glscopeclient demo - analysis of full duplex 100 Mbit Ethernet

Demonstrating glscopeclient performing various protocol decodes and signal integrity analysis on a full duplex 100baseTX Ethernet link.Data was acquired using a custom test fixture with SMA connections to the scope, a Teledyne LeCroy WaveRunner 8104-MS.
Get the code:
github.com/azonenberg/scopeha...
See updates on my Twitter:
/ azonenberg
Development discussion and support: #scopehal on Freenode

[@reps]

Is all processing happening on PC? Impressive software either way!!

[@AndrewZonenberg]

Yes. All of the protocol decoding, clock recovery, etc is my own open source code running on the PC. For this demo I'm pulling waveform data from a Teledyne LeCroy WaveRunner 8104-MS. The native data format is essentially raw 8-bit ADC samples plus some metadata for sample rate, trigger offset, etc. Everything past that is my code. Libscopehal also supports Rigol, R&S, Siglent, and Agilent scopes to some degree, although the LeCroy driver is the most well developed and tested since that's what I have in my lab. Nobody's offered up a Tek to test on yet, and we've never tested the Agilent driver on a modern Keysight scope but it might work.

[@chrizlax]

Really nice snapping window management, and the small text is neat too. Too many applications waste all my monitor space without any increase in usability, yours looks really efficient and fast to get the graphs on the screen you want. I liked the minimal config as well, e.g for the PLL, no need to set anything but the expected frequency. Is autoscaling planned for the stats plots? (e.g BER) Seems it would save you some time?

[@AndrewZonenberg]

BER plots are actually hard coded right now to I think 0 to 1e-12 on the Y axis. Pretty much all of the stats plots already autoscale on the Y axis, but since the X axis is shared among plots in a group it's a bit trickier. When you have plots with incompatible X axis units in one group (say volts and picoseconds) it draws against the underlying value without regard to units. Which can lead to some display issues. As far as X axis autoscaling, I try to use that very carefully because it's easy to end up annoying the user. Eye patterns and BER bathtub curves both take over the entire waveform group's X axis and force it to be +/- 1 UI from center. I haven't used it elsewhere yet.