-
#125 Reply
Posted by
precaud
on 09 Aug, 2018 00:55
-
That's was mistake in my conceptual thinking, whatever. I assumed, measuring the signal output at one side of the power divider would compensate for the total loading effect on the source. But it doesn't, it attenuates the loading effect to a lower impedance seen by the VNA - assuming using a resistive power divider. Anyway, measuring both the transformer input and output with high impedance R and A, then calculating A/R removes the loading effect of the transformer from the result. That's nearer to a virtual zero source impedance than the power divider method.
Well there can be good reasons for looking at it either way. I'm just concerned that future readers of this thread will look at all the measurements as if they're the same, when they're not.
rx8pilot:
You'd normalize the "FRA" setup by setting the VNA to sweep A/R phase and amplitude, the replace the transformer by a "through" and normalize. The 3577A then normalizes these traces (resulting in "A/R/D1" / "A/R/D2" as traces). This is at least how I'd normalize this setup. precaud pls confirm.
Yes, and make sure whatever you'll be loading the xfmr with is in place for the sweep you use for normalizing.
precaud:
At above some 10MHz the high impedance input mode of the 3577A doesn't work too well anymore, because of the 50 Ohm impedance of the internal and external cables involved. For lower frequencies, high Z should work fine and wouldn't require normalization anyway.
Yes, for true high-Z measurements that wold be true. But not so much for a device with a 50 to 75 Ohm term. You can use normalize to test the impedance sensitivity of the measurement setup. Ex: normalize a "thru" sweep with 50 Ohm term, then change to a 75 Ohm term and it will plot the deviation. Anything that is not a flat line is a problem... phase deviation would be the most sensitive.
That's phenomenal bandwidth you're getting through that xfmr! What is it rated for?
-
#126 Reply
Posted by
rx8pilot
on 09 Aug, 2018 01:10
-
This thread has thrown me down a rabbit hole of thoughts, self-education, and experiments. Only learning more about magnetics and properly measuring them with a 25-year-old VNA can get me out of the hole.
-
#127 Reply
Posted by
T3sl4co1l
on 09 Aug, 2018 03:05
-
This thread has thrown me down a rabbit hole of thoughts, self-education, and experiments. Only learning more about magnetics and properly measuring them with a 25-year-old VNA can get me out of the hole.
If it gets more people measuring, it can't be a bad thing.
Tim
-
-
That's phenomenal bandwidth you're getting through that xfmr! What is it rated for?
Don't know. Found it by chance on the evilbay (looking for North Hills after someone mentioned their transformers here), and decided to give it a try. No datasheet, no information, no similar part number found on the North Hills website. BTW, they've got some app notes about transformers and other interesting stuff.
So I just could guess: It's a RGB + sync video transformer, so I expected it to have wide BW to be useful, or it would be a common mode choke (North Hills calls them "Humbucker"). Former analog monitors having BNC RGB inputs achieved video bandwiths in excess of 100MHz.
http://www.nhsignal.com/products-wideband-video-isolation.html
-
#129 Reply
Posted by
ehughes
on 09 Aug, 2018 21:01
-
For what its worth, Omicron got at least 1 sale as a result of the tear down. I just ordered one for QA checking out transducers. Once it is up an running, I'll post some measurement results.
-
#130 Reply
Posted by
precaud
on 11 Aug, 2018 18:01
-
-
#131 Reply
Posted by
T3sl4co1l
on 11 Aug, 2018 18:12
-
Bandwidth is maximum at Zo, and drops off slowly for Z <> Zo. The first-order approximation is thus: if LL and Lp are constant, then the LF cutoff is Z / (2*pi*Lp) and the HF cutoff is Z / (2*pi*LL).
For the frequencies and materials we're talking about here, this should be pretty close. Note that the HF cutoff is actually a transmission line effect, not a dominant pole, so it will have slightly different response into other resistances (I forget if this is better or worse for BW). For worse quality materials (like laminated iron), the permeability dependence on signal level and frequency is probably an important factor.
Tim
-
#132 Reply
Posted by
precaud
on 12 Aug, 2018 15:25
-
This is a pretty interesting video, Ray Ridley comparing the three commercially-available isolation xfmrs at low frequencies.
Unit "P" is clearly the Picotest.
Unit "O" is the Omicron.
I do wish he had also plotted transmission curves vs freq over the whole range of the transformers, measured into the same load R.
I had the Ridley 15MHz unit and it had a 10dB dip starting at around 10kHz, regardless of load.
-
-
The Omicron Labs B-WIT 100 is only supposed to be driven at 0dBm not 13dBm. Interesting video from Omicron Labs where product "R" gets tested.
-
#134 Reply
Posted by
precaud
on 12 Aug, 2018 16:38
-
Yeah, the 15MHz Ridley unit I had looked similar in terms of flatness, though smoother at the high end than that one. The freq resp dip bugged me and so I sold it (perhaps hastily).
However, one could argue that, as long as the response does not change with level or load, and the variations are not extreme, for most FRA work, ultimate flatness is not critical...
-
#135 Reply
Posted by
Wolfgang
on 12 Aug, 2018 17:29
-
Unfortunately, neither Ridley nor Omicron tried to be especially objective with their presentations.
What would have been helpful are S21 curves that go over the whole frequency range, at given load impedances (e.g. 1/5/10/20/50 Ohms),
plus a curve for the allowable signal level for, say, 1% distortion over frequency, also for different load impedances. Also phase response is important
to know about.
It is very clear than one single device hardly fits all needs (High bandwidth, tolerant to strong signals with no distortion, low capacitances, not bulky ...)
There will versions with a deep low frequency limit - they are heavy and bulky, have a high coupling capacitance
There are smaller ones, which are small signal only, have medium frequency ranges, but lower capacitances, ...
...and the HF ones with very high bandwidth, very small signal tolerance, and small capacitance.
All marketing claims made in this industry has to be taken with a huge ROCK of salt. So have these videos.
-
#136 Reply
Posted by
Kleinstein
on 12 Aug, 2018 19:36
-
If a really good injection transformer is that difficult, I would consider thinking about an active solution, with optical isolation, thus a kind of analog opto-coupler. In a simple form something like 2 photo-diodes in parallel to a load resistor. Without amplification it would be likely limited to low power (e.g. a few 10 mV) - but chances are it could work from DC to beyond 10 MHz with low coupling capacitance. If it needs an amplifier for the output this could be battery powered for good isolation.
-
#137 Reply
Posted by
Wolfgang
on 12 Aug, 2018 19:44
-
Yes, this is a valid approach. If it is battery-powered and has a very fast (also low noise) optical isolator it competes very well with transformer solutions, especially at the low frequency range.
-
#138 Reply
Posted by
ehughes
on 16 Aug, 2018 18:12
-
We just got our Bode100. It is a very nice piece of equipment and the software that come along is useful. We are doing some analysis of a 1.8MHz piezo ring. Specifically, we are able to evaluate the epoxy in a transducer assembly.
Yet nice piece of equipment and easily worth the money for our QA setup.
-
-
I was browsing the Picoscope site (I'm interested in their flexres 16bit/8bit scopes with arb gen that can do FRA) and found quite a nice video link about measuring loop gain with a DIY signal injection transformer. Looks like a nanocrystaline core from a common mode choke with about 7 metres of TP each for primary/sec - its at 9:00 minutes in, but the whole video is quite well done. Didn't achieve the bandwidth that others have here tho
https://youtu.be/wKs8VyERZXU
-
#140 Reply
Posted by
precaud
on 20 Aug, 2018 04:32
-
Here's a good reason NOT to use the Jensen IsoMax transformer for gain- and phase- margin measurements above 20kHz - above which the phase response changes with impedance... caused by the bandwidth increasing with increasing load R. The North Hills units are all better in this range.
-
#141 Reply
Posted by
DualTriode
on 20 Aug, 2018 21:34
-
Hello,
What you have shown is that the leakage inductance and load resistance form a low pass filter.
What you have not shown is that accuracy of the VNA Bode Plot is diminished.
The Bode Plot Phase Margin is independent of the transformer phase.(within limits) The VNA software has provisions to compensate for the attenuated transformer HF gain. (within limits)
DT
-
#142 Reply
Posted by
precaud
on 20 Aug, 2018 23:01
-
What you have not shown is that accuracy of the VNA Bode Plot is diminished.
I haven't "shown" it, but I guarantee you it definitely is diminished, unless some post-processing is used to correct for it (Open/Short/Load). But even then, if the change is not fairly linear over the entire measurement range, the compensation math won't b able to correct for it accurately.
The Bode Plot Phase Margin is independent of the transformer phase.(within limits).
Not in my experience.
The VNA software has provisions to compensate for the attenuated transformer HF gain. (within limits)
Yes, some VNAs have "provisions" for it, but I have never seen any example of someone "normalizing" their VNA with the transformer in place prior to making a control loop measurement.
-
#143 Reply
Posted by
rx8pilot
on 20 Aug, 2018 23:12
-
Curious why you started at 4R7.
What are you using to test it? How did you calibrate? I may do the same test to see if the results are similar (although I may use higher value loads) - but would want to match what you did.
-
#144 Reply
Posted by
precaud
on 20 Aug, 2018 23:36
-
Curious why you started at 4R7.
I didn't; I started at 1R0, and ended at 4R7
Most of what I use these for (output impedance measurements) loads the transformer between 1R0 and 5R0 on the high end. See the setup in this post:
https://www.eevblog.com/forum/blog/eevblog-1104-omicron-labs-bode-100-teardown/msg1720265/#msg1720265For the measurement I posted, the setup is the same xfer function setup as others I've posted; both inputs at 1M, R measures xfmr input, T measures the output.
What are you using to test it?
AP Instruments 102B
How did you calibrate?
Huh? You want me to explain the 102B's cal procedure?
I may do the same test to see if the results are similar (although I may use higher value loads) - but would want to match what you did.
Please do. I get the same results regardless of the analyzer used.
-
-
@voltsandjolts Nice video from Picoscope, thanks. Something has been bugging me about using 1:1 transformers and them terminating the secondary with 5 or 10 ohms. I would have gone for 2:1 or most probably 3:1 for a 5 ohm load. If I get time this week I will have a stab at a 3:1 transformer and see how it goes.
-
#146 Reply
Posted by
DualTriode
on 21 Aug, 2018 05:21
-
Hello,
Open/Short/Load calibration is used for the impedance accessory test fixture and is not part of the VNA frequency gain / phase plot.
If you look in the Keysight E5061B User’s Manual you will the Open/Short/Load calibration procedure for Impedance Analysis option and optional impedance test fixture(s).
The available adjustment for the Frequency Gain / Phase Margin test is signal level. There is no adjustment needed for transformer phase.
The available test signal adjustment is for amplitude. The generator output level is adjusted to be within the small signal range of the DUT error amplifier. Over the LF range the signal level may need to be trimmed to reduce noise. Over other frequency bands the signal may need other tweaking. This tweaking is a iterate process looking for the cleanest Bode plot. This tweaking is more about the DUT than the injection transformer. (within limits)
DT
-
#147 Reply
Posted by
Wolfgang
on 21 Aug, 2018 11:05
-
I agree completely.
The properties of your isolation transformers determine your dynamic range, even if the phase and amplitude response can be calibrated out by a good VNA.
What I do not understand in a lot of cases why manufacturers insist so much on a 1:1 ratio. I do not know a lot of examples except PFC where high signal injection amplitudes at low frequencies are really needed; In the absolute majority of cases you need rather small amplitudes in the millivolts range in order not to drive the DUT out of its linear range.
I did it with COTS ISDN transformers and a matching network., and it worked fine.
-
#148 Reply
Posted by
precaud
on 21 Aug, 2018 12:42
-
Open/Short/Load calibration is used for the impedance accessory test fixture and is not part of the VNA frequency gain / phase plot.
My point exactly. So what exactly are you referring to when you said "The Bode Plot Phase Margin is independent of the transformer phase.(within limits)" ?
If you look in the Keysight E5061B User’s Manual you will the Open/Short/Load calibration procedure for Impedance Analysis option and optional impedance test fixture(s).
Yes, its unfortunate E5061B users need to buy an optional fixture to do it. We're not limiting our discussion to any particular analyzer. What one are you using?
The available adjustment for the Frequency Gain / Phase Margin test is signal level. There is no adjustment needed for transformer phase.
And that's where we disagree. If you make a measurement in a freq range where the xfmr magnitude and/or phase is not linear, it transfers directly into your measurement. How could it be otherwise? If would try it, you would see it is so.
The available test signal adjustment is for amplitude. The generator output level is adjusted to be within the small signal range of the DUT error amplifier. Over the LF range the signal level may need to be trimmed to reduce noise. Over other frequency bands the signal may need other tweaking. This tweaking is a iterate process looking for the cleanest Bode plot. This tweaking is more about the DUT than the injection transformer. (within limits)
Yes, that is for control loop testing. For impedance testing, the amplitude considerations are different. But it's not either/or; the same transformer can be used for both measurements, as long as you characterize it at the load it will be seeing, and use it within that range.
No matter what, the transfer response of the xfmr at the load it will see needs to be paid attention to.
Something has been bugging me about using 1:1 transformers and them terminating the secondary with 5 or 10 ohms. I would have gone for 2:1 or most probably 3:1 for a 5 ohm load. If I get time this week I will have a stab at a 3:1 transformer and see how it goes.
For impedance testing, I agree, and look forward to what you come up with. I'm gathering bits together to do a 2:1. For control loop, a 1:1 into 10 Ohms or above should be sufficient.
What I do not understand in a lot of cases why manufacturers insist so much on a 1:1 ratio.
Because they're easy to make. And they generally do the job.
-
-
Open/Short/Load calibration is used for the impedance accessory test fixture and is not part of the VNA frequency gain / phase plot.
My point exactly. So what exactly are you referring to when you said "The Bode Plot Phase Margin is independent of the transformer phase.(within limits)" ?
So if you look up the AP Instruments Model 102B Manual (I found one here:
http://www.apinstruments.com/files/102Bman.pdf), Page 54 (PDF numbering) / 50 (document numbering) shows your typical FRA setup used. The VNA calculates the frequency response from the ratio of its inputs Channel A and Channel B. So the Source impedance and transformer phase / amplitude cancels out from the equation. No calibration required, just a good matching between Channel A and B (which one expects from a decent VNA).
Within limits of course, since one needs a minimum amplitude of source signal coupled through the transformer to get a useful signal above the noise floor.
And that's where we disagree. If you make a measurement in a freq range where the xfmr magnitude and/or phase is not linear, it transfers directly into your measurement. How could it be otherwise? If would try it, you would see it is so.
Same reason as above, the measurement of Channel A / B ratio cancels the transformer out of the equations.