Help to make a true floating probe?

[FriedMule]

Hi I have now for over a year tried to find out on how to make a floating probe, but without getting any wiser.
It shall not be high Voltage, differential nor anything other than a 100MHz probe that is genuine floating.
I have looked at transformers, chips, optically fibers and other solution, but I am in way over my head.
How would you make the isolating part and with what type of component?

[BrianHG]

https://www.eevblog.com/forum/beginners/floating-probe!-for-$2-50/

[Berni]

https://www.eevblog.com/forum/beginners/floating-probe!-for-$2-50/

This is pretty neat. Too bad the thread ended there, sounds like one could get some impressive bandwidth out of it with perhaps a more careful component selection and tweaking.

[T3sl4co1l]

Not so hard to do, it's just that you need quite intense sources (preferably lasers I guess) to get useful signal currents out of photodiodes, which have to be small enough (low capacitance) to get useful signal bandwidth.  And then the noise floor (equivalently, dynamic range) in that bandwidth isn't great, plus you can't wrap negative feedback around it to account for nonlinearities.  (Which at least makes photodiodes better than phototransistors, which have hFE nonlinearity).

The really nice ones use an isolated ADC, which is of course a challenge if it's high speed (10s Gbps).

Tim

[FriedMule]

I have done a lot of reading upon all your answers and have now decided to try on a differential probe thingy instead.
The design is mostly stolen from what I could find on the internet's images. Since I have no illusions of my own capabilities do I in no way think or hope to get  no ware near the almost 2 GHz that the original design was capable of. And I do not need so high a speed, but still funny to reach as high a frequency as possible:-)

The U1 chip is a ADA7927-1 https://www.mouser.dk/datasheet/2/609/ADA4927-1_ADA4927-2-878322.pdf
Every signal trace is 11.8 mil (0,3 mm) and the power traces is 19.8 (0,5 mm).
On each end are a BNC female plug. The board is 4 layers (Signal, ground, +5V, 5V). The ground layer do I maybe only let go under the signal part, while the other 2 layers will fill the whole board. Below the slot in the PCB do I imagine putting a part of the power supply circuit, while the transformer, the bridge and all that will be on a separate board.

To the right of the PCB have I put the BNC-out but can't find out if it's okay to run the signal GND to the GNDD layer and the main signal through a trace, it does make the impedance totally different.

Any comment, tips tricks or answers on my questions? :-)

[David Hess]

Check out how the Tektronix A6902 20MHz isolation amplifier was designed.  It uses optical coupling up to 2.5 kHz and then transformer coupling above that.

http://w140.com/tekwiki/wiki/A6902

[Berni]

Yeah with this sort of design you are not getting anywhere near 2GHz because of conflicting design parameters.

To get a lot of common mode input range out of it you need the input resistors to be large in value. This is what actually throws off the impedance as the input side is going to be far from 50 Ohm. Also you typically don't want probe inputs to be as low impedance as that anyway. This can make it hard to get a flat frequency response out of it, especially when you connect a coax cable to the input that will start to resonate at certain frequencies due to not having its end terminated properly.

Also these kind of diff amp designs are heavily dependent on the tolerances of the passives in its feedback network to get a good common mode rejection out of them.  When the inverting and non inverting side are slightly different a common mode signal turns into a differential one and gets amplified. At higher frequencies the parasitic inductance and capacitance start to matter too, so to get good common mode rejection out of them they typically have some trim pots and trim caps around them to fine tune it for the best response.

That Tek way of doing it sounds like a pretty good idea, but yeah might be a bit complex for DIY

[BravoV]

Check out how the Tektronix A6902 20MHz isolation amplifier was designed.  It uses optical coupling up to 2.5 kHz and then transformer coupling above that.

http://w140.com/tekwiki/wiki/A6902

About the A6902, the unit it self, not including the probes, is much-much bigger than today's common sized DSOs.

My DS1104Z (identical size to popular DS1054Z) stands above the gigantic Tek A6902, yeah, its heavier than the scope it self.