Single Ended to Differential Transformers. What's The Difference?

[vk4ffab]

I see these used, often interchangeably, in schematics. So what is the actual difference between A and B other than galvanic isolation between input and output on A. They are both wound 1 to 1. I honestly do not see any difference. Assume B has its ground connection connected, dummy who drew the schematics did not do an electrical rules test HAHAHA

[szoftveres]

If you leave out the 100nF capacitors then obviously the voltage balun provides DC isolation whereas the current balun provides DC ground reference. And I guess that's it, i.e. whether (or not) isolation or DC reference is needed at the secondary side. Just like isolation transformer, or auto-transformer.

[T3sl4co1l]

Eww, the connectors aren't on grid.  Hmm, B's ground is floating, too.

The phasing of B is... suspect, but maybe not out-and-out wrong.

Well, assuming those are fixed in the most obvious manner:

These are a simple transformation (ha) of a circuit, with a simple rule: voltage-mode vs. current-mode.

The reason cannot be understood with an ideal transformer: in that case, there is no difference, no advantage to one or the other.

Understanding the nonideal transformer brings relevant insight:



In A, stray capacitance (or even more nuanced: transmission line effects) affects balance and bandwidth, setting an upper cutoff; magnetizing inductance (and coupling capacitance where applicable) sets a lower cutoff.

In B, magnetizing inductance affects balance, but bandwidth is essentially unlimited (the two windings can be a single transmission line).  But I think a mistake has been made here:

Mentally flip around the bottom winding (pins 1/2), so that the bottom output (nets aren't labeled, so I'll use C4/C9 to identify them instead; so, C9) comes from the left side and the dot is on the right.  Now we have a common-mode choke, going between input+C9, through a transformer core, to C4+GND.  If J3 has a rising step applied, C9 voltage rises immediately, given by the divider of (transmission line) Zo into C9 termination resistance (let's say it's R to GND).  If Zo = R, this is 50%, after a delay of ~zero.  One electrical length of time later (the propagation delay of the TL), C4 rises to half, and reflects back its load (let's say R again).

It's a 2:1 network (say 100Ω input, 50+50Ω output), but a balun it ain't.  It's a power combiner.

If T5 windings have dots on the same side, it's a CMC and a "current balun" in the usual way (follow the same analysis as above).

Tim

[vk4ffab]

If you leave out the 100nF capacitors then obviously the voltage balun provides DC isolation whereas the current balun provides DC ground reference. And I guess that's it, i.e. whether (or not) isolation or DC reference is needed at the secondary side. Just like isolation transformer, or auto-transformer.

Thanks, good to know I am not going dumb LOL

Eww, the connectors aren't on grid.  Hmm, B's ground is floating, too.

The phasing of B is... suspect, but maybe not out-and-out wrong.

Well, assuming those are fixed in the most obvious manner:

These are a simple transformation (ha) of a circuit, with a simple rule: voltage-mode vs. current-mode.

The reason cannot be understood with an ideal transformer: in that case, there is no difference, no advantage to one or the other.

Understanding the nonideal transformer brings relevant insight:



In A, stray capacitance (or even more nuanced: transmission line effects) affects balance and bandwidth, setting an upper cutoff; magnetizing inductance (and coupling capacitance where applicable) sets a lower cutoff.

In B, magnetizing inductance affects balance, but bandwidth is essentially unlimited (the two windings can be a single transmission line).  But I think a mistake has been made here:

Mentally flip around the bottom winding (pins 1/2), so that the bottom output (nets aren't labeled, so I'll use C4/C9 to identify them instead; so, C9) comes from the left side and the dot is on the right.  Now we have a common-mode choke, going between input+C9, through a transformer core, to C4+GND.  If J3 has a rising step applied, C9 voltage rises immediately, given by the divider of (transmission line) Zo into C9 termination resistance (let's say it's R to GND).  If Zo = R, this is 50%, after a delay of ~zero.  One electrical length of time later (the propagation delay of the TL), C4 rises to half, and reflects back its load (let's say R again).

It's a 2:1 network (say 100Ω input, 50+50Ω output), but a balun it ain't.  It's a power combiner.

If T5 windings have dots on the same side, it's a CMC and a "current balun" in the usual way (follow the same analysis as above).

Tim

Thanks, that confirms most of how I was thinking. OH and BTW I broke my own rules never ask a question full of incorrect information. I never even noticed the phasing dots, these are my schematics for my personal use and while I know where the phasing should be, the schematic does not show that which to others is confusing. But you answered all permutations and that certainly helped clear things up in my mind.

What got me thinking about this was looking over a schematic I had downloaded years ago and I was going to email the author (a ham) to ask him why he chose 1 configuration over the other, but it turns out he is SK (dead). But anyway, I am designing and building a new transceiver from the ground up and am trying to incorporate all the things I have learned over the last couple of years and hopefully end up with a thing that is good enough to be a daily driver. Build lots of crap in my time, but I think this time I should be golden, or at least hope I am LOL

[G0HZU]

If you can get access to a 4 port VNA it is possible to measure the transformer as a 4 port device as in the image below.

This can then be exported from the VNA as an s4p file and it can be used as a component with four connections in a linear RF simulator.

See the second image to see the s4p model configured as a 4:1 transmission line transformer. This assumes the transformer is wound with bifilar windings to mimic a 100R transmission line.

It could also be configured as in your initial schematics and the impact on performance can be compared for each topology. This will show the differences you can expect from a 'real' transformer, due to the transmission line effects and the core material etc.

It would also show what would happen if you connected it with the wrong phasing. You only need to export one s4p model from the VNA to explore all this stuff in a simulator.

Providing you measure it correctly with the VNA and you export a valid s4p model, the simulated results should be extremely close to the results you would get with a real transformer when it is connected up in the various different ways.

[mark03]

Hmmm.  For those of us without four-port VNAs, would this method be applicable?

https://scikit-rf.readthedocs.io/en/latest/examples/metrology/Measuring%20a%20Mutiport%20Device%20with%20a%202-Port%20Network%20Analyzer.html