EEVblog #1104 - Omicron Labs Bode 100 Teardown

[JS]

Hello,

All this talk of termination resistors.

Is this with a resistor intentionally placed across the secondary winding plus the resistance to ground internal to the VNA?

That means that we have the external resistor across the secondary coil in parallel with the instrument internal resistor to ground.

Different related topic.

It is not so much about adding additional resistance to extend or reduce bandwidth it is about resistance being used to tune or critically dampen the native inductance of the coil. Another way to think of it goes like this. Inductance is often added to RF circuit to increase Q and extend bandwidth with inductive peaking. 
 
Thanks DT

The resistor is added because you want low resistance to use the transformer, so you want to know the response of the transformer in its intended way. You want such resistance to reduce the impact of the insertion of the transformer onto the DUT.

JS

[Wolfgang]

What about also *driving* the transformer from a very low impedance !

Its the old rule of thumb: The lower 3dB corner sits where inductive impedance is ca. 4 times the feeding and terminating impedances.
So, when the feeding impedance is just a few ohms and the inductance stays the same, the low frequency corner is improved.

[precaud]

All this talk of termination resistors.
Is this with a resistor intentionally placed across the secondary winding plus the resistance to ground internal to the VNA?

Yes. But the VNA Rterm is usually set to 1MegOhm for these measurements.

That means that we have the external resistor across the secondary coil in parallel with the instrument internal resistor to ground.

Yes, but its not always actually present in use. It depends on what you're using the transformer for, and what the load presented to it is..

It is not so much about adding additional resistance to extend or reduce bandwidth it is about resistance being used to tune or critically dampen the native inductance of the coil.

That works really well if the circuit on the secondary side is fixed, and you can tailor it to the xfmr. I used to do that when designing transformer-coupled mic preamps, optimizing the xfmr RC termination for flattest input impedance and/or freq response. (Deane Jensen was a mentor of mine, and showed us all how to do this back in the 80's.)

But this RC compensation doesn't work when the circuit on the secondary side is an unknown load which you're injecting a signal into. That load is in parallel to whatever Rterm (if any) is across the 2ndary. In that case we want no termination resistor, but we need to deliberately terminate the xfmr with a range of loads it will encounter to see how its characteristics change.

Here's an example, a circuit HP used to recommend for their VNA's to measure output impedance of DC voltage regulators. This is an easy circuit to use with VNAs because of the 1 Ohm current multiplier. Using short and load compensation, I get good results with it down to 10mOhm or better.

The transformer they recommended was the North Hills 0017CC, shown earlier in this thread, a 50 Ohm 1:1 unit rated at 5MHz BW (7.5MHz actual). In this circuit the xfmr is operating at load impedance of (1 Ohm + Zdut) at each freq. At those loads the xfmr bandwidth is varying dramatically with load Z. The useable bandwidth of the 0017CC at 1 Ohm is less than 100kHz (more like 10kHz if you want to measure phase accurately, see the plots a couple posts up for the similar 0016PA). And that's why HP's Z plots for this technique only go to 100kHz, even with the 500MHz analyzer. The xfmr bandwidth into 1 Ohm is the limiting factor.

Hope this helps.

[precaud]

What about also *driving* the transformer from a very low impedance !

Yes, it extends the bandwidth some, but my VNAs all have 50 Ohm outputs, so an external amplifier with direct-coupled output is required to do it.

[Wolfgang]

I did it the primitive way and I added a voltage divider with 50 Ohms impedance for the generator and only a few Ohms to drive the transformer.
Loss is welcome because levels should be very low and it also dampens out resonances.

See here:

https://electronicprojectsforfun.wordpress.com/injection-transformers/

[precaud]

I did it the primitive way and I added a voltage divider with 50 Ohms impedance for the generator and only a few Ohms to drive the transformer.
Loss is welcome because levels should be very low and it also dampens out resonances.

For control loop response, that works, since you need only small injection signal. How high are you trying to measure? Most ones I've seen only look at 100kHz and below.

See here:
https://electronicprojectsforfun.wordpress.com/injection-transformers/

Yes, I've seen that, it is relevant for control-loop-response but not for using a xfmr for output isolation (breaking ground loops for low-level measurements, aka "braid error").

[Wolfgang]

Hi,

my measurement go up to a few 100kHz only. They are mostly for linear power supply for measurement circuits where low noise is an issue (like preamps for noise measurements, ...). Funny enough, even battery powered supplies have noise peaks.

For the "braid error" issue I am working on another approach injecting current and measuring voltage response of a PSU. In these cases I do not need a transformer,
but an active injector, the same ideas like the PicoTest ones. Stability can then be inferred by extracting data from the Nyquist plot of the output impedance.

This is work in progress, but in September I can get my hands on a Keysight E5061B-3L5 (VNA from 5Hz to 3GHz), and then I can properly measure all my homebrew stuff without improvisation.

There are nice appnotes from Keysight how a VNA can be used to measure milliohm impedances in PDN networks. I learned a lot from those.

[Kleinstein]

What about also *driving* the transformer from a very low impedance !

Yes, it extends the bandwidth some, but my VNAs all have 50 Ohm outputs, so an external amplifier with direct-coupled output is required to do it.

If there is no need for high power, one could use a simple shunt resistor at the input side of the transformer. Even with a 50 Ohms output it lowers the impedance, but also the amplitude. If needed (e.g. longer cables) one could add series resistance to get back 50 Ohms impedance for matching.

For the normal use, one would also measure the transfer characteristics of the transformer. So the measurements could be done beyond the 3 dB point. With a not so large load resistance an the secondary side the extra loading by the circuit should not have that much effect. With correction it's no so much die transfer function that matters, but how stable this function is.

Besides the forward transfer function, one might have to care about the coupling capacitance. If the point where the injection transformer is used is high impedance, this is likely the more important limitation. So in a power supply the transformer should go right between the output and the divider - the point between the divider and the FB amplifier would be more like bad choice, as it is high impedance and a few 100 pF can really make a difference.

[precaud]

my measurement go up to a few 100kHz only. They are mostly for linear power supply for measurement circuits where low noise is an issue (like preamps for noise measurements, ...).

Most of these xfmrs will work for you, then.

For the "braid error" issue I am working on another approach injecting current and measuring voltage response of a PSU. In these cases I do not need a transformer,
but an active injector, the same ideas like the PicoTest ones. Stability can then be inferred by extracting data from the Nyquist plot of the output impedance.

This is work in progress, but in September I can get my hands on a Keysight E5061B-3L5 (VNA from 5Hz to 3GHz), and then I can properly measure all my homebrew stuff without improvisation.

I look forward to seeing your results!

There are nice appnotes from Keysight how a VNA can be used to measure milliohm impedances in PDN networks. I learned a lot from those.

Agreed. The "shunt-thru" technique can give good results into higher freqs.

[DualTriode]

Hello,

Now that the conversation regarding the termination resistor is near complete take a look at page 8 figure 4.1 of the Omicron Lab injection transformer manual. The recommended injection resistor value is between 1R and 10R.

https://www.omicron-lab.com/fileadmin/assets/Bode_100/Accessories/B-WIT_100/B-WIT-B-LFT-User-Manual-V1.1.pdf

Now looking at injection transformer phase measurements:

The transformer phase plot shows the phase difference between the transformer primary and secondary. Some folks here say that the injection transformer is no longer useful when phase exceeds some value plucked from the chart. This needs a closer look.

If you look at the Signature Bode 100 gain vs Phase Margin chart, the point where gain is equal to 0dB the PM gives a very good indication of power supply stability. Doesn’t the injection transformer phase effect the PM of the power supply? The short answer is no. 

Thanks DT

[precaud]

Now that the conversation regarding the termination resistor is near complete take a look at page 8 figure 4.1 of the Omicron Lab injection transformer manual. The recommended injection resistor value is between 1R and 10R.

More precisely, it can be used with R's from 1R to 10R, with 10R recommended.

If you look at the Signature Bode 100 gain vs Phase Margin chart, the point where gain is equal to 0dB the PM gives a very good indication of power supply stability. Doesn’t the injection transformer phase effect the PM of the power supply? The short answer is no. 

Well of course not. That isn't the question. The question has to do with the accuracy of the measurement.

[T3sl4co1l]

If you want infinite bandwidth from such a transformer, and don't mind a flat 6dB insertion loss: split the transmission line in the middle, and add terminations here.  You need an R+C across the (now open) ends of these windings, where:
R = Zo
C >= 2.5 * k * (377Ω) * t_half / Zo^2

k = effective dielectric constant (note this is less than the material k itself, when the dielectric is mixed, e.g., microstrip, twisted pair, foam, etc.)
t_half = transmission line electrical length (units of time)

This terminates the half-windings, so the transformer looks like a pair of Zo resistors from P1 to S1 and from P2 to S2.  Obviously, this breaks transformer action.

One might rightfully note this isn't a transformer at all anymore, and the circuit can be greatly simplified by removing the transformer altogether; now you have coupling capacitors (the termination resistors are completely optional, unless you liked the insertion loss), and the CMRR is still just as bad, although it's bad at low frequencies too (oh well? ).

To restore transformer action, connect across the cut ends (primary middle to primary middle, secondary middle to secondary middle) with inductors, of similar value (i.e., L ~ Zo * t_half).  Now you get asymptotically zero insertion loss at low frequencies, and at high frequencies, a 6dB shelf instead of the dips and peaks.

Aside of all of this, CMRR is asymptotically bad in the HF limit.  To address this, place a CMC on either side of the transformer, as much (equivalent) inductance as you can afford.  Add damping R+C across the transformer (P1 to S1, and P2 to S2) to control resonances, if necessary (these will be a large impedance, so will have little impact on HF response, despite their location).  The CMC inductance will resonate with the transformer isolation capacitance (and coupling/termination capacitance C, if used, as above), so this prevents that resonance from getting too peaky.  An isolation impedance in the low kOhms is very reasonable to achieve this way, and while that doesn't sound terrifically good, understand that we're talking with respect to radio frequencies here, where a thin wire in semi-free space is unlikely to reach half a kohm Zo.  This is doing as well as you can, given the limitations of real electromagnetism and not just some inductors and capacitors in a SPICE model.

Tim

[Wolfgang]

Now looking at injection transformer phase measurements:

The transformer phase plot shows the phase difference between the transformer primary and secondary. Some folks here say that the injection transformer is no longer useful when phase exceeds some value plucked from the chart. This needs a closer look.

If you look at the Signature Bode 100 gain vs Phase Margin chart, the point where gain is equal to 0dB the PM gives a very good indication of power supply stability. Doesn’t the injection transformer phase effect the PM of the power supply? The short answer is no. 

Thanks DT

What you write reminds me of a hefty discussion following a Texas Instrument Application note where they used a garden variety line transformer as an injection transformer and claimed its characteristics are irrelevant because it was "outside the loop".

It is OK to claim that (provided the secondary resistance is small enough) it does not alter the characteristics of the loop, but what it does alter is the amplitude and phase you measure at the loop output. You could calibrate this out using a VNA, but it is definitely a precision issue. When you look at the characteristics of the magnetic materials used for such transformers, the range of mu for a given frequency is 2 to 1. You can expect an coresponding low precision range for band corners. If your measurements are to be trusted, the phase shift induced by your transformer must be a) small and b) well known.

[precaud]

Here's a plot I made last year of the 10Hz - 10MHz magnitude response including some of the transformers mentioned in this thread, with 50 Ohm generator output, and xfmr terminated with 1 Ohm.

Curves from top down:
The dark blue is a toroid Tokin 10mH CM choke.
Magenta is the Jensen Iso-Max mentioned here several times. It is coax wound on a standard EI core, no doubt with carefully-chosen core material...
Green is the North Hills 0016 PA
Yellow is the North Hills 0017 CC
Red is a North Hills 1307 LB 75 Ohm unbal to 110 balanced transformer, used as a makeshift step-down xfmr, with one side of the 2ndary terminated in 50 Ohms and the output taken from the other side. Even this mild stepdown is sufficient to extend phase linearty vs Z to just beyond 100kHz, and so this is the xfmr I use with the circuit shown in post #102.
Cyan is a Ridley 15MHz Injection Transformer, a 10:1 step-down design, IIRC.

I would expect that the B-WIT 100 and the home-rolled version would behave pretty much the same, rolling off somewhere around 100kHz.

It would be interesting to make a 2:1 or 3:1 version to extend the 1 Ohm bandwidth. Hmmm...

EDIT : I learned yesterday that the Ridley is a 3:1 step-down winding ratio.
This makes it even more compelling to do a home-roll 2:1 or 3:1 version...

[capt bullshot]

Hey guys and girls, I've just found out the secret sauce that North Hills adds to their wide bandwidth transformers

.

.

.

.

.

Yes, its "Bauschaum" (polyurethane foam)


This is a NH 12369 (obviously) video isolation transformer. Anyone's got a datasheet? I didn't find one. Will do some measurements later.

[T3sl4co1l]

Interesting that there's silver mica caps in there.

Pot cores are quite good, having more A_L than a toroid of the same size -- which means less wire length and more transformer bandwidth.  (Too bad amorphous isn't available in shapes!)

Tim

[capt bullshot]

transformer pr0n









The unit has 4 channels, three of them intended for video and one sync.
The sync channel has a small cap (labelled "100") across the input BNC.
The video channels have the silver mica ("471") connected from input (presumably BNC shell) to output (BNC shell).
The transformers are wound using twisted enamelled wire, the video channels have thicker wire than the sync channel.

[capt bullshot]

Some results:

The video channel has a magnetizing inductance of 112mH and the input to output coupling capacitance is 1.13nF (measured quite similar on all three channels).
For the sync channel, the results were 977mH and 1.9nF.

The frequency response (-3dB) for the video channels is 50Hz ... 78MHz, but the 45° phase shift response is from 50Hz to 14MHz - some tricks happening here - Note the bumps in the frequency response above 1MHz. This response looks the same for all three video channels. Termination was 50 Ohm for all measurements.



The sync channel shows a clean response from the low end of the HP3577A (5Hz) to 7.5MHz.



Estimating the wire length with the TDR shows a pretty clean impedance over length, somewhat varying around the expected 75 Ohm. Maybe 1.8m for the video transformers and around 6.8m for the sync. Assuming the real length in this ballpark, the high frequency response is way much better than the home rolled transformer using the nanocrystalline / amorphous core.




Pretty good stuff this is!

[precaud]

The frequency response (-3dB) for the video channels is 50Hz ... 78MHz, but the 45° phase shift response is from 50Hz to 14MHz - some tricks happening here - Note the bumps in the frequency response above 1MHz. This response looks the same for all three video channels. Termination was 50 Ohm for all measurements.

I'm guessing you're using the 50 Ohms term in the 3577A. If you terminate the xfmr right at its output and set the analyzer to 1MegOhms, the mag/phase correlation at the high end should make more sense.

Pretty good stuff this is!

Agreed!

[capt bullshot]

I'm guessing you're using the 50 Ohms term in the 3577A. If you terminate the xfmr right at its output and set the analyzer to 1MegOhms, the mag/phase correlation at the high end should make more sense.

Yes, I'm using the internal termination. Shifting the termination towards the transformers output and setting the input to 1M doesn't make it any better, in contrary one can see even more bumps in the frequency response. I've used a rather short coax cable to connect the terminated transformer output to the 3577A. Maybe one would get better or at least other results using a compensated 10:1 oscilloscope probe.

In general, I'd prefer to see the transformers response when fed and terminated by its nominal impedance (75 Ohm for video stuff). I don't have the minimum loss pads to match to 3577A's 50 Ohm ports, so for now I'll stay with the 50 Ohm measurements.

And for the records, I'm making the measurement from Output to R input, in contrary to using a power divider and A/R as I prefer to see the system response over the transformers response fed from a virtual zero ohm source. Using the power divider and A/R measurement gives different results with notably extended low frequency response.

[DualTriode]

Now looking at injection transformer phase measurements:

The transformer phase plot shows the phase difference between the transformer primary and secondary. Some folks here say that the injection transformer is no longer useful when phase exceeds some value plucked from the chart. This needs a closer look.

If you look at the Signature Bode 100 gain vs Phase Margin chart, the point where gain is equal to 0dB the PM gives a very good indication of power supply stability. Doesn’t the injection transformer phase effect the PM of the power supply? The short answer is no. 

Thanks DT

What you write reminds me of a hefty discussion following a Texas Instrument Application note where they used a garden variety line transformer as an injection transformer and claimed its characteristics are irrelevant because it was "outside the loop".

It is OK to claim that (provided the secondary resistance is small enough) it does not alter the characteristics of the loop, but what it does alter is the amplitude and phase you measure at the loop output. You could calibrate this out using a VNA, but it is definitely a precision issue. When you look at the characteristics of the magnetic materials used for such transformers, the range of mu for a given frequency is 2 to 1. You can expect an coresponding low precision range for band corners. If your measurements are to be trusted, the phase shift induced by your transformer must be a) small and b) well known.

Hello,

I don’t know anything about a garden variety (straw man) line transformer.

I do agree there may be a falloff in amplitude. Both the Keysight and Bode 100 have built in provisions to adjust the test amplitude to within the small signal requirements of the error amplifier control loop. Even the best specialized injection transformers may sometimes need a little help.

I have seen no evidence that the change in phase in the injection transformer is evident in the phase measurement between the input of the error amplifier loop and the power supply output.

Any transformer, even the best specialized injection transformers may get a little wobbly if pushed into the anti-resonance/ resonance frequencies.

Thanks DT

[precaud]

In general, I'd prefer to see the transformers response when fed and terminated by its nominal impedance (75 Ohm for video stuff). I don't have the minimum loss pads to match to 3577A's 50 Ohm ports, so for now I'll stay with the 50 Ohm measurements.

When it comes to xfmr-coupling, I'm a fan of "whatever works"   

And for the records, I'm making the measurement from Output to R input, in contrary to using a power divider and A/R as I prefer to see the system response over the transformers response fed from a virtual zero ohm source.

I'm not using a power divider, either. What is your "virtual zero ohm" source ?

Using the power divider and A/R measurement gives different results with notably extended low frequency response.

To me, characterizing a xfmr is set up like an FRA measurement; A/R, source Z same as it will be used, no power divider, high-Z inputs, and xfmr terminated in whatever the desired load is.

[rx8pilot]

To me, characterizing a xfmr is set up like an FRA measurement; A/R, source Z same as it will be used, no power divider, high-Z inputs, and xfmr terminated in whatever the desired load is.

How do you calibrate/normalize this setup?

[capt bullshot]

I'm not using a power divider, either. What is your "virtual zero ohm" source ?

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.

In practical appliation, e.g. for the said video stuff, one cannot take the loading effect out of the system, so I normalize R (transformer replaced by a "through") and then put the transformer in place to see the system response.

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.

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.

[capt bullshot]

Couldn't leave that stuff alone ...

So I bodged a pair of minimum loss pads (50 Ohm <-> 75 Ohm) together and measured the video transformer again:
-3dB from 60Hz to 100MHz, +/-45° phase at 60Hz / 13Mhz, similar to my 50 Ohm measurement with the low frequency response shifted up (as one would expect from the increased source / sink impedance). Still some smallish bumps in the amplitude response above 1MHz.

Now, after some thinking, I replaced the "through" calibration piece (a male-to-male BNC adaptor) with a piece of 75Ohm coaxial cable of similar length to the guessed wire length within the transformer - somewhere in the 1.8m range.
After normalizing to this "through reference", the result is (guess what):
same as above for -3dB, but -45° phase at 45MHz.

So the internal wire length of the transformer causes some delay that leads to the misinterpretation -3dB implausible to phase.

BTW:
TDR'ing the transformer through the minimum loss pads shows a quite flat line - so these transformers clearly are made for 75 Ohm system impedance and have an impressive bandwidth then.

BTW 2:
The 3577A's "Length" setting achieves the same result as normalizing with the "through cable". Just set it to the negative value of your additionally introduced cable length to compensate for this.

BTW 3:
Pushing Leo Bodnar's fast pulses through the transformer gives a risetime of 4ns (vs. 400ps w/o the transformer, measuring the scope and matching pads). The transformer also adds some ringing to the square wave.