My first project: Low-cost differential probe

[cyr]

Using a 10x passive probe won't help with common-mode range if you do it like this, I would change the design so that it has two probe inputs instead (something like the Tek AM502)

http://www.barrytech.com/tektronix/tektm500/tekam502.html

However, I'm not sure how well a pair of cheap passive probes will match (probably not very well, destroying your CMRR).

Still it would be an improvement on using the math function on the scope, better usable range and only one scope channel used...

[poorchava]

I think you still need to use insulation 1:1 mains transformer. In case your scope is powered from different phase than DUT, you can have like 400V between scope ground and either input of the probe. It's even worse if you need 2 or more channels at the same time: common mode may be a problem. Not because attenuation, but rather plain semiconductor damage because of high voltage.

[Mechatrommer]

In case your scope is powered from different phase than DUT, you can have like 400V between scope ground and either input of the probe. It's even worse if you need 2 or more channels at the same time: common mode may be a problem. Not because attenuation, but rather plain semiconductor damage because of high voltage.

are you saying 2 very far apart isolated system? if yes then lets assume the worst, like what? 1K, 5K, 10K, 100KV ground differential?. no differential probe can survive that AFAIK if its beyond their rated limit, arching is another issue.its beyond the realm of engineering we are talking something similar to ESD. we have to limit our scope of discussion, ie assuming we can bring both ground potential to equal (or within limit) then all the specs (CM) will become make sense.

- So this thing can not be used on probing high voltage like mains powered AC to DC power supply circuit ?

he is still lacking for the attenuator or coupler there.

- How good is this probe's performance compared to using the common two probes trick with both ground clips connected to probe differential signal ?

by using external diff probe:
1) you save one channel of your dso, ie you can use 2 diff probes to measure 2 separate diff signals using your 2 channel dso. and also using math function of your dso will reduce its performance (ds1052e tested)
2) theoritically... you can tune the diff probe's circuit to get better CMRR compared to 2 passive probes trick. but... thats the "most fun" part if you want to beat the aftermarket products. opamp's CMRR spec is just the limiting factor of what you can achieved (i'm yet to get any closer). once you put your attenuator, you are on your own. in my case, moving my finger near and around the attenuator changes the CMRR, magic! but i have to admit, its my limitation in knowledge and experience and i'm restricting myself to jellybeans parts (for a reason).

[BravoV]

2) theoritically... you can tune the diff probe's circuit to get better CMRR compared to 2 passive probes trick.

Any idea what is the ball park number (quantitatively) for best case scenario ?

Honestly I'm actually interested to build one too, its just I need to know what should I expecting, worst and best case scenarios to judge if its worth it.

PS : I do have a 1000 VA isolation transformer.

[hammil]

- So this thing can not be used on probing high voltage like mains powered AC to DC power supply circuit ?
- How good is this probe's performance compared to using the common two probes trick with both ground clips connected to probe differential signal ?

1. Not without attenuation, no.
2. I don't think that will always work, since the scope still measures the voltage referenced to mains earth. I'll look into it further, though. However, either way it still uses two channels, whereas my solution should only use the one.

FWIW Hammil, I would deal with the attenuation inside the case and PSU outside the case (if you've run out of space).

Conducted noise might only be 16uV, radiated noise (L1 & tracks) is another matter altogether.

That might be an option, yeah. I've ordered a few boards from oshpark, so I should be able to see if there's a visible effect from noise.

Using a 10x passive probe won't help with common-mode range if you do it like this, I would change the design so that it has two probe inputs instead (something like the Tek AM502)

http://www.barrytech.com/tektronix/tektm500/tekam502.html

However, I'm not sure how well a pair of cheap passive probes will match (probably not very well, destroying your CMRR).

Still it would be an improvement on using the math function on the scope, better usable range and only one scope channel used...

Thanks for the thoughts. What I'll probably end up doing is having a (selectable?) x10 or x100 attenuator inside the case, and advise that the probe be used in x1 mode.

Out of curiosity, though, if I did have two probe inputs, where would I connect them, and how would this improve the common-mode range?

I think you still need to use insulation 1:1 mains transformer. In case your scope is powered from different phase than DUT, you can have like 400V between scope ground and either input of the probe. It's even worse if you need 2 or more channels at the same time: common mode may be a problem. Not because attenuation, but rather plain semiconductor damage because of high voltage.

I can see why that would be a problem, however using a transformer will add a high-pass filter unless I used some switching trickery, which would vastly increase the complexity of the design. Thank you for the information, though.


One solution for the common-mode problem, I believe, would be to simply have a (selectable, of course) attenuator referenced to mains earth itself.... Other than that, I can't seem to find any simple fixes. I may just have to accept this as a design limitation.

[Mechatrommer]

Any idea what is the ball park number (quantitatively) for best case scenario ?

depending on your luck, if you are hobbiest. somewhere around 80dB at low freq, and/or 60dB at higher freq if you are an engineer.

[Marco]

In some expensive scopes that have isolated channels (Tek used to do that) the channels are isolated on digital level

Won't you be hanging a huge amount of stray capacitance on the DUT in that case?

[codeboy2k]

FWIW Hammil, I would deal with the attenuation inside the case and PSU outside the case (if you've run out of space).

Conducted noise might only be 16uV, radiated noise (L1 & tracks) is another matter altogether.

Yeah I was concerned about that too, especially saw L1 so close to the output BNC.  I would like to see the entire power supply capped over and shielded on the PCB; if not, at least shield L1.

[hammil]

Yeah I was concerned about that too, especially saw L1 so close to the output BNC.  I would like to see the entire power supply capped over and shielded on the PCB; if not, at least shield L1.

It was either that, or have it closer to the amplifier, which would be much more sensitive. Looking at it now.. I could have L1 on the other side of the PCB, which would help, I'd imagine.

How easy/difficult is it to make a basic metal shield can? I'm really not too 'handy', unfortunately.

On an unrelated note, I have updated the schematic to include a x10 attenuator that I believe attenuates both the differential and the common-mode components of the input.

In fact, it would only require a few extra resistors and jumpers to make a selectable x1/x10/x100 attenuator there.

[fcb]

Hi Hammil,

I don't think I would bother with x1 option. The second design should be 'worry-free' in operation, very low risk of cooking anything.  If you need x1 then I would put a second opamp on the output that has a switchable gain.  If you wanted to make it really mean - make the front end switchable AND add a switched-gain opamp in the back.  You can at least choose opamps based on specific places in the circuit.

I would still get rid of the switcher in the unit - if you want to keep it, add some inductance in the power lines and island the switcher ground. Again though, I would definetley put the PSU outside the 'head' unit - there's another possible benefit, if you fitted something like a 9 pin D on the side of the unit (or something smaller - perhaps a six pin Hirose HR10A), you could then have a couple of control lines from the external supply - these could drive two small relays, which make up an attentuator & gain control.  If your worried about current drain - there's some neat latching relays available.

[Mechatrommer]

(for no particular reason) let me highlight this, since this is the most interesting part for me. have fun doing that thats where you will learn about matching circuit and parasitics.

[fcb]

You might find this Bob Pease show informative:

[Marco]

somewhere around 80dB at low freq, and/or 60dB at higher freq if you are an engineer.

Tektronix's universal differential probes don't even get that ...

http://www.tek.com/datasheet/differential-probe-differential-probes

[alm]

10x probes are typically 1% accurate in attenuation (at least the expensive ones that actually come with a datasheet). This limits your CMRR to 40 dB max (at DC). 1x probes are better, but limit the BW to a few MHz. Tektronix made some special 10x probes with adjustable attenuation (P6055/P6135?). These had a variable resistor in parallel with the scope input to vary the attenuation ratio. I guess you could do something similar by designing the amplifier for >> 1 Mohm input impedance and put a 990k + 25k trimmer in parallel.

60 dB requires better than 0.1 % matching between the two channels. Are you going to match the caps in the input attenuator own to better than 0.1 %? What about the voltage coefficient (the capacitance of a cap changes with applied voltage). I believe that 'cheap' differential probes (eg. the ~$300 ones) have their CMRR limited by the voltage coefficient of the attenuators.

[hammil]

I don't think I would bother with x1 option. The second design should be 'worry-free' in operation, very low risk of cooking anything.  If you need x1 then I would put a second opamp on the output that has a switchable gain.  If you wanted to make it really mean - make the front end switchable AND add a switched-gain opamp in the back.  You can at least choose opamps based on specific places in the circuit.

I would still get rid of the switcher in the unit - if you want to keep it, add some inductance in the power lines and island the switcher ground. Again though, I would definetley put the PSU outside the 'head' unit - there's another possible benefit, if you fitted something like a 9 pin D on the side of the unit (or something smaller - perhaps a six pin Hirose HR10A), you could then have a couple of control lines from the external supply - these could drive two small relays, which make up an attentuator & gain control.  If your worried about current drain - there's some neat latching relays available.

That's a pretty good idea. Especially since I'm considering making different versions of the board. ( A JFET-input and a transformer isolated one, if you were curious ). The power supply/control board would almost certainly be identical between the versions, so that might be handy. For a connector, I'll probably go with a .1" header, just to cut down the cost.

Relays would be awesome, but they would cost around £3 each from what I can see, which as it stands would easily double the BOM cost.

I guess you could do something similar by designing the amplifier for >> 1 Mohm input impedance and put a 990k + 25k trimmer in parallel.

Absolutely - I was planning to do something very similar

60 dB requires better than 0.1 % matching between the two channels. Are you going to match the caps in the input attenuator own to better than 0.1 %? What about the voltage coefficient (the capacitance of a cap changes with applied voltage). I believe that 'cheap' differential probes (eg. the ~$300 ones) have their CMRR limited by the voltage coefficient of the attenuators.

I'll just have to get the best trimmer caps I can reasonably include... The main limiting factor here is cost; I'd like this to fill the gap in the market of 'entry level' differential probes.

Thank you very much for the info and advice

[robrenz]

Take a look at this DXC100A 10X-100X probe manual from Lecroy. It has schematics and calibration procedures  .  This gives an indication of what is involved in optimizing CMMR and response.  I have one of these and the Preamble/Lecroy DA1855A 100MHz diff amp it mates with.

[Mechatrommer]

just for the "mailbag update" i've ordered few 0.1% 1Mohm resistors, varieties chocollete of var caps 4-60pF, 2-30pF, 4-10pF etc from digikey, and only 2 Bourns 1Kohm multiturn trimpot (i wish i can get more but they are expensive), i wish i can get 0.05 - 0.01% resistor but, thats what my "error budget" tells me for now. if your design space does not limit you, maybe you can build your own or buy variable butterfly capacitor, just like in robrenz's 7a22 teardown. maybe it has more resolution for matching and stability purpose, i'm not sure. pcb layout is another issue, high impedance node of the attenuator likes to couple with anything near it including pcb gnd plane and copper pour, i cant afford to keep ordering prototype pcb from itead my budget is limited

[hammil]

Take a look at this DXC100A 10X-100X probe manual from Lecroy. It has schematics and calibration procedures  .  This gives an indication of what is involved in optimizing CMMR and response.  I have one of these and the Preamble/Lecroy DA1855A 100MHz diff amp it mates with.

Awesome This will come in very handy

just for the "mailbag update" i've ordered few 0.1% 1Mohm resistors, varieties chocollete of var caps 4-60pF, 2-30pF, 4-10pF etc from digikey, and only 2 Bourns 1Kohm multiturn trimpot (i wish i can get more but they are expensive), i wish i can get 0.05 - 0.01% resistor but, thats what my "error budget" tells me for now. if your design space does not limit you, maybe you can build your own or buy variable butterfly capacitor, just like in robrenz's 7a22 teardown. maybe it has more resolution for matching and stability purpose, i'm not sure. pcb layout is another issue, high impedance node of the attenuator likes to couple with anything near it including pcb gnd plane and copper pour, i cant afford to keep ordering prototype pcb from itead my budget is limited

Try http://oshpark.com/ . $5 per square inch, 3 good quality boards, 12-day turnaround.

[Marco]

60 dB requires better than 0.1 % matching between the two channels. Are you going to match the caps in the input attenuator own to better than 0.1 %? What about the voltage coefficient (the capacitance of a cap changes with applied voltage).

I'm intrigued by Douglas Smith's balanced probe as a way to sidestep this problem .... he uses a pure resistive divider with coax just like DIY single ended passive high frequency probes, but uses 2 for differential measurement with a BALUN/combiner instead of a difference amplifier.

I wonder if this wouldn't be superior for the HF part of the spectrum, obviously it can't work down to DC ... but you can use a separate measurement circuit (with it's own high impedance attenuators) for the LF and combine them.

[ericjackson05]

Sorry my English, which is not my mother tongue. I try my best to explain explictly.(-:
I have been designed a differential probe according  to your circuit diagram. Some problems occur in the debugging process.
1. The provided current of button cell such as CR2032 is not enough for this application, the voltage of which is pulled down to below 2V. two AA batteries is OK.
2.The output current of MAX865 is not enough. positive and negtive 12V pulled down to +8V and -7V or so;
3. The input signal amplitude of AD8130 is only -2.5V to 2.5V according to chip manual and my test, which is not rail-to rail. Which show strong limitation.
 My email is ericjackson05@qq.com, any one can contact with me if you have interest. (-: