Spindle Encoder Crashing Computer

When I try to run the spindle with the spindle encoder connected, it spins anywhere from about a few seconds to about a minute, then crashes the computer. I can run the spindle with the spindle encoder disconnected for an unlimited amount of time, no problem. We installed a hall effect sensor on the spindle for the index. (We want to use it for spindle orientation for the tool change later.) It matters not if the index wire is connected to the 7i76. If the encoder wires are connected, and I try to spin the spindle, the computer crashes to an immediate black screen and does an immediate reset.

The machine is an XYZA mill. I'm using a 5i25, 7i76, and 7i73 from Mesa. The Linuxcnc version is 2.7.14.

On the 7i76 I have the W4, W5, and W6 jumpers to the left for single-ended mode.

SERVO_PERIOD = 1000000

The encoder has a resolution of 256 ppm. hm2_5i25.0.encoder.00.count shows 1280 counts per revolution. This may be due to getting four for one from quadrature and the spindle pulley ratio?

What should I try next?

My first guess would be that there is a wiring problem.

Yes, This is likely wiring related, probably a ground loop communicating switching noise current from the spindle VFD --> motor frame --> encoder shield --> host PC

if you have a shielded encoder cable, it should be terminated only at one end
(not connected to 7I76 ground)

Other possible fixes:

Line filter on VFD line power

Common mode choke (giant ferrite bead) on VFD motor output ( around 3 UVW wires, not the motor frame ground connection)

Common mode choke on encoder cable

Thanks for the wisdom. I'll try this soon!

I just got back to this. Here are the things I've tried so far:

1. Verified that yes there was switching noise on encoder output with oscilloscope.
2. Installed common mode chokes on vfd output.
3. Installed dedicated grounding rod for vfd per vfd manual.
4. Made sure the encoder cable shield is not connected to 7i76. It was not.
5. Retested for noise with o-scope. Found no noise but the computer still crashes. (Remember it runs perfectly with the encoder disconnected.)

I was preparing to purchase and install the line filter for the vfd, but then ran the following test.
I unscrewed the encoder from the spindle and spun it with a rechargeable drill. It crashed the same way with the 3 phase power not even applied to the spindle or servos so I'm now assuming it cannot be switching noise. The differential signal looked fine on the scope until the computer crashed.

What should be my next step in the troubleshooting process?

That's really odd

Bad PC that crashes on some FP operations?

Thanks for the tip. Can you think of a way I could test this?

Does our working mpg hand wheel give us any clues?

Using the halmeter shows that linuxcnc is processing the data from the encoder correctly, position and velocity, until the computer crashes.

Its all pretty weird so I dont really know how to test other that try a different PC

Just had another thought: is it possible that there's a short in the encoder/encoder 5V wiring? That might cause a computer reboot if momentarily shorted

Peter,
You were right about the PC being the culprit. A new motherboard solved the crashing computer problem.

Another question:
Our spindle encoder is differential, but the index sensor, mounted on the spindle rather than the servo, is ttl. Can I set the jumpers on the 7i76 so that phases A and B are differential and the index is ttl?

I think I'll repost this as a new thread.