Pathology: Differential Mismatch
The amplitude of the signal on the p side of the differential pair (channel 2, teal color) is -250 mV + 1050 mV = 800 mV. The amplitude of the signal on the n side of the differential pair (channel 3, green color) is -100 mV + 500 mV = 400 mV. This is an imbalance between the two lines that make up the differential signal of 100%.
There are three probable causes of this asymmetry:
- You are not measuring the correct lines that make up the differential pair.
Verify the pins that are being measured are the signals on the two lines that make up the same differential pair.
- One probe is set as a 50% attenuating probe or gain of 2x.
The absence of reflections at the transitions suggests that the ends of the lines are well terminated and the probe is not loading the line very much. This suggests the use of active probes. It may be possible that one probe is set for either attenuation or gain. Verify the gains of the probes on each line are the same.
- The output impedance of one of the drivers is twice the impedance as the other line.
The absence of reflections suggests each line is well terminated. The different voltages in each line could be due to different voltages launched, which might be due to different series terminations either on the circuit board or on-chip. Check the series terminations at the drivers for the two lines.
A differential signal is the difference in voltage between the two lines that make up a differential pair. The average voltage between the two lines is the common signal component and the difference in voltage is the differential signal component.
If the two voltages that make up the differential pair signal level are not switching the same magnitude, the common signal component will be modulated. This can cause EMI problems and if the common mode rejection ratio of the receiver is not large enough, will contribute to differential noise. A differential system should never be designed with asymmetrical drivers.
Usually there is some asymmetry in the path that causes the imbalance in the two voltages. This should be found and eliminated.
Any asymmetry between the two voltages that make up the differential pair signals will cause a distortion in the differential signal and the creation of some common signal. For example, in the figure below, one line is skewed a time equal to the rise time.
The common signal, in red, has a modulation to it and the differential signal in blue has a longer rise time. Usually the impact on the differential signal is small. If not, this is an indication of a really big problem which should have been caught in the design phase.
An asymmetry between the voltage levels of a differential pair are usually very slight, due to a rise time difference, a skew between the two paths, or a different output impedance of the two drivers of the differential pair.
When a large difference is observed, like 2x difference in amplitude, this is usually an indication of a set-up problem, or an on-chip source impedance problem. If there were a termination asymmetry, you would see ringing in the received signal due to the mismatch in termination.
Set up problems could be due to the scope, the probe or the specific lines that are being measured.
Impact and Effect
The quality of the differential signal determines the bit error rate. Anything which distorts the rise time of the differential signal will contribute to jitter and collapse of the eye.
Any asymmetry between the lines that make up the differential pair, such as intra-pair skew, impedance difference, driver impedance or termination impedance, will potentially cause some of the differential signal to convert to common signal. This will generate common signal and will distort the differential signal. If the asymmetry is significant enough, it may result in an increase in the bit error rate.
All asymmetries in the driver and in the interconnect will affect the asymmetry in the voltage between the two lines that make up the differential pair. This will result in a distortion of the differential signal and creation of a common signal.
The common signal by itself may not cause a problem, but it indicates the differential signal has been distorted. Some of the common signal will be converted into differential noise by the receiver. If the distortion of the differential signal is low enough to keep the received eye within the eye mask limits of the application, the distortion may not be a problem and can be tolerated in the design.
Every controllable asymmetry should be minimized to provide extra margin, so that small and unavailable asymmetries do not bring the system down.
When asymmetries are observed, as with every measurement, it should be verified that the measurement is not an artifact.
The next step is to identify the source of the asymmetry. Places to look are:
- Diver output impedance
- Line to line impedance difference due to cross section variation
- On-chip termination impedance difference
- DC blocking cap differences
- Skew in the driver
- Skew in the two lines due to a length difference
- Skew in the two lines due to glass weave effects
See also topics related to:
- Differential pairs
- Glass weave effects
- Broadside verse edge coupled differential pairs