Poor mans differential probe

[NivagSwerdna]

I don't have a differential probe but I do have a 2 channel scope (Rigol 1052E).... if i put one channel on one side of the input and the other channel on the other side and use MATH- does that achieve the same thing?

[abraxa]

If you can live with the fact that both probes need to be connected to a common ground, then yes.

[NivagSwerdna]

If you can live with the fact that both probes need to be connected to a common ground, then yes.

I 'think' that is OK in this case. Thanks

[schmitt trigger]

You really have to ensure that the probes are matched as close as possible, which essentially means a couple of things:
> Use the same probe model from the same vendor.
> Compensate the probes

[David Hess]

If you can live with the fact that both probes need to be connected to a common ground, then yes.

Real differential probes also have both probes connected to a common ground whether they make this ground available or not.  On modern probes, this ground runs through the oscilloscope to the power connection to the device under test and it is severed like by using an isolated oscilloscope input, malfunctions including destructive failure are possible.

You really have to ensure that the probes are matched as close as possible, which essentially means a couple of things:
> Use the same probe model from the same vendor.
> Compensate the probes

This helps but is not enough for the best performance.  Unless x1 probes are used, the probe attenuation factor can significantly degrade the low frequency common mode rejection ratio and even if it does not, the oscilloscope's gain difference between channels will.

They used to make special probes which had trimmable DC attenuation for differential use but an alternative is to use the oscilloscope's variable function to trim the combined probe and channel gains to match before doing the subtraction.  Analog oscilloscopes all supported this and some DSOs do also.

[MarkL]

Using math for differential measurements will work, but you need to make sure that both signals, Ch1 and Ch2 in your case, are within the dynamic range of the ADC of the scope.  If one channel or the other has the output of the ADC pinned to a positive or negative value, you're going to get some strange looking waveforms with the resultant Ch1-Ch2.

Some scopes warn about over-range conditions and some don't.  To be sure you're not over-range, you should make sure the signal on each channel is fully on the vertical area of the screen, and then do the math for Ch1-Ch2.

[Ian.M]

Also, on a DSO using waveform math for the difference, if the common mode signal is large compared to the differential signal you are going to run into problems with limited ADC resolution.  e.g. an 8:1 signal level ratio will loose you three bits of resolution on the result (compared to the normal single channel resolution).

[David Hess]

Also, on a DSO using waveform math for the difference, if the common mode signal is large compared to the differential signal you are going to run into problems with limited ADC resolution.  e.g. an 8:1 signal level ratio will loose you three bits of resolution on the result (compared to the normal single channel resolution).

I almost posted about this.  If the subtraction is done after the signal are digitized, then the quantization noise from each digitizer adds.  If the subtraction is done in the analog domain, then this is not a problem but only very old DSOs worked this way.  As a practical matter, you just have to deal with it.

[bitseeker]

I don't have a differential probe but I do have a 2 channel scope (Rigol 1052E).... if i put one channel on one side of the input and the other channel on the other side and use MATH- does that achieve the same thing?

Basically, yes. Specifically, earlier posts dive into the nitty gritty. For getting a general look, it'd probably be fine. Be sure that each input does not exceed the limits of the probes or instrument. Even if the differential signal is just a few volts, the maximum voltage above ground must still be adhered to.

[macboy]

I don't have a differential probe but I do have a 2 channel scope (Rigol 1052E).... if i put one channel on one side of the input and the other channel on the other side and use MATH- does that achieve the same thing?

Basically, yes. Specifically, earlier posts dive into the nitty gritty. For getting a general look, it'd probably be fine. Be sure that each input does not exceed the limits of the probes or instrument. Even if the differential signal is just a few volts, the maximum voltage above ground must still be adhered to.

I would add that the probes must be identical, and must be calibrated to respond identically. The easiest way to achieve that is to first individually adjust the compensation on each probe, then activate the math A-B and connect both probes to the cal signal - this gives you a 100% common mode signal. Then adjust one probe to get the displayed signal to a minimum (maximally flat). You may also need to slightly adjust the vertical gain on one channel. This optimizes your common mode rejection. Any residual signal that you can't tweak out will also give you some idea about the limitations of this method.

[alm]

As David Hess already mentioned earlier, 1x probes may provide better common mode rejection than common 10x or 100x probes assuming you can live with their voltage and bandwidth limitations. They are very well matched by definition as long as the cable is the same length, since they are basically just a few hundred Ohm in series with the 1 MOhm scope input.

[David Hess]

As David Hess already mentioned earlier, 1x probes may provide better common mode rejection than common 10x or 100x probes assuming you can live with their voltage and bandwidth limitations. They are very well matched by definition as long as the cable is the same length, since they are basically just a few hundred Ohm in series with the 1 MOhm scope input.

Even with 1x probes, any mismatch in the gain of the two vertical inputs will degrade common mode rejection but 10x and 100x probes just make this worse.

The solution is to use the oscilloscope's variable function to trim the gain of one of the channels so the combined gain of each probe and channel match.  On an analog oscilloscope, you can *see* when they are matched because the noise visibly nulls out.

[MarkL]

...
The solution is to use the oscilloscope's variable function to trim the gain of one of the channels so the combined gain of each probe and channel match.  On an analog oscilloscope, you can *see* when they are matched because the noise visibly nulls out.

This works on analog scopes, but on the digital scopes I've used, the scope will take any fine adjustments of the vertical scale into account in the math so that it will *still* be subtracting the absolute values from each other, thereby preserving any probe or front-end mismatch.  Perhaps some DSOs work as you say?

I've been asking manufacturers for an independent offset and gain adjustment for each channel which is applied before and independently of anything else.  It would be handy for this application and in others to zero out short-term probing and front-end errors.  Perhaps it already exists; maybe someone can point out a scope that has it.

[bitseeker]

Depending on one's definition of poor as it relates to test equipment, there's this to consider:

https://www.eevblog.com/forum/testgear/new-low-cost-($170)-100mhz-differential-scope-probe-from-micsig/

There's 2-for-1 intro pricing mentioned in the thread. Find a fellow "poor man" to share the cost and order two.

I just wish it had 10X instead of 50X minimum.

[David Hess]

...
The solution is to use the oscilloscope's variable function to trim the gain of one of the channels so the combined gain of each probe and channel match.  On an analog oscilloscope, you can *see* when they are matched because the noise visibly nulls out.

This works on analog scopes, but on the digital scopes I've used, the scope will take any fine adjustments of the vertical scale into account in the math so that it will *still* be subtracting the absolute values from each other, thereby preserving any probe or front-end mismatch.  Perhaps some DSOs work as you say?

Old DSOs work this way.  I have not tried it on a new DSO.  If you had to multiply a channel by a constant before subtracting that would be annoying and add further quantization noise and I have assumed that most do it this way.

Adjusting the gain is actually pretty easy in the analog domain without resorting to variable gain in the signal path; just trim the ADC reference with an offset.

I've been asking manufacturers for an independent offset and gain adjustment for each channel which is applied before and independently of anything else.  It would be handy for this application and in others to zero out short-term probing and front-end errors.  Perhaps it already exists; maybe someone can point out a scope that has it.

Most modern DSOs either have vertical position controls which also operate as offset controls or separate position and offset controls.  Separate offset controls are a premium feature.

I have not looked into gain (variable) trimming.  I know some either do not support it at all or require it to be done as part of the math.