DIY active injector for bode plotting (alternative to transformer injector)

[TopQuark]

Hi, much effort have been spent on building better and better isolation transformers for use in bode plotting. e.g.
https://www.eevblog.com/forum/testgear/diy-transformer-for-use-with-bode-plots/

I was using my isolation transformer to measure open loop response yesterday, and was quite unhappy with the noisy low frequency response measurements, mainly due to insufficient drive strength of the setup at low frequency and the high open loop gain of the DUT. My isolation transformer also only has a bandwidth of 10MHz or so, and can't really fully utilize the 100MHz frequency range of the bode plotting function.

I thought to myself, there has got to be something better. So I went on to picotest's website and noticed they sell a so called solid state injector, which uses active components to drive the DUT.
https://www.picotest.com/products_J2110A.html
While I liked the concept, I didn't like how they implemented the solution. The injector doesn't seem to be isolated w.r.t. to the signal input or ground, and there's a maximum usable input voltage range of +-12v. The 45MHz bandwidth is not terrible, but not great either.

I wanted something that breaks up ground loops, and was willing to give up frequency response below 10Hz. I want ample drive strength that is not dependant on transformer magnetics or source drive impedance. Being able to work up to 100MHz would be great.

After a bit of experimentation and ideation, I came up with this solution. BUF634A powered by a couple of 9v batteries, control input is capacitively coupled to the BUF634A input, that's it. This gives us theoretically 210MHz of BW, 250mA of drive strength, and floating operation.  Building this is so much easier than trying to eek out the last bit of performance out of a injection transformer. 

There's a bit of gain peaking around 70MHz or so, I blame construction technique and termination impedance mismatch for that, I'll fix that later. Besides that, the frequency response is really flat, and is very usable from 10Hz to 100MHz, much better than my injection transformer.

Testing the active injector in a real circuit reveals the strength of the active injector, high drive strength especially at lower frequencies. No longer limited by core saturation, with its low impedance 250mA output and high impedance input, the active injector can really shove a signal to a feedback loop proper to test the open loop response of systems at low frequencies, where loop gain is high. Measuring open loop response of a buck regulator, the active injector gives beautiful, noise free, text book loop response measurements with only 0.2Vp-p of stimulus, while the transformer based measurement is a bit all over the place below a few hundred Hz, even with a few times higher stimulus amplitude. The active injector really brings out the best of the scope's bode plot function, I no longer have to willfully ignore the lower frequency range to make my graphs look neater. 

I think I'll still keep my injection transformer for some occasions, but for measuring loop response of closed loop systems, IMO the active injector is miles ahead of any transformer. If you like bode plotting, I think you should definitely build an active injector of some sort.

[nctnico]

Looks neat. But how about using a transformer to get the signal into the buffer? The way the circuit is now has quite a large capacitance between input and output which could mess things up in some cases. Another option would be to use some kind of opto-isolator. Heck, it might not even matter to use a square wave at low frequencies.

[TopQuark]

Looks neat. But how about using a transformer to get the signal into the buffer? The way the circuit is now has quite a large capacitance between input and output which could mess things up in some cases. Another option would be to use some kind of opto-isolator. Heck, it might not even matter to use a square wave at low frequencies.

Then we are back to pulling all our hair out trying to build an isolation transformer with flat response from 10Hz to 100MHz, a problem that even gurus on the forums or companies can't seem to solve. 

Most of the time the DUT won't be swinging all over the place w.r.t to the input anyways. (if it does, you have a bigger problem to worry about  )