz and it was rock solid balanced.
Interstage transformers, very simple solution. Go bifilar windings. Super balanced beyond 20kHz for the Monolith.
The issue of phase/amplitude imbalance between the two winding ends was most pronounced on transformer winding style, and I found to begin generally over 10k. Layered windings like Lundahl were all over the place, and could not be corrected. Unknown winding styles like Onetics were also highly imbalanced. Stray capacitance and its resonance with the transformer leakage and source impedance is ultimately the root cause.
Addition of load resistors or RC networks never corrected the issue that I was able to determine.
My two solutions:
Input transformers are less difficult to find a good splitter. The Jensen / Cinemag options are very nice. Nothing beat the Tribute nanocrystalline core- I could sweep that thing out to 50 kHz and it was rock solid balanced.
Interstage transformers, very simple solution. Go bifilar windings. Super balanced beyond 20kHz for the Monolith.
Yeah, I even switched up the scope probes and channels, the imbalance stayed where it was so I know the probes are ok. I dont have any of the exotics you mentioned, I do have a Jensen input transformer JT-11P4 but it doesnt have center taps.
But based on your comments it should be possible to keep balance using the widely easily obtainable Jensen line. This does seem to come down to quality then.
10K is what I'm seeing too using run of the mill transformers. I'd really like to pick up a pair of Jensens, if those stay solid I wont have to bother sourcing a hard to get transformer.
How about this... Rather than using a center tapped secondary transformer to do the phase splitting and possibly have to deal with this amplitude to frequency issue. Has anyone tried a phase splitter like this? (second schematic in the pdf). By centering the ground reference between two matched resistors? I would think this method would bypass the possibility of having the amplitude issue I'm seeing because the coil/core/capacitance/quality is all taken out of the equation completely? The antiphase would occur on either side of the precision resistors right? And amplitude shouldn't change with rising frequency right?
https://www.jensen-transformers.com/wp-content/uploads/2014/08/as060.pdf
https://www.jensen-transformers.com/wp-content/uploads/2014/08/as060.pdf
A solution that worked!
OK, I used a non center tapped transformer I had, Jensen JT-114P
Hooked up to produce an antiphase signal across matched resistors. This method does not exhibit any of the amplitude problems seen while trying to use a center tapped transformer to produce a phase split right out of its coil. I ran this one all the way out to 100K with no shifts in amplitude between the two phases. And of course they are perfectly antiphase provided you match the resistors well. I can literally invert one of the phases on the scope, lay it over the other trace and the two hide each other perfectly in terms of both amplitude and phase match across the whole spectrum.
I've been wanting to try to experiment with PP using a transformer phase splitter instead of cathodyne or LTP, so this experiment is good news.
In the middle of all this my Rigol function generator bit the dust, just stopped producing output. So I had to use my old 1970's vintage Leader analog audio generator (the kind that uses a variable capacitor for the signal frequency). I forgot how much I really like an analog audio generator, the traces are actually better than the digital function generator sine wave!
Here is the transformer I actually used with the schematic above that I had on hand.
https://www.jensen-transformers.com/wp-content/uploads/2014/08/jt-11p4-11.pdf
Also I didnt even bother to load the secondary, amplitude and phase still stayed rock solid to very high frequencies.
OK, I used a non center tapped transformer I had, Jensen JT-114P
Hooked up to produce an antiphase signal across matched resistors. This method does not exhibit any of the amplitude problems seen while trying to use a center tapped transformer to produce a phase split right out of its coil. I ran this one all the way out to 100K with no shifts in amplitude between the two phases. And of course they are perfectly antiphase provided you match the resistors well. I can literally invert one of the phases on the scope, lay it over the other trace and the two hide each other perfectly in terms of both amplitude and phase match across the whole spectrum.
I've been wanting to try to experiment with PP using a transformer phase splitter instead of cathodyne or LTP, so this experiment is good news.
In the middle of all this my Rigol function generator bit the dust, just stopped producing output. So I had to use my old 1970's vintage Leader analog audio generator (the kind that uses a variable capacitor for the signal frequency). I forgot how much I really like an analog audio generator, the traces are actually better than the digital function generator sine wave!
Here is the transformer I actually used with the schematic above that I had on hand.
https://www.jensen-transformers.com/wp-content/uploads/2014/08/jt-11p4-11.pdf
Also I didnt even bother to load the secondary, amplitude and phase still stayed rock solid to very high frequencies.
Last edited:
When you want to measure the difference between the windings, solder 2 prcision resistors of the rated impedance (300 or 600) to both ends, join the other ends and measure the voltage between the midpoints xformer and resistors.
To avoid different capacities loading keep all connections as short as possible and use a bnc connector directly soldered to the midpoints. But put the scope to high impedance.
So a subtraction is done and the rest is the imbalance
To avoid different capacities loading keep all connections as short as possible and use a bnc connector directly soldered to the midpoints. But put the scope to high impedance.
So a subtraction is done and the rest is the imbalance
Before doing an oscilloscope Pseudo differential measurement:
You need to put both probes on the square wave calibrator, and adjust the compensation screw.
Do not change the channel the probes are on.
Then set one channel to invert, and do a ch1 + ch2 (add function).
Any residual signal that you see, is the error of the Pseudo differential measurement.
you may have to use a larger square wave signal if you need to measure more differential volts (because you can not change the channel gain after calibration, the 2 channels may not have equal gain at the new channel gain setting, you have to recompensate the probes and check the measurement error.
With out proper compensation, the channel gains may not be equal, and the probes loss versus frequency may not be equal, you will have an uncalibrated measurement.
Look at the error signal to see how accurate your measurement is at that particular setting of the scope, and those particular probes, on those particular channels.
You need to put both probes on the square wave calibrator, and adjust the compensation screw.
Do not change the channel the probes are on.
Then set one channel to invert, and do a ch1 + ch2 (add function).
Any residual signal that you see, is the error of the Pseudo differential measurement.
you may have to use a larger square wave signal if you need to measure more differential volts (because you can not change the channel gain after calibration, the 2 channels may not have equal gain at the new channel gain setting, you have to recompensate the probes and check the measurement error.
With out proper compensation, the channel gains may not be equal, and the probes loss versus frequency may not be equal, you will have an uncalibrated measurement.
Look at the error signal to see how accurate your measurement is at that particular setting of the scope, and those particular probes, on those particular channels.
6a3summer and bansuri. Yes I calibrated the probes on same channel and the probes are all matching ones that came with the scope. I inverted one phase, overlayed them, and liked the results. But I didn't put the scope on add, just chop. Ill do that. I didn't want to cut the 6 inch leads on my Jensens yet but i did twist them and i know that worked because before twisting them the scope would jump just by moving things, after twisting they were stable. I cant remember if the scope was on hi impedance, ill check.
In any case, so far, I'm convinced that virtual Grounds are a better approach to obtaining antiphase using a Transformer than trying to get it directly from a coil with a CT. But i know next to nothing in theory, only what happened when I simply swept across a CT Transformer and saw one phase grow by as much as 20% over the other phase in amplitude at 30khz. its like one half of the coil takes over the other and wins the higher the frequency goes. This simply doesn't happen with the virtual ground no CT and a reasonably quality Transformer like Jensen.
My goal of phase splitting with an input Transformer as opposed to a tube is about a PP Amp I'm contemplating which I will draw up and post separately. It will have the option of inserting or bypassing the driver stage and phase split right at the input and that switchable too by optionally using the already provided antiphase of the DAC. So basically input Transformer with RCA, no input Transformer with XLR as that is already antiphased. And an optional driver gain Stage if your DAC has enough already and you are using the DAC volume control. But thats another subject. I had to first learn here how to get a perfect antiphase from a Transformer input. And I emphasize learn as I really depend on the minds here, baby steps for me.
In any case, so far, I'm convinced that virtual Grounds are a better approach to obtaining antiphase using a Transformer than trying to get it directly from a coil with a CT. But i know next to nothing in theory, only what happened when I simply swept across a CT Transformer and saw one phase grow by as much as 20% over the other phase in amplitude at 30khz. its like one half of the coil takes over the other and wins the higher the frequency goes. This simply doesn't happen with the virtual ground no CT and a reasonably quality Transformer like Jensen.
My goal of phase splitting with an input Transformer as opposed to a tube is about a PP Amp I'm contemplating which I will draw up and post separately. It will have the option of inserting or bypassing the driver stage and phase split right at the input and that switchable too by optionally using the already provided antiphase of the DAC. So basically input Transformer with RCA, no input Transformer with XLR as that is already antiphased. And an optional driver gain Stage if your DAC has enough already and you are using the DAC volume control. But thats another subject. I had to first learn here how to get a perfect antiphase from a Transformer input. And I emphasize learn as I really depend on the minds here, baby steps for me.
That sounds good, but in practice does not always work that way. Thus far you have only tested with hand-matched secondary load resistors and a probe as burden. Depending on the load in-circuit, you will add unintended additional burden, which will upset that 'perfect' balance you have right now. Think triode, grid capacitance, Miller capacitance, feedback nodes, and just general strays. It will not be a guarantee that the two ends carry balanced impedances. If a split winding transformer does what it's supposed to do, it will be more tolerant of side/side burdens and continue driving matched voltage/phase.
As input transformer duty, may not be an issue, but this generalized solution completely falls apart once you enter interstage transformer territory. So don't be surprised if you find slightly altered response once you remove your probes and get the actual circuit running in its final state.
In practice:
Not all transformer phase splitters are equal, there are excellent, good, and bad.
They can suffer from unequally distributed capacitance, to the laminations; and from the two ends of the primary winding to the two ends of the secondary winding (different at one end versus the other end).
They also can have un-equal leakage reactance from the ends of the primary, to the ends of the secondary windings.
"All things being equal", they are not equal.
I have never used an input transformer. Not a phase splitter, nor a single phase transformer.
But I have used a single phase interstage (bifilar); a single ended to push pull interstage; a push pull primary to push pull interstage; and a tapped inductor (autotransformer) phase splitter.
None were perfect, but all were quite good.
If you use a pair of 10Meg Ohm / 10pF probes, and use a signal generator and 1k series resistor to drive a bifilar wound primary (1k = rp of the tube you will use to drive in the actual amplifier), then the probes will have extremely low load on the secondary (much less than any load resistor across the secondary).
Just my opinions.
Not all transformer phase splitters are equal, there are excellent, good, and bad.
They can suffer from unequally distributed capacitance, to the laminations; and from the two ends of the primary winding to the two ends of the secondary winding (different at one end versus the other end).
They also can have un-equal leakage reactance from the ends of the primary, to the ends of the secondary windings.
"All things being equal", they are not equal.
I have never used an input transformer. Not a phase splitter, nor a single phase transformer.
But I have used a single phase interstage (bifilar); a single ended to push pull interstage; a push pull primary to push pull interstage; and a tapped inductor (autotransformer) phase splitter.
None were perfect, but all were quite good.
If you use a pair of 10Meg Ohm / 10pF probes, and use a signal generator and 1k series resistor to drive a bifilar wound primary (1k = rp of the tube you will use to drive in the actual amplifier), then the probes will have extremely low load on the secondary (much less than any load resistor across the secondary).
Just my opinions.
Last edited:
I see, there is one wiring I have not tried, maybe loading was my problem all along of one coil taking off over the other. What if I use a CT secondary AND load it like a virtual CT I did above. Each side would have its own half of the coil and they would each see matched load? I have not tried that yet. All the CT splitting I have tried so far was with no resistors either PP or as a virtual gnd. And all the transformers I've tested with CT had one side of the coil run away from the other side above 10khz.
Are you saying the key to making it more bullet proof in a real circuit is to still use a CT secondary but load it with another virtual ground and grounding the CT? Maybe thats my mistake all along when I measured the CT transformers, the CT was depending on the ground common of the two long O probes? So that let one side run away from the other.
Leakage inductance might not be the same for both halves of the xfmr winding. That will unbalance it.