EMRFD Message Archive 3513

Message Date From Subject
3513 2009-09-02 06:43:07 w1kilofoxtrot Measuring wideband twisted wire transformers
I've just been blessed with an HP 8753 network analyzer, and I'm trying to learn how to use it while I also work on a 20W, 50V MOSFET amplifier.

According to my calculations, I should be able to get a good output match with a 4:1 (2t:1t) transformer. Measuring return loss is pretty simple; I terminate the 200 ohm side with a resistor/capacitor combination that approximates the push pull output stage transistor characteristics, and look at a sweep.

What I'm having a moment with is how to measure the through loss, given I have a 50 ohm source, and the transformer is designed to give me a 4:1 response. Should I add a 150 ohm resistor to the source, and terminate the 50 ohm side in the detector? Or, is there a better way to make this measurement? I'm not quite sure how to calibrate the instrument in this topology... I sure don't want to add another transformer to the mix to get back to 50 ohms...

BTW, I'm planning on using a trifilar design, but will change it a bit so the output and input are on opposite sides of the ferrite cores I'm using. I'll let you know how it comes out!

Thanks for any hints you might be able to offer...

73,
Steve
W1KF
3514 2009-09-02 07:06:37 Chris Trask Re: Measuring wideband twisted wire transformers
>
> According to my calculations, I should be able to get a good
> output match with a 4:1 (2t:1t) transformer. Measuring return
> loss is pretty simple; I terminate the 200 ohm side with a
> resistor/capacitor combination that approximates the push pull
> output stage transistor characteristics, and look at a sweep.
>
> What I'm having a moment with is how to measure the through
> loss, given I have a 50 ohm source, and the transformer is
> designed to give me a 4:1 response. Should I add a 150 ohm
> resistor to the source, and terminate the 50 ohm side in the
> detector? Or, is there a better way to make this measurement?
> I'm not quite sure how to calibrate the instrument in this
> topology... I sure don't want to add another transformer to
> the mix to get back to 50 ohms...
>

You make two identical transformers and connect them back-to-back. The loss of each transformer is then half of what you measure. It's also a better way to measure the return loss.


Chris

,----------------------. High Performance Mixers and
/ What's all this \ Amplifiers for RF Communications
/ extinct stuff, anyhow? /
\ _______,--------------' Chris Trask / N7ZWY
_
3515 2009-09-02 11:42:14 Lou Burke Re: Measuring wideband twisted wire transformers
I would think that you could simply terminate the 200 ohm port of the transformer and look into the 50 ohm port with the analyzer to measure 50 ohms across you intended bandwidth or viseversa.

Lou, W7JI

----- Original Message -----
3516 2009-09-02 11:55:57 Chris Trask Re: Measuring wideband twisted wire transformers
>
> > I've just been blessed with an HP 8753 network analyzer, and I'm
> > trying to learn how to use it while I also work on a 20W, 50V MOSFET
> > amplifier.
> >
> > According to my calculations, I should be able to get a good output
> > match with a 4:1 (2t:1t) transformer. Measuring return loss is pretty
> > simple; I terminate the 200 ohm side with a resistor/capacitor
combination
> > that approximates the push pull output stage transistor characteristics,
> > and look at a sweep.
> >
> > What I'm having a moment with is how to measure the through loss,
given
> > I have a 50 ohm source, and the transformer is designed to give me a 4:1
> > response. Should I add a 150 ohm resistor to the source, and terminate
the
> > 50 ohm side in the detector? Or, is there a better way to make this
> > measurement? I'm not quite sure how to calibrate the instrument in this
> > topology... I sure don't want to add another transformer to the mix to
get
> > back to 50 ohms...
>
> I would think that you could simply terminate the 200 ohm port of the
> transformer and look into the 50 ohm port with the analyzer to measure
> 50 ohms across you intended bandwidth or viseversa.
>

Just try doing that with a VNA that has 50-ohm terminals. That's what
the question was about.

Chris

,----------------------. High Performance Mixers and
/ What's all this \ Amplifiers for RF Communications
/ extinct stuff, anyhow? /
\ _______,--------------' Chris Trask / N7ZWY
_
3517 2009-09-02 12:06:16 Niels A. Moseley Re: Measuring wideband twisted wire transformers
Chris Trask wrote:
>>> I've just been blessed with an HP 8753 network analyzer, and I'm
>>> trying to learn how to use it while I also work on a 20W, 50V MOSFET
>>> amplifier.
>>>
>>> According to my calculations, I should be able to get a good output
>>> match with a 4:1 (2t:1t) transformer. Measuring return loss is pretty
>>> simple; I terminate the 200 ohm side with a resistor/capacitor
> combination
>>> that approximates the push pull output stage transistor characteristics,
>>> and look at a sweep.
>>>
>>> What I'm having a moment with is how to measure the through loss,
> given
>>> I have a 50 ohm source, and the transformer is designed to give me a 4:1
>>> response. Should I add a 150 ohm resistor to the source, and terminate
> the
>>> 50 ohm side in the detector? Or, is there a better way to make this
>>> measurement? I'm not quite sure how to calibrate the instrument in this
>>> topology... I sure don't want to add another transformer to the mix to
> get
>>> back to 50 ohms...
>> I would think that you could simply terminate the 200 ohm port of the
>> transformer and look into the 50 ohm port with the analyzer to measure
>> 50 ohms across you intended bandwidth or viseversa.
>>
>
> Just try doing that with a VNA that has 50-ohm terminals. That's what
> the question was about.
>
> Chris

Hi Chris and gang,

Why wouldn't one put a 150 ohm resistor in series with the 200-ohm
output; the VNA+150 ohm will correctly terminate the 200 ohm output. Of
course, you'd have to compensate the measurement results for the voltage
drop.

Regards,
Niels PA1DSP.
3518 2009-09-02 12:19:44 Lou Burke Re: Measuring wideband twisted wire transformers
Keep in mind that the transformer should see a resistive load. I measure broadband xmfr's all the time using this method with my Array Solutions 4170 analyzer

Lou
----- Original Message -----
3519 2009-09-02 12:48:42 w1kilofoxtrot Re: Measuring wideband twisted wire transformers
Now, why didn't I think of that? Given it's your article (QEX 2005) that I'm working from, I'll try to make two identical transformers for the through measurements.

I'll still want to look into the 50 ohm side, and see what it looks like with the 'realistic' termination
3520 2009-09-02 13:52:21 Chris Trask Re: Measuring wideband twisted wire transformers
>
> >
> > Just try doing that with a VNA that has 50-ohm terminals. That's
what
> > the question was about.
> >
> > Chris
>
> Hi Chris and gang,
>
> Why wouldn't one put a 150 ohm resistor in series with the 200-ohm
> output; the VNA+150 ohm will correctly terminate the 200 ohm output. Of
> course, you'd have to compensate the measurement results for the voltage
> drop.
>

Because then you don't have to compensate the measurement results, you
instead measure the loss directly.

Chris

,----------------------. High Performance Mixers and
/ What's all this \ Amplifiers for RF Communications
/ extinct stuff, anyhow? /
\ _______,--------------' Chris Trask / N7ZWY
_
3521 2009-09-02 13:55:21 Chris Trask Re: Measuring wideband twisted wire transformers
>
> > You make two identical transformers and connect them back-to-back.
> > The loss of each transformer is then half of what you measure. It's
> > also a better way to measure the return loss.
>
> Now, why didn't I think of that? Given it's your article (QEX 2005) that
> I'm working from, I'll try to make two identical transformers for the
through measurements.
>
> I'll still want to look into the 50 ohm side, and see what it looks like
> with the 'realistic' termination on the transistor side. That gets me a
> step closer to having a match when I power up. But, I'll also do a return
> loss with the far side terminated in 50 ohms.
>
> Should be a properly instructive exercise, I think.
>
> Thanks! I'll let you know how it comes out...
>

You'd be surprised how many people overlook the 2-transformer method of
measuring losses. It does require that you construct them carefully, but
then part of the learning process is to make them consistent, and that comes
with practice, practice, practice....

Chris

,----------------------. High Performance Mixers and
/ What's all this \ Amplifiers for RF Communications
/ extinct stuff, anyhow? /
\ _______,--------------' Chris Trask / N7ZWY
_
3522 2009-09-02 13:58:42 Niels A. Moseley Re: Measuring wideband twisted wire transformers
Chris Trask wrote:
>>> Just try doing that with a VNA that has 50-ohm terminals. That's
> what
>>> the question was about.
>>>
>>> Chris
>> Hi Chris and gang,
>>
>> Why wouldn't one put a 150 ohm resistor in series with the 200-ohm
>> output; the VNA+150 ohm will correctly terminate the 200 ohm output. Of
>> course, you'd have to compensate the measurement results for the voltage
>> drop.
>>
>
> Because then you don't have to compensate the measurement results, you
> instead measure the loss directly.
>
> Chris

You'd have to make two transformers instead of just one. Furthermore,
they have to be identical. At least for me, compensating is faster than
winding an additional transformer.

Niels.
3523 2009-09-02 14:27:30 Lou Burke Re: Measuring wideband twisted wire transformers
I guess I'm missing something here......my analyzer has a BNC input that wants to see 50 ohms. If I put the 50 ohm port of a 4:1 xmfr on the analyzer and terminate the 200 ohm port with a 200 ohm resistor and sweep the input over the desired range I see the 50 impedance or variations of the port impedance across the bandwidth of the xmfr.

Or I can look at the 200 ohm port and terminate the 50 ohm port and do the same test.

Lou
----- Original Message -----
3524 2009-09-02 14:59:27 Chris Trask Re: Measuring wideband twisted wire transformers
>
> I guess I'm missing something here......my analyzer has a BNC input that
> wants to see 50 ohms. If I put the 50 ohm port of a 4:1 xmfr on the
analyzer
> and terminate the 200 ohm port with a 200 ohm resistor and sweep the input
> over the desired range I see the 50 impedance or variations of the port
> impedance across the bandwidth of the xmfr.
>
> Or I can look at the 200 ohm port and terminate the 50 ohm port and do the
> same test.
>

Yes, you are missing something. He specifically stated that he wanted
to make an insertion loss test, not a return loss test. Those are two
completely different tests.

Chris

,----------------------. High Performance Mixers and
/ What's all this \ Amplifiers for RF Communications
/ extinct stuff, anyhow? /
\ _______,--------------' Chris Trask / N7ZWY
_
3525 2009-09-02 16:32:39 Lou Burke Re: Measuring wideband twisted wire transformers
Yes I did miss it.....sorry for wasting bandwidth.

Lou
----- Original Message -----
3526 2009-09-03 07:11:56 iq_rx Re: Measuring wideband twisted wire transformers
Hi Chris and all,

Good thread. Yes, the two transformer method is the way to go. That is a direct measurement of insertion loss.

At very high frequencies (>100 GHz or so) sometimes all we have is S11. After carefully calibrating the network analyzer, we obtain a measurement of transmission loss by terminating the opposite end of the transformer with a short, offset short, and matched load. The short is a perfect mirror at the far end of the network, so we get everything back out the input--minus the losses in the network. But the matrix math is complex (pun intended), and it's more than a bit counter intuitive that we can calculate transmission loss to a matched load from reflection loss using a short. It works, but even above 100 GHz we always try to make a direct S21 measurement whenever possible, using two identical transformers back to back. We are skeptical of our one-port transmissi
3532 2009-09-04 14:37:16 w1kilofoxtrot Re: Measuring wideband twisted wire transformers
Well, I made one transformer, and decided to see if the math worked out. All the following refer to the Mar/Apr 2005 QEX article Chris published.

It's a 2t:1t twisted 24 ga bal-un, to become the output transformer for my amplifier. Inductances on my core are 16.6 uH primary, 4.2 uH secondary. Makes sense, given the 4X change we'd expect.

Measuring the coupling coefficient gives me a k of .994, at 14 MHz. Seems a fair number so far.

I calculate a mutual inductance of 8.325 uH, and here's where it gets wierd. The primary leakage inductance calculates to -50 nH. (Equation 6) Yep, negative 50 nanohenries. I decided to suspend disbelief for a minute, and calculated the secondary leakage, (equation 7) which then results in 37.5 nH.

If I accept that result, the equivalent inductances look way small;
3533 2009-09-04 15:37:37 Chris Trask Re: Measuring wideband twisted wire transformers
Chris

,----------------------. High Performance Mixers and
/ What's all this \ Amplifiers for RF Communications
/ extinct stuff, anyhow? /
\ _______,--------------' Chris Trask / N7ZWY
_
3534 2009-09-04 16:40:09 w1kilofoxtrot Re: Measuring wideband twisted wire transformers
Chris,

Thanks for the additional reference. I'll read up and digest that.

I'm glad to see fundamentally I'm on track. One of my biggest concerns when I do analysis is imputing too much accuracy to my measurements. I'm pretty sure that most of what I do is 3 significant digits as at best, so I do the round off before I do the math. I can see where the accuracy of my little LC meter (basically this one: ironbark.bendigo.latrobe.edu.au/~rice/lc/index2.html) might be the limiting factor. My Fluke PM6304 is still en-route, and I don't know it's calibrati
3535 2009-09-04 17:18:53 Chris Trask Re: Measuring wideband twisted wire transformers
>
> Thanks for the additional reference. I'll read up and digest that.
>
> I'm glad to see fundamentally I'm on track. One of my biggest concerns
> when I do analysis is imputing too much accuracy to my measurements. I'm
> pretty sure that most of what I do is 3 significant digits as at best, so
> I do the round off before I do the math. I can see where the accuracy of
> my little LC meter (basically this one:
> ironbark.bendigo.latrobe.edu.au/~rice/lc/index2.html) might be the
> limiting factor. My Fluke PM6304 is still en-route, and I don't know
> it's calibration status.
>
> It's hard learning stuff 30 years after when you should have... but
> it's great to learn something new!
>

You should have been here many years ago when I first tried to learn how
to do this sort of thing. Nobody anywhere had put together a concise
procedure for measuring the model parameters for wideband transformers.
I've written three papers on this subject and with this last one I think I
finally got it right.

Even worse, a procedure for calculating the values for the matching
capacitors was not to be found anywhere. The values that Dye and Granberg
gave in their Motorola application notes were derived experimentally, or so
I'm told by people who were there at the time. It was thanks to a Philips
application note that I was able to figure out how to do it as a design
procedure.


Chris

,----------------------. High Performance Mixers and
/ What's all this \ Amplifiers for RF Communications
/ extinct stuff, anyhow? /
\ _______,--------------' Chris Trask / N7ZWY
_