14
\$\begingroup\$

I've encountered differential signals in a few places, such as a differential output audio amplifier and now in a project working with DMX which is similar to RS-485. (Here's a similar question about RS-485.)

Looking at a waveform from a DMX lighting controller for example, I connected the channel 1 probe to D+, channel 2 probe to D-, and both ground leads to ground.

It produces this display:

Differential Signal Measurement

While this is usable, I know it's still not the correct way to look at differential signals.

What is the correct way? I've heard of "differential probes;" does that mean I need to purchase new probes?

\$\endgroup\$
  • 3
    \$\begingroup\$ You have been given answers for when you really neeed to, but realistically speaking, most of the time you don't need to. Perhaps you spend some time looking at both channels and convince yourself that the differential signals are properly mirror images of each other. Once you have done that, absent high noise, you can simply look at the non-inverted half of the pair when you want to examine timing, data, etc - if there's any further reason to doubt the differential implementation, you can go back and look (is one half of the pair delayed?). \$\endgroup\$ – Chris Stratton Jul 20 '13 at 15:38
  • \$\begingroup\$ Also if you have single ended to differential transmitters, or differential to single ended receivers, be sure to examine the behavior of those - some can have suprising delay. \$\endgroup\$ – Chris Stratton Jul 20 '13 at 15:39
  • \$\begingroup\$ Thanks Chris. I posted this question and answer because there wasn't already one on EE.SE about it, and because I think points like yours (you don't usually need to) are just as valid to someone learning how. Knowing how and when it's actually necessary are both totally valid points. \$\endgroup\$ – JYelton Jul 20 '13 at 23:53
14
\$\begingroup\$

The reason you can't measure differential signals quite as easily with an oscilloscope has to do with the fact that oscilloscopes are (generally) not floating. The ground lead on the probes are connected to the oscilloscope chassis, which in turn is earth grounded. Because of this, anything you connect the ground lead to will also be connected to earth ground. (As videos I link below demonstrate, this is dangerous if measuring high voltage!)

When you measure two random points with a multimeter, the meter is floating, so you're not connecting either point to actual earth ground, which lets you measure differences between points without concern that you're creating a short circuit.

In low voltage signal applications, tying one side of a differential signal to ground can cause problems and might damage a transceiver.

There are two ways to measure differential signals with an oscilloscope:

If you have a two-channel oscilloscope, connect one side of the signal to channel 1, and the complementary signal to channel 2. The ground leads stay unconnected.

Since you are interested in the difference between the signals, you want to subtract channel 2 from channel 1. Most scopes provide a way to add or subtract channel 1 and channel 2 inputs. On some scopes you might have to add channel 2, but invert it so that you are effectively subtracting it.

In this image, the scope has an A-B mode which subtracts channel 2 from 1:

Differential Measurement Subtracting Second Channel

The other way is indeed using differential probes, and provides better results without reducing the number of usable channels on the oscilloscope. (And are usually designed for safer high-voltage measurements.) However these probes are expensive.

W2AEW does a superb job explaining these concepts in his video on differential measurements using oscilloscopes. There's also a video by BTC Instrumentation which shows the channel subtraction method in more detail.

\$\endgroup\$
  • 2
    \$\begingroup\$ Please why The ground leads stay unconnected ? \$\endgroup\$ – The Beast Dec 4 '16 at 22:04
  • \$\begingroup\$ From hardware engineers I get to hear the probe ground needs to be connected to ground. \$\endgroup\$ – Ginu Jacob Dec 21 '16 at 7:00
  • \$\begingroup\$ Probe ground already is connected to ground - on the other side. However - true, it does not hurt and actually yields better results when conntected to ground on both sides - one could use the calibration ground pin for that. \$\endgroup\$ – Rok Jarc Dec 11 '19 at 8:52
  • \$\begingroup\$ Or do my way, just take off the Earth wire in the plug. LOL! \$\endgroup\$ – GeneCode Jan 9 at 5:46
4
\$\begingroup\$

Other people have already explained how to configure your two oscilloscope channels to measure a difference between two signals.

Of course, you run into a problem if you want to measure some other signal besides that differential signal. I take it your scope has only two channels.

In such a case, you'd need a differential probe. But, as JYelton mentioned, differential probes are expensive.

However, as long as you're not trying to measure very high voltages or very high frequencies, you can simply make a differential amplifier circuit on a protoboard. You can power it from a battery or perhaps from the device you're measuring.

Differential Amp

Just make sure you use an op amp with adequate voltage swing, make sure it's stable, and you're good to use this simple circuit as a cheap differential probe.

\$\endgroup\$
2
\$\begingroup\$

For differential measurements, most scopes with at least two inputs support the following features:

  • INVERT channel A or B (or both), it is usually found near position or AMPL/DIV setting.
  • ADDing channel A and channel B, it is usually found where you enable a channel.

INVERT one channel, ADD both channels et voila, your differential measurement.

\$\endgroup\$
1
\$\begingroup\$

Measuring 485 (or suchlike) signals differentially is within the features of most oscilloscopes these days and this has already been answered satisfactorily but I would ask "why bother" or "what are you trying to achieve"?

If you were debugging a flaky 485 link I think you'd stand a good chance of seeing the problem single-ended. It's likely any faulty connection will pass its flakiness from one of the 485 wires to the other via the terminator. I'm not saying this is 100% but I am saying that on the several 485 lines I've had to debug I've never felt the compulsion to measure differentially. If my received data were flaky I wouldn't hesitate to look at the single-ended output from the data receiver - after all, this is what feeds the MCU and it is the MCU that is telling me the data flaky.

I've done a few high speed (80Mbps) links that transmitted and received differentially. The data (purposefully) was tri-level edge information in order to magnetically couple to the receiver. I've never felt the compulsion to debug these using differential probes either - I've measured single ended to make sure the data looked OK and then gone onto the single-ended output from the receiver chips to see what the trouble was.

I guess for analogue differential signals, differential measurements are more important because what may look "nasty" on one line could look fine differentially.

\$\endgroup\$
  • \$\begingroup\$ Andy just to clarify, if you measure a differential signal "single-ended" does this mean tying the device ground to the oscilloscope ground and measuring one half of the signal? If so, can that ever create a problem for the transceiver sending or receiving the signal? \$\endgroup\$ – JYelton Jul 20 '13 at 23:55
  • \$\begingroup\$ @JYelton no I'm not advocating that I'm just saying looking at one of the wires with respect to ground and leaving the other one to go about its normal business tells you most things on digital differential comms. \$\endgroup\$ – Andy aka Jul 21 '13 at 10:49

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.