My first project: Low-cost differential probe

[hammil]

I always hated the problem with mains-earth-referenced oscilloscopes. Not only are the input grounds all shorted together, it can be bloody dangerous to probe around other earthed boards if you don't know exactly what you're doing.

The solution, of course, is to use an active/differential probe, however these can cost hundreds if not thousands of pounds/dollars. So, I decided to roll my own. I honestly don't believe this has been done before, but I'm sure someone will prove me wrong.

I call it: The OmniProbe! (Yes, I had to give it a wank name.)

It consists of a differential amp, a little input protection, and some power ICs. That's it. Plug one end into the scope, and the other into your probe, and it puts at least 4M of isolation between the signal and mains earth (It's likely to be significantly higher, need to measure it properly), with ~6M of differential impedance.

The main component in the design is the AD8130 differential amp, specifically designed to link signal paths together while keeping the grounds isolated. Input impedance isn't quite JFET-level, but it's decent. Bandwidth is good too: 290MHz with a 0.2V signal, or 175MHz with a 2V signal. Unfortunately where it suffers is the maximum differential input; the datasheet specifies only 5V before performance starts to degrade, somehow (again, I need to take proper measurements). If higher-voltage signals need to be measured, however, a x10 attenuator will work quite nicely. Noise and biasing in the chip is all extremely low.

A little aside: I noticed that with op-amps, you have to choose two of these three, regardless of cost: high working voltage, high bandwidth, high input resistance. I haven't yet found a chip that has everything.

The whole device runs from a CR2032 battery, which is boosted up to 6V with a switch-regulator, then converted to +/- 12V, which gives the amplifier it's best performance.

Input protection is just a couple of zeners, back to back, to limit the maximum input voltage to the amplifier. Anything more would increase capacitance between the inputs. The diodes alone add around 6pf, so I included a jumper that you can remove if you want to disconnect them.

I also added an adjustable isolation output resistor to serve as compensation for the scope's capacitance.

The PCB is designed to be mounted in a Pomona 2391. Overkill, perhaps, but it's the easiest way to shield a prototype design like this, I find.

Schematics and PCB previews are attached. Oh, the PCB images are a little old, so they're just for illustration really.

[GK]

Not bad for a 1st project, but those 500R input protection resistors are going to limit your bandwidth.

That AD part has a differential input capacitance of 3pF. That alone (not including strays) with 1k works out to -3dB at a little over 50 MHz.

You can fix that problem by putting small value capacitors in parallel with the 500 ohm resistors.

[Mechatrommer]

good to see there's another diff probe project popping in... here is my venture (i dont want to update every little crannies of the details until i get a firm setup, not yet).
High Voltage 10-100MHz Differential Probe Investigated

1) is your intention to make the "input attenuator" separate? (external?) since 0.2V or 2V differential is close to useless (for my app only sorry )
2) the good thing is you are using battery (portable). i am using bipolar supply so i can concentrate on the diff amp circuit. i was considering using 2x 9V (rectangle) battery but still no luck finding the suitable enclosure.

since you are using switching circuit. i will be interested looking at your circuit noise floor. and your device bandwidth profile and cmrr. cheers.

[Marco]

What do you intend to use for the probe head?

[hammil]

Thanks for all the feedback, everyone

Not bad for a 1st project, but those 500R input protection resistors are going to limit your bandwidth.

That AD part has a differential input capacitance of 3pF. That alone (not including strays) with 1k works out to -3dB at a little over 50 MHz.

You can fix that problem by putting small value capacitors in parallel with the 500 ohm resistors.

That's superb, thank you - should've thought of that myself

1) is your intention to make the "input attenuator" separate? (external?) since 0.2V or 2V differential is close to useless (for my app only sorry )
2) the good thing is you are using battery (portable). i am using bipolar supply so i can concentrate on the diff amp circuit. i was considering using 2x 9V (rectangle) battery but still no luck finding the suitable enclosure.

since you are using switching circuit. i will be interested looking at your circuit noise floor. and your device bandwidth profile and cmrr. cheers.

Max 'guaranteed' differential input is listed as ~5V, but in terms of absolute ratings, it's 17V. I need to check what kind of performance impact it has when 5 < Vdiff < 17.

To keep parasitics to a minimum, I was planning to have nothing on the board in terms of attenuation, but of course one could buy a basic x10 attenuation BNC module fairly cheaply, in addition to the one on the probe. The amplifier is easily sensitive enough to handle significant attenuation, so with two x10's, one could theoretically measure up to 500V.

I'm using component values over and above those specified in the datasheet, to minimise noise. One good thing about the AD8130 is that the CMRR is given as 70dB at 10MHz, which is excellent.

What do you intend to use for the probe head?

Anything that can connect to a BNC will fit, but I suppose ideally it would be any good quality passive probe, with at least a switchable x10 attenuator.

[codeboy2k]

nice design. I like the battery operation.

I'm using component values over and above those specified in the datasheet, to minimise noise. One good thing about the AD8130 is that the CMRR is given as 70dB at 10MHz, which is excellent.

Although that CMRR is good, since mechatrommer specifically asked about the switcher, I'll talk about that here.  What you really care about is the op-amp PSRR when dealing with switching noise on the power rails, and that op-amp has about 55dB PSRR at your switching frequency, which for the MCP1624 is 500Khz. The voltage doubler operates at 24Khz.

With your MCP1624 circuit, and a 10uF output cap, assuming that your circuit draws 30mA total (just my guess), the worst case switching ripple will be at the high end of the the switcher duty cycle, say at 80%, it's going to be on for 900ns, thus, 30ma = Cout * dV/dt, dV is the ripple, dt is the FET on time.

Heck, lets go worst case and assume 100mA draw from the switcher(!) and 80% duty cycle.. then 100ma = 10uF * dV/900ns , dV = 9mV power  supply ripple.

9mV power supply ripple at 500 KHz, and the op-amps PSRR at 500 KHz is 55dB, then that 9mV power supply ripply will result in an output voltage of 10^(log(9mv) - (55/20)) = 16 uV ripple at the op-amp output (due to the power supply ripple).

That's a worst case guess I just did, and it depends if 16uV of error is acceptable to the designer Chances are the circuit is not drawing more than 50ma, and maybe even just a few 10s of mA, so that power supply ripple will be even less and that error at the output will be even less still.

The PSRR at 24 KHz is about 85 dB, but I didn't do the math for its ripple and noise contributions. But 85dB is much better, and the noise contribution will be much lower.

If it was me, after building the first cut of this circuit, I'd certainly want to measure its current draw, know that value, and figure out what its ripple should be, then measure that to know its true value, and decide if the output error due to power supply ripple after rejection is OK for me and my design goals.

Cheers!

[Marco]

Anything that can connect to a BNC will fit, but I suppose ideally it would be any good quality passive probe, with at least a switchable x10 attenuator.

In that case you don't really need the input resistors, you don't need to worry about input capacitance (in fact you need a little more probably) and you should have a 1 meg transistor parallel with the input I think.

[poorchava]

I would add a possibility to switch between 1M and 50R modes. 50R is useful and many lower-range scopes don't have that.

[sorin]

In my opinion this don't offer any type of protection, or solution for your problem.
As I know, you need a isolated amplifier(optical or magnetic isolation).

[Mechatrommer]

i've downloaded the AD8130 diff opamp, quite good candidate as a replacement part for my device thanks for the link. the output swing will depend on the gain, Vo = GAIN X Vi, the OP indicating he's using unity gain, i wonder if that 70dB CMRR @ 10MHz is also valid at unity gain? i havent checked the DS gotta pick the kid now . chiaow.

[poorchava]

As I know, you need a isolated amplifier(optical or magnetic isolation).

Two problems:
-isolation amps are VERY expensive
-they have really modest bandwidth (ie. AD210 , -3dB@20kHz)

In some expensive scopes that have isolated channels (Tek used to do that) the channels are isolated on digital level, after the analog frontend and conversion. High speed (XX MHz) analog isolation is 'a bit' hard to do.

[sorin]

this circuit  is useless, for what you want to use it.
suppose that you want to measure two points, 40 v and 41 V  respectively potential referred to ground.
the differential voltage is just 1Volt, but what will happen with your circuit?

[hammil]

9mV power supply ripple at 500 KHz, and the op-amps PSRR at 500 KHz is 55dB, then that 9mV power supply ripply will result in an output voltage of 10^(log(9mv) - (55/20)) = 16 uV ripple at the op-amp output (due to the power supply ripple).

Thank you very much for the analysis

The standby currents for all the ICs is very low, and by my estimate it would only draw a few mA from the supply in operation. I need to test that, however. Plus, 16uV is almost invisible even at 500uV/div on a scope, so I think it's well within the acceptable limit.

Again, really appreciate the time you took with this. I had no idea I'd get this much feedback!

In that case you don't really need the input resistors, you don't need to worry about input capacitance (in fact you need a little more probably) and you should have a 1 meg transistor parallel with the input I think.

I'll have to test that out - I'll find some more information about passive probes. However, I'd really like a higher differential impedance than 1M, and I'd like to take advantage of the low input capacitance (in terms of effect on the circuit-under-test).

Regarding the input resistors, I would definitely like the option to use the clamping diodes, as a differential voltage of 18V can easily damage the AD8130. If the resistors weren't there, a voltage in excess of around 12V would effectively short out the inputs, possibly destroying the diodes and/or causing problems with the circuit. Of course, if you didn't need the device to probe around devices with unknown voltages, one could just short out the pads.

this circuit  is useless, for what you want to use it.

I'm sorry you feel that way, however it seems it is quite clear to me and other users why a device such as this would be useful.

[hammil]

I would add a possibility to switch between 1M and 50R modes. 50R is useful and many lower-range scopes don't have that.

I actually considered this to begin with, however adding a switch in series would increase the dreaded input capacitance by around 6pf. I could use a jumper, however... perhaps a later version of the board might have something

One could of course use a 50R BNC terminator, but it would be nice to have everything on one board.

[Mechatrommer]

this circuit  is useless, for what you want to use it.

it will depends on what the OP want to measure (low vs hi volt CM). if its within the spec values (diff and CM voltage) then the OP is good to go even without an attenuator, i think. if he has to provide an attenuator, then thats the biggest part, every aspects esp CMRR will rely on the attenuator (front end) matching. there's nothing worst than a low CMRR (i can live with low PSRR) ie an error voltage that follows signal shape and frequency. and that is so far what i figured out from my short experimentation.

What you really care about is the op-amp PSRR when dealing with switching noise on the power rails

a "not so good" "random noisy output" can be averaged in dso, but not the non-rejected CM voltage. i was just asking for output noise level (since my circuit's noise is coming from another aspect, ie opamp gain), but that is not so much an issue compared to CMRR.

[Mechatrommer]

I would add a possibility to switch between 1M and 50R modes. 50R is useful and many lower-range scopes don't have that.

I actually considered this to begin with, however adding a switch in series would increase the dreaded input capacitance by around 6pf. I could use a jumper, however... perhaps a later version of the board might have something
One could of course use a 50R BNC terminator, but it would be nice to have everything on one board.

i believe what he meant is that an opamp's output that capable of driving 50ohm load, it has nothing to do with the input stage. i checked the AD8130 it only specified up to 100ohm load, hence it cannot properly show a "halved magnitude" on 50ohm terminated coax. if you dont terminate your coax, then your signal reaching the scope will be out of proportion (due to coax capacitance). terminating it with 50ohm will show you 33% of the signal, but i'm not sure about the flatness with this mismatched characteristic impedance 100ohm opamp output - 50ohm coax - 50ohm terminator setup. the easy fix for me is we need another buffer on the output that is capable to drive 50ohm load. all opamps range that i used are capable except they all combined cost alot more than your AD8130.

edit: sorry i re-checked the datasheet. 100ohm load is actually the 50ohm + 50ohm terminated coax. so yes, this amp alone can be used without any buffer...
"The AD8129/AD8130 can directly drive loads of as low as 100 ?, such as a terminated 50 ? cable."
i always got mixed up with this 50ohm cable and 100ohm load 

but still you cannot make the 50ohm terminator to be "on the board" with long coax hanging after that to dso. a terminator must be on the dso side.

[sorin]

I'm sorry you feel that way, however it seems it is quite clear to me and other users why a device such as this would be useful.

With "is useless" I mean that , dont resolve this problem:

I always hated the problem with mains-earth-referenced oscilloscopes. Not only are the input grounds all shorted together, it can be bloody dangerous to probe around other earthed boards if you don't know exactly what you're doing.

Don't that is useless in general.

[Mechatrommer]

I'm sorry you feel that way, however it seems it is quite clear to me and other users why a device such as this would be useful.

With "is useless" I mean that , dont resolve this problem:

I always hated the problem with mains-earth-referenced oscilloscopes. Not only are the input grounds all shorted together, it can be bloody dangerous to probe around other earthed boards if you don't know exactly what you're doing.

Don't that is useless in general.

ok i missed that statement, actually its already in the OP. if dangerous voltage is the concern, then its useless AD8130 is only speced to ±10.5V CM, nothing dangerous about it.
and how you are going to use Pomona 2391 for 2 input and 1 output device? it only got 1 input 1 output?
http://www.pomonaelectronics.com/pdf/d2391_1_01.pdf

[codeboy2k]

9mV power supply ripple at 500 KHz, and the op-amps PSRR at 500 KHz is 55dB, then that 9mV power supply ripply will result in an output voltage of 10^(log(9mv) - (55/20)) = 16 uV ripple at the op-amp output (due to the power supply ripple).

Thank you very much for the analysis

The standby currents for all the ICs is very low, and by my estimate it would only draw a few mA from the supply in operation. I need to test that, however. Plus, 16uV is almost invisible even at 500uV/div on a scope, so I think it's well within the acceptable limit.

You're welcome. Even though the quiescent currents are very small, when it's driving your scope, especially if you use the 50ohm input setting on your scope, it might draw a few mA or more.  It will never draw 100mA or even 50 or 30 mA like I said, I was just trying to pick a worst-worst case. And thus 16uV (worst, worst case!) of noise is insignificant, at 500uV/division on your scope, so your switcher will work just fine.

For a more realistic  10mA , just plug it in:
ripple =  (Ton*I)/Cout = (900ns*10ma) / (10uF) = 900uV power supply ripple
900 uV @ 55 dB PSRR = 10^(log(900uV) - (55/20)) = 1.6uV at the output of the opamp 

Cheers!

[fcb]

Nice idea. A few comments though.

1. Assuming you want to use the pomona case, you'll have insulate the input bulkhead BNC chassis from the output.
2. Why bother with an internal supply? You're going to compromise the noise performance probably by doing so.
3. Did you build a prototype to check what the CM range actually turned out to be?
4. Most scope probes are x10 attenuation and most differential probes are x10/x25/x50/x100 at least - and you pay mega$$$ for highBW differential, even modest tek 100MHz probes are $1K.
5. High Z is easier with attenuation...

Why not build a differential probe with a much higher (useful) range and a lower -3dB, now that would be useful for SMPSU work - have a peek at the PICO TA041 for sensible specs.

[codeboy2k]

ok i missed that statement, actually its already in the OP. if dangerous voltage is the concern, then its useless AD8130 is only speced to ±10.5V CM, nothing dangerous about it.

The OP said he would use a divider on the input, i.e. 1 10:1 probe. 
But that has it's own problems, like then you don't need the 500 ohms at the input of the opamp, and the probe adds capacitance that might need compensation at the opamp inputs, plus the probe needs a 1M at the input to work as a proper divider...

but..

I don't think OP can do this, since they are differential inputs, you need to divide them both down before the input, not divide them relative to each other like a 10:1 probe would do it.

Best probe is just the bare end of the coax. divide both inputs at the op-amp.

and how you are going to use Pomona 2391 for 2 input and 1 output device? it only got 1 input 1 output?

He's using one BNC center pin and shield as the two inputs.  There's nothing wrong with this approach, but care has to be taken to isolate the BNC shield on the input side from the metal case of the Pomona box and the output BNC shield. (oops fcb already said it !)

[hammil]

ok i missed that statement, actually its already in the OP. if dangerous voltage is the concern, then its useless AD8130 is only speced to ±10.5V CM, nothing dangerous about it.

Not dangerous voltage as such, but dangerous current. It's not uncommon to have a logic rail capable of 10A or more, and it would be very bad news if you managed to short that out. And of course, with a x10 attenuator one could measure up to +/- 100V common mode, with 50V differential, which will suffice for a lot of power supplies.

Of course, this is just one application. I tried to design it to have low noise, and good signal integrity. And to be honest, I'm learning a lot of this as I go.

He's using one BNC center pin and shield as the two inputs.  There's nothing wrong with this approach, but care has to be taken to isolate the BNC shield on the input side from the metal case of the Pomona box and the output BNC shield. (oops fcb already said it !)

Damn, that's a good point. Should be possible, though, with some isolation materials and a little bodging.

Again, thanks to everyone for the feedback

[hammil]

Nice idea. A few comments though.

1. Assuming you want to use the pomona case, you'll have insulate the input bulkhead BNC chassis from the output.
2. Why bother with an internal supply? You're going to compromise the noise performance probably by doing so.
3. Did you build a prototype to check what the CM range actually turned out to be?
4. Most scope probes are x10 attenuation and most differential probes are x10/x25/x50/x100 at least - and you pay mega$$$ for highBW differential, even modest tek 100MHz probes are $1K.
5. High Z is easier with attenuation...

Why not build a differential probe with a much higher (useful) range and a lower -3dB, now that would be useful for SMPSU work - have a peek at the PICO TA041 for sensible specs.

1. Yeah, that's a very good point. Should be possible to do, though.
2. Need to get power from somewhere, and using a battery with a couple of ICs seemed the easiest way. The noise should be pretty low as well - 16uV at the worst worst case, as codeboy2k calculated.
3. Working on it
4. 5. Yeah, that's definitely something I'll consider, but I reckoned an external attenuator would do a better job.

Thanks

[BravoV]

Interesting thread, thanks for sharing.

Noob questions :

- 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 ?

[fcb]

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.