Trying to understand 1130 MRAM sporadic failures using Active FET Probe

USING ACTIVE FET PROBE TO WATCH SIGNAL DURING FAILURE

I had been plagued by the inability to watch what happens when the sporadic failure occurs since the oscilloscope probe and the logic analyzer probes both loaded the circuit enough that the bit never was dropped. 

I purchased an active FET probe which introduces much much less loading on a circuit. It was a Tektronix P6201 probe which was sold as new-old stock. I also purchased a used Tektronix 1101 Power Supply which the probe requires for operation. With both shipments finally received, I checked out everything before putting it to use.

TESTING AND CALIBRATING THE 1101 POWER SUPPLY

I downloaded the manuals and began testing the power supply. I certainly didn't want a bad supply to blow up an expensive probe. The box produces +15V, -15V and +5V and delivers it over a unique cable to the probe. I got the voltages set exactly correctly and validated proper behavior of the overcurrent protection. 

ACTIVE PROBE VOLTAGE LIMITATIONS ARE WORRYING

The probe uses a field effect transistor to achieve a very high impedance, thus limiting the load on the circuit being measured. However, the FET has a pretty low threshold for maximum voltage. The manual for the probe gave the range of voltages that could safely be measured.

The circuit I will be probing is pulled up to +3V inside the IBM 1130 and then discharges to ground to set a bit into the storage buffer register. A capacitor is involved in the circuit thus there is the possibility of overshoots so I wanted to be quite cautious. 

The probe itself can tolerate levels of no more than +/- 0.6V, a 1.2V p to peak and with a DC offset built into the probe can be attached to lines that are +/- 5.6V, no more. It comes with two attenuators - a 10X and an 100X - that increase the max swing and offset. The 100X can connect to a +/- 200V line and record up to 120V swings, while the 10X has a max level of 6V and swing of 12V and a max +/- volt limit of 56. 

The probe can be put in AC or DC coupling mode, but even in AC mode the swing is going to be 3V or perhaps a hair more. I needed at least the 10X attenuator to protect the probe. 

The advantage is that the attenuators cut loading to the circuit being measured. The probe alone has an resistance of 100K ohm and 3 pF, but the attenuators bring the resistance up to 1M ohm and capacitance down to 1.5 pF. 

SET UP THE PROBE WITH 100X ATTENUATOR FOR MAXIMUM SAFETY

I did my first measurements using the 100X attenuator.  I hooked it to the backplane pin where my signal arrives to set the flipflop for B register bit 14, then set another probe on the scope to a pin for the detection of a parity error. This would give me the trace on my signal at the time that the bit failed to set and the 1130 stopped with a parity check. Once the Tektronix power supply for the probe finally arrived, I was able to attach it to the 1130 and record some data. 

I had the machine set to loop through the memory of the 1130 continually reading the data. In the past, after a few hundred thousand reads I would see bit 14 fail to set when the 1130 MRAM board is trying to set it as a 1 value. This triggers a parity check since the parity generated by the MRAM board is based on all the 1 bits, but the 1130 is missing one so that it finds the parity does not match.

The machine ran flawlessly. Just like with the ordinary scope probe attached. I detached the active FET probe and the machine then took some parity checks. This time it is dropping bit 2, not bit 14, but it is on about the same frequency as before and just as predictable. Just . . . not measurable.

When I force it to record a trace, the pulses look reasonable. The signal drops sharply to ground, then rises back to 3V. There is a small uptick in the middle of the low point, but it might be an artifact because I didn't have the shortest ground path for the probe.

I am going to experiment with the 10X attenuator and make some other changes, hoping that I can get a failure while I am watching. I sure hope that I didn't spend many hundred dollars for the active FET probe and will still fail to observe the problem.