Cheap differential probe - AD8130

[borys]

Hello to all Members!!
I am DIY stuff maker and need to build a simple differential probe. I am going to use it in some SMPS measurements, I do not need very high bandwidth (couple MHz will do the job  ). I have made a first prototype and it seems to work well, but there is a few things that I am not sure about.
The probe is based on AD8130 amplifier, which works as a unity gain differential amplifier (page 33 on datasheet).
http://www.analog.com/media/en/technical-documentation/data-sheets/AD8129_8130.pdf
I just have added a few resistors + compensation caps at the inputs and x10 x100 switch.

1.What would be the best way to compensate the offset which is present on the measurements ? (putt it at the input or shift a bit OP-amp reference )
2.I have powered the circuit from two small 12V batteries to make the things easier, do I really have to stabilize the supply voltage to the op-amp (let me say at +/-9V) ?
3. The probe GND must be allways connected to the DUT GND ?
4. I have used a twisted pair cables a my probe, I do not know if they can be a single/short wires coming out the probe box or I have to use a proper screened cable (adds a bit more capactance) ?

Sorry if my questions are bit stupid, learning process requires it..

Bellow some scope shoots, the test signal is from ebay signal generator (do not laugh ) 1MHz square wave, yellow line is my reference (150MHz x10 rigol probe) and purple one is from my probe.

Thanks
Peter

[lincoln]

>4. I have used a twisted pair cables a my probe, I do not know if they can be a single/short wires coming out the probe box or I have to use a proper screened cable (adds a bit more capactance) ?

this should help: https://www.eevblog.com/forum/projects/lo-z-probe/

[cat87]

First off, nice job with the probe, the signal looks OK for this kind of set-up. Really Impressed. Looking forward to seeing this thing in action 

1.Maybe play with the op-amp gain and/or with the input attenuation resistor network.
2.Well, basically, yes, it's a must. Your signal on the output will depend , among other things, on the slew rate of the  op-amp. No problems with the one you chose, but once the batteries will get drained, it will decrease in performance. Also, it will have an effect on the amplitude of the output signal. So, maybe stick an LDO on each rail and get things as stable as possible. Another way would be to get a  buck-boost  power supply, powered from the batteries, to get things stable even when the batteries are really low. but that requires some shielding of the PSU portion from the rest of the board.
Also, have a look at the graph where it says "power supply rejection versus frequency", that will give you some idea of how the well the op-amp rejects noise picked up from the power supply, because that will also influence you final waveform output.
3.Yes and no. Yes, you always have to connect the "+" and "-" of the probe to the DUT. No, the GND of the probe is never connected to the DUT. That goes to the oscilloscope GND. Because it's a differential probe, you measure a signal differentially i.e. wherever you connect the "-", that's going to be you reference (you can hook up the "-" to the GND of the DUT or to any other part of the circuit, given the probe can withstand that much  voltage difference). If you connect the GND of the Diff Probe to the GND of the DUT, you defeat the whole purpose of a diff. probe.
4.It depends. If you'll be measuring DC level voltages (0Hz to about 50KHz), then the wires are perfectly OK. But if it's AC measurements you;re interested in, asyou have showm in your photo i.e. the 1MHz signal, then yes, you need some proper screened cable, in order to limit the amount of noise pick-up and also get a controlled input impedance on the input of your probe.

Don't worry, the questions are not stupid, it's how people learn stuff

As a side note, the op-amp you've chosen can handle really high bandwidth, but you might want to look for other methods of building you Diff Probe, because an op-amp in differential amplifier mode  has severe limitations as far as CMRR is concerned. Also the bandwidth will be much less then what those fancy charts show in the data sheet.

Also, it's  always better to buffer you input signal (if you want to have a bandwidth above 1MHz) with some really high impedance, something like a JFET differential stage.  The input impedance of  resistor network and the op-amp alone will not cut it.

[Mechatrommer]

2.I have powered the circuit from two small 12V batteries to make the things easier, do I really have to stabilize the supply voltage to the op-amp (let me say at +/-9V) ?

if its from battery... no.

3. The probe GND must be allways connected to the DUT GND ?

no. every reasonably priced KV range diff probe that i know dont do this.

4. I have used a twisted pair cables a my probe, I do not know if they can be a single/short wires coming out the probe box or I have to use a proper screened cable (adds a bit more capactance) ?

every reasonably priced KV range diff probe that i know use only single wire cabling. pure copper or resistive type i'm not sure. but i think there is reason not to use coupled cable pair. and imho its better to follow the trend of the more experienced one.

Sorry if my questions are bit stupid, learning process requires it..

no thats not stupid. i guess those are questions that coming out of every diff probe diyers that've done it.

Bellow some scope shoots, the test signal is from ebay signal generator (do not laugh ) 1MHz square wave, yellow line is my reference (150MHz x10 rigol probe) and purple one is from my probe.

the real fun is to see signal of common mode rejection, ie when both probe connected to the same signal.. doing sort of single ended measurement like that (one probe to the ground) is not really fun

[borys]

Thanks for advice !!!
It really helps me. After today I know that probe cables (in my simple setup) can not be twisted/screaned with GND.
With higher amplitude fast transients there is massive ringing at the square voltage curves (primary in SMPS). After I have removed twisted GND screen from probe cables, and left them 25cm each the ringing is gone.

Question no 5. Adding the small value resistor at the probe tip (let me say 50R - 1kR) should dump a bit the resonances coming from the probe cable capactance/inductance and rest of the circuit ??


Big thanks.
Regards

[Mechatrommer]

Question no 5. Adding the small value resistor at the probe tip (let me say 50R - 1kR) should dump a bit the resonances coming from the probe cable capactance/inductance and rest of the circuit ??

highly probably yes (apparently), but you need to spec it yourself...

[borys]

Thanks for advice's!!

I have another few questions.
6. Would it be advisable to putt a ground plane around the low voltage part of the probe (op-amp, compensation circuit, supply) ?
7. I do not have the CMRR regulation pot at the resistor divider, is it really needed ?

The pcb's of the probes I have seen didn't had any gound planes and allways there was CMRR pot present just at the front of the probe (by the resistor network divider).
Is there any way that I can do some HOME MADE measurements that proof that this pot is really neaded ??

Thanks
Peter

[Mechatrommer]

6. Would it be advisable to putt a ground plane around the low voltage part of the probe (op-amp, compensation circuit, supply) ?

follow the opamp datasheet on recommended pcb layout..

7. I do not have the CMRR regulation pot at the resistor divider, is it really needed ?

definitely yes...

Is there any way that I can do some HOME MADE measurements that proof that this pot is really neaded ??

as i said.... doing cmr signal probe will be fun, ie both probe to the signal... the output should be zero... you will get nonzero trust me proving the trimpot is necessary..

[cat87]

I would like to contribute my 2 cents to this:
6. Yes, you can put a ground plane if you'd like, BUT, do NOT put  one under the op-amp. Leave a small patch under it that has no copper. Putting a ground plane under your op-amp screws up your CMRR and couples in  noise at high frequency (>1MHz). Basically it's a no-no.
7.Yes, definitely. On most designs, the CMRR is mostly dictated by how well the PCB is laid out

[Richard Head]

Putting a ground plane under your op-amp screws up your CMRR and couples in  noise at high frequency (>1MHz). Basically it's a no-no.

That's interesting. Please elaborate.

[Marco]

1.What would be the best way to compensate the offset which is present on the measurements ? (putt it at the input or shift a bit OP-amp reference )

I think you could use a trimmable opamp buffered voltage near 0, with some good decoupling, instead of the ground for the feedback divider of the AD8130.

[borys]

Thanks again!!

I will prepare a new board with the CMRR pot and do some more tests. After a few resistors are tweaked the probe seems to work a bit better.
I did some tests on real life animal (SMPS current protection) and all stuff survived.

The pcb of the probe is single layer and I was going to try the ground plane only on the top layer, but when I looked at the pcb's of some genuine probes there is no ground plane at all --> so I will not use it.

I think the bottom layer ground plane would work as a capacitor (two conductors splitted with dielectric) and it will provide some coupling in between the tracks at the high frequences).  (??)

Regards Peter

[borys]

Bellow a next prototype boards proposition.
Now which one to go for ??

[cat87]

Putting a ground plane under your op-amp screws up your CMRR and couples in  noise at high frequency (>1MHz). Basically it's a no-no.

That's interesting. Please elaborate.

Well, by having a ground plane under the op-amp, the input capacitance on the input will increase, because of capacitive coupling of the input pins to the ground plane.  And, for high frequencies, high input capacitance is something you want to avoid. Also, the power pins will also be capacitively coupled to the input pins, so there will be a lot of noise fed  into the input,  from the output switching of the op-amp itself.

Therefore:

Bellow a next prototype boards proposition.
Now which one to go for ??

The one without the ground plane, in my opinion. Or, just remove the copper underneath the op-amp

[borys]

I have made two more boards. The one with no ground plane did not work well, I was not able to compensate it fully. With flat top/bottom part of the square wave there was always some overshoot present, compensation was not working OK, square was distorted a bit.

Than I have made a board with ground plane, some of the capacitance have increased so I had to increase compensating caps a bit. After probe has been compensated it is working IMHO OK. I did some tests on SMPS with good results. Bellow some photos.

Thanks to all for help !!