VFD speed noisy?

The spindle control on the Bantam / LinuxCNC system seems to be suffering from noise. Although I can set a constant speed demand within LinuxCNC, the VFD itself seems to be having a difficult time maintaining a nice, smooth speed. 

The VFD itself is a rather unusual device, being of New Zealand origin and dating back to the mid 1990s. I acquired a few of these after the demise of Wavedriver Ltd (a maker of electric vehicle drives and components) at the hands of Powergen (national utility company, run by a bunch of grubby thrusters). It works well and is unusual in having a single phase 240V input and a power rating of 5.5kW - most single phase VFDs max out at ~4kW these days, as you are expected to use 3-phase above that sort of power.

Luckily I grabbed a copy of the manual some years back, so I can look up the functions and connections.

Without diving into the detailed info about setting up the analogue inputs (I've already done that of course), this high level diagram tells me what I need to know:

And this is what it actually looks like in the cabinet. The cover is off the VFD and there's a scope and probe draped over the wiring, otherwise it's actually reasonably tidy in there.

And here's the signal connections on the VFD end:

The 3 wires at the VFD that I am interested in are:

• T13: +10V reference voltage
• T14: Analogue input 1
• T15: Analogue 0V

And yes, the screen is already connected to the analogue 0V / ground.

At the other end of the cable, the "digital potentiometer" within the Mesa 7i76 sits on an electrically isolated island, with the signal passed across the isolation barrier by means of a high speed digital opto, followed by a low pass filter and opamp, presumably powered from the +10V analogue reference voltage (although I can't be arsed to check that).

A point worth registering is that the analogue island isn't grounded at either end. There are a few different options here:

• Ground one or both ends of the cable, noting that the analogue ground and screen are the same (possibly not helpful?), as I seem to recall I ran out of wires in the screened cable. So perhaps grounding the VFD end of the screen would make sense.
• Slap a cap between the analogue control voltage and the analogue ground / screen - to dampen any differential mode noise. A lossy electrolytic might work here.
• Put some ferrite clamps on the cable - to attenuate any common mode noise.

Here's what it sounds like - and looks like on the scope. Lots of very energetic bursts of ringing going on there.

• Result of fitting a ground wire to the screen - zich.
• Result of slapping a cap on the analogue voltage - zilch.
• ...and I couldn't find my stash of ferrite clamps, so this will have to wait until a load of Chinese clamps of unknown spec arrive tomorrow.

So although annoying, this isn't critical. However, it's been persisting for a long time and the current hiatus (replacing the PC) would seem like a good opportunity to fix it.

I'll update as and when progress is made.....