RF balun, impedances and ...wtf?

[Yansi]

Hello
just came across an interesting stuff. Was looking for a reference schematic how to terminate properly an RF IC with differential current outputs. The easiest way is of course using a proper RF transformer. While looking for some inspiration for part numbers and schematic concepts, I found a few many of them... let's say interesting.

Looking at the datasheet, for example this one: https://ww2.minicircuits.com/pdfs/TC1-1T+.pdf  They do say the transformer ratio is 1:1. Does that mean that all three windings have the same number of turns (1:1:1) or does that mean the center tap is really a center tap, and the ratio is indeed 1:0.5:0.5? I think the latter should be right.

Now looking at some amateur implementations of such balun for the RF IC, I am becoming a little confused even though I should understand quite well how the balun works. See the example #1 here:

The T1-1T is obviously a 1:1 transformer, according to the datasheet. By using any of the two assumptions about the turn ratios above, the circuit just ain't right, or is it? Why is he transformer terminated both sides? Does not make sense.

Example #2. This seem not right either.

This is even more crazy. I found the TTWB4 transformer to be 1:4 ratio. So 16x gets the impedance transformed.  Again, why is the transformer doubly terminated? Doesn't make sense to me.  Don't you just need to terminate once, to convert the differential current outputs into single ended voltage out at 50ohm?

What I think should be enough is just use the 1:1 balun (1:0.5:0.5 like the T1-1T or TC1-1T+), terminate both leads from the RF IC with 25ohm and voila, have a 50ohm out? 
OR use the same transformer, current drive the primary directly and then terminate single resistor 50ohm on the secondary?  Should not this be enough?


Am I missing something or am completely dumb? As I could find a shitton of other schematics on the web, that has the balun implemented wrong, as far as I can tell.  The only schematic I could find in a minute or two that seemed to be correct (using only two 25ohm resistors) was the manufacturer EVALuation board for that IC.

[T3sl4co1l]

You'd have to e-mail them to be sure.  The datasheet does not specify port impedances or ratios.

It might be reasonable to assume the secondary is CT'd, and 50 ohms each side (so the overall ratio is 50 ohms to 25 ohms -- because the 50 ohm loads act in parallel -- or sqrt(2)/2, which is an inconvenient ratio to wind a transformer at, so it must have a dB or two of mismatch), because they need to measure the amplitude and phase balance somehow, and that will be into a 50 ohm system.

But if it's another impedance, all the measurements are still perfectly valid: everything's in dB, and one would assume they used appropriate matching networks, de-embedded with the fixture, to determine the measurements.  So it's not necessarily the case that the test environment around the transformer was 50 ohms.

Note also that a transformer does have a characteristic impedance, but it can still be useful at different impedances, as long as you don't mind reduced performance (bandwidth, loss, maximum power).  This might account for anything using oddball impedances, but the same ratios (like 100:50:50 ohms, for a 1.4:1:1 winding).

Also, anywhere they're putting termination resistors on both sides, it may well be that they're expecting a mismatch, and are swamping that just in case.  Maybe it is a 1:1:1 (i.e., 50:200:200 ohm) transformer, but no one cares?

Tim

(P.S. I probably still got the ratios wrong. Hey, you get what you pay for. )

[rfeecs]

or does that mean the center tap is really a center tap, and the ratio is indeed 1:0.5:0.5? I think the latter should be right.

You are right.  Both sides have the same number of turns, just one has a center tap.

When in doubt, look at the S-parameters:
https://www.minicircuits.com/WebStore/dashboard.html?model=TC1-1T%2B

They have 5 port S-parameters!  I guess that makes sense for 5 pins.  Put it into a circuit simulator and it works as described above.

[dmills]

This gets subtle...

Current outputs are nominally high impedance (That is sort of the definition of a current source), but they have a compliance voltage range, as well as a current range (often programmable on things like DDS parts).

The trick is to terminate in a manner that gets you maximum power output into whatever your reconstruction filters designed source impedance is while also respecting the devices voltage compliance range at all frequencies, and while avoiding DC in the transformer windings as far as possible (which causes non linearities). Sometimes getting max power means a primary side impedance not equal to 50 ohms, in which case a ratio other then 1:1 is required.

The trap is that at the sample rate these parts usually run, the transformer leakage inductance is not negligible, which is why you often see resistive termination on the primary side (It ensures compliance voltage limits are respected for energy up where the leakage inductance makes the transformer appear open circuit).

Not saying there is not a lot of stupid design out there, but this is somewhere where the non ideal behaviour of real parts starts to matter.

73 Dan.

[T3sl4co1l]

Note that, for CC outputs, you can use the pin capacitance as part of the shunt capacitor in a lowpass or bandpass filter.  No source termination is needed, and you get double the power output as a result.  (The single-terminated filter prototype is a bit different from the more common double-terminated case.)  This is fairly common with RF amplifiers, where the (collector or drain) output impedance is very high.

The downside is this: not only is the voltage (at the source) about double (because the impedance is about double), but the voltage gain varies up and down across the passband (that is, the voltage developed at the source).  If you have a hard cutoff for the voltage compliance range, it might not be worthwhile at all to pursue a single-terminated filter.  In other words, the double-terminated filter has flat and predictable voltage gain properties, so you can operate closer to the voltage limit at all frequencies.

Note also that, if the transformer is well behaved, you can integrate it anywhere within the filter, with a little adjustment necessary to account for its inductances and capacitance.  It might be desirable to have the first pole or two of the filter to be balanced, at the source, so the source sees a controlled (CM and diff) impedance.  Then the rest of the filter follows the transformer.  That way, you only need a few extra components to provide the balanced part of the filter, rather than a fully balanced filter, and that way you don't suffer the uncontrolled parasitics of the transformer.

Tim

[Bud]

They do say the transformer ratio is 1:1. Does that mean that all three windings have the same number of turns (1:1:1) or does that mean the center tap is really a center tap, and the ratio is indeed 1:0.5:0.5?

It is usually the impedance ratio which is specified for RF transformers. CT would sit at half of that impedance value.

This is even more crazy. I found the TTWB4 transformer to be 1:4 ratio. So 16x gets the impedance transformed. 

No, 1:4 is impedance ratio, so it would be 50:200 Ohm transformation.

Am I missing something or am completely dumb?

As dmills said. Read the IC datasheet for voltage compliance range and then see if the numbers add up in that design.

[Yansi]

Thank you for making some good points. I have already noted the importance of being compliant to the compliance range of the DAC. (no pun intended). 

TTWB4 indeed, it is an impedance ratio! I am not quite used to denote transformers using impedance ratios, only turn ratios. So I saw just the 1:4. I stay being corrected, thanks for noticing!  But it still does not make that circuit correct, as the R21 (including C32) are both not needed.

Thinking about the circuit, putting a single termination on the secondary, might even produce better impedance match by a bees dick, as the transformer's parasitics will get shunted by the termination. But it very depends on how well behaved the transformer is, if it will make any difference.

The balun should be for an Analog Devices DDS chip, so the output current indeed (especially with that one I need) quite limited (5mA min, 10mA typ. peak) , including the voltage range being quite limited.  I will double check, when it will be put into design.

[chrisl]

But it still does not make that circuit correct, as the R21 (including C32) are both not needed.

It all depends on what is connected to C32....     If a high Z voltage driven device with a DC bias is followed then both R21 and C32 are required.

[KE5FX]

This stuff is a mess.  I've occasionally seen reference designs from major manufacturers that have invalid diff-to-SE impedance matching and/or double termination.  There are some subtleties: consider that a transformer with a 1:1 impedance ratio has a 1:1 turns ratio.  But if there's a center tap, the turns ratio on either side of it must be 2:1, and the resulting impedance ratio at either end of the winding referred to the center tap is 4:1. 

In other words, a 50-ohm match looking into the primary of the T1-1T-KK81 transformer could be obtained in any of three different ways:

• A single 12.5-ohm resistor on one half of the secondary winding, and nothing across the other half
• A single 50-ohm resistor across the whole secondary, nothing at the center tap
• A pair of 25-ohm resistors, one going to RF ground on each side of the center tap (as shown in your example)

Why does the third option require two 25-ohm resistors instead of two 12.5-ohm resistors?  Because the voltages at either end of the secondary winding are out of phase.  With a symmetrical load, this makes the center tap effectively disappear.  Your example circuit explicitly says it's got a "differential current output," and that implies that the output Z is infinite.  You will have a "50 ohm match" at the primary only if that condition prevails, which it probably won't.  That's fine if whatever's driving it doesn't care.  In any case, you would not use a circuit like that if you wanted to drive a 50-ohm load, because it would be double-terminated.

A traditional 2:1 balun by itself will suppress common-mode noise and even-order distortion components, but that's about all it will do.  If you don't care about either of those, you might as well go single-ended.  For best overall performance, consider the 180-degree couplers in the Mini-Circuits catalog instead of a traditional balun. 

With the LMK61E2's LVPECL outputs, for instance, the data sheet recommends either using a balun or just one of the single-ended outputs, leaving the other terminated in an unused 50R load.  A 2:1 balun will not give you any additional output power over the SE output, but you can pick up another 2 dBm with an SBTCJ-1WX+.   I keep expecting the eval board designers and data-sheet authors to realize that those parts exist, but it doesn't seem to be happening.

Edit: also note that the hybrid coupler is really needed only if the source actually cares about 50-ohm output at each of its differential ports.  At low frequencies it probably doesn't, in which case you may get the same output amplitude boost by using the (much cheaper) 1:1 transformer with no resistive termination at all.  Voltage compliance may be a problem if you ever let it run open-circuit, though.

[G0HZU]

Macom are generally much better/clearer at presenting winding information in their datasheets for similar RF transformers. I'm not so impressed by their web based component selector though...

https://www.macom.com/products/passives/transformers--baluns


The datasheet example below is for a 75 ohm transformer but they do give some winding/turns information in the datasheet.

https://cdn.macom.com/datasheets/MABA-008965-CF1160.pdf

[Yansi]

But it still does not make that circuit correct, as the R21 (including C32) are both not needed.

It all depends on what is connected to C32....     If a high Z voltage driven device with a DC bias is followed then both R21 and C32 are required.

There was an amp behind, but it had a separate cap on its input. This is just a duplicate cap for no reason.