Built-in Self Tests (BIST)

Topics:

Overview

A number of self tests are implemented in the recording system to debug combinations of subsystems.

The signal and noise tests are qualitative; see the 'Noncritical Failures' section for better ways to evaluate probe performance.

Test Functional Area
Base Station PC -> BS (slot) connection
Heartbeat Probe -> BSC FPGA connection
SerDes Link HS -> BSC communication
I2C/Memmap Probe configuration interface
EEPROM Part number/version data storage
Shift Registers Shank and base configuration network
Parallel Serial Bus Probe -> BSC data connection
Signal Path Signal fidelity check
Noise Level Amplifier noise check 
Terminology:
    HW     = hardware
    SW     = software
    PC     = computer
    FPGA   = microcomputer
    I2C    = inter-IC bus
    BS     = base station card
    BSC    = base station connect card
    HS     = headstage
    PSB    = parallel serial bus
    POR    = power-on reset
    PRBS   = pseudo random binary signal
    SerDes = serializer/deserializer

The Tests

Base Station

This test checks the connectivity between PC and BS FPGA board over the PXIe interface.

Heartbeat

The heartbeat signal generated by the PSB_SYNC signal of the probe is routed via the SerDes to the BSC FPGA. The BSC analyzes the presence of a 1 Hz clock signal.

The presence of a heartbeat signal indicates the functionality of the power supplies on the HS for serializer and probe, the POR signal, the presence of the master clock signal on the probe, the functionality of the clock divider on the probe, and basic communication over the SerDes link.

The functionality of the SerDes link is quantitatively verified using the integrated PRBS test of the SerDes component pair.

I2C/Memmap

This test verifies the functionality of the I2C interface and the probe's memory map. The function reads a register on a memory map address, writes back a modified value, reads this again, and writes the original value again to restore the probe to its settings before test.

The test is successful if the modified value can be read back, and if an ‘Ack’ byte has been received for each R/W operation.

EEPROM

The test verifies the correct access to the EEPROM on flex, HS and BSC. The function writes data to an unused location on the EEPROM, and reads it back.

The test is successful if the data can be read back, and if a ‘Ack’ byte has been received for each R/W operation.

The function tests the flex, HS and BSC EEPROM consecutively. The function returns an error after the first failed test.

Shift Registers

This test verifies the functionality of the shank and base shift registers. The function configures the shift register two times with the same code. After the 2nd write cycle the SR_OUT_OK bit in the STATUS register is read. If OK, the shift register configuration was successful. The test is done for all registers. The configuration code used for the test is a dedicated code (to make sure the bits are not all 0 or 1).

If this test fails, a shank is likely broken.

Parallel Serial Bus

A test mode is implemented on the probe which enables the transmission of a known data pattern over the PSB bus. This test records a small data set and verifies whether that matches the known data pattern.

This way the PSB bus on the probe and all data lines between probe and serializer are verified.

Signal Path

The probe is configured for recording of a test signal which is generated on the HS. This configuration is done via the shank and base configuration registers and the memory map. The AP data signal is recorded and the frequency and amplitude of the recorded signal are extracted for each electrode. If at least 90% of the electrodes show a signal within the frequency and amplitude tolerance, the function returns a pass value.

  1. The function requires more than 30 seconds to complete.

  2. This test and the noise test are somewhat qualitative. If all other tests besides {signal, noise} pass, the probe is probably alright.

Noise Level

The probe is configured for noise analysis. Via the shank and base configuration registers and the memory map, the electrode inputs are shorted to ground. The data signal is recorded and the noise level is calculated. The function analyses if the probe performance falls inside a specified tolerance range (go/no-go test).

The probe passes the noise test if fewer than 20% of the electrodes have their AP noise value out of spec, and if fewer than 20% of the electrodes have their LFP noise value out of spec.

The function returns a pass/fail value.

  1. The function requires more than 30 seconds to complete.

  2. This test and the signal path test are somewhat qualitative. If all other tests besides {signal, noise} pass, the probe is probably alright.

Critical Failures

These hardware errors may be recoverable by reconnecting or replacing parts.

Headstage Problems

Bad headstages are exceedingly rare. If you are getting errors like:

Error 44 'NO_LINK: No headstage detected on port X'

It is probably a poor connection between the probe flex and headstage. Keep playing with that. Sometimes changing the pairing between probe and headstage will help because the tiny ZIF connectors are not manufactured to super tight tolerances and sometimes the ZIF clasp can weaken and fail to close tightly onto the flex.

The next thing to suspect is a bad 5-meter cable. Sometimes the leads can wiggle loose inside the Omnetics connectors after many dis/connect cycles.

Finally, while it is very unlikely that the headstage has an electronic issue, for a 1.0 headstage, you can try running the HST dialog with the headstage tester dongle.

Shift Register Error

You should verify the probe is actually broken by trying an alternate headstage and/or cable. If this problem won't go away...

The shift registers are used to program settings into the probe. Depending upon how badly damaged the probe is, it might "run," but you have no way of knowing which electrodes are connected or which references are being used.

Future versions of SpikeGLX/imec firmware may permit the use of a multishank probe that only has one or two broken shanks, but that is not possible yet. Do not throw such probes away.

Parallel Serial Bus Error

This most commonly occurs when the probe flex is not seated well in the headstage. It will usually clear if you carefully close the clasp on the ZIF while pushing the flex squarely and firmly into the ZIF all the way.

Noncritical Failures

Signal Path Error / Noise Level Error

These two BISTs are somewhat qualitative and oversensitive. If the only thing wrong is that one or both of these two tests fails, the probe may well be fine. We suggest manually surveying the probe which will give you much better information about both the probe and the environment. Do the following:

Common Setup

  1. Run the probe in a saline bath with reference and ground connected together and to a wire placed in the solution with good conductance (Ag/AgCl, Pt). Immerse deeply so all electrodes are in solution. This is the same as the input-referred noise setup in the probe 'User Manual.'

  2. Select External referencing mode for the probe.

Assess Probe Internal Noise

Do this by applying maximum environment noise filtering:

  1. On the IM Setup tab, check Filtered IM stream buffers. This causes the ShankViewer to apply both band-pass filtering and CAR (Gbldmx).

  2. In the Graphs Window apply these Settings:

As you hover the mouse over the traces, observe stdv readouts at the bottom of the Graphs Window (stdv is effectively RMS-offset). Depending upon probe model, the specs for this are 7-9 uV, but in practice, anything below 10 uV means the probe is ok. Remember there will always be a handful of outlier channels.

  1. Open the ShankViewer to look at the whole probe. Set the following:

The AP range should be ~50 uV on this scale. Look at the uniformity of the value over the probe. Hot (bad) channels will be bright yellow or white.

You may also see some black (zero) channels. These could be dead, or they could be saturated. Check that by un/checking the -<T> filter as you observe the trace in the Graphs Window.

In any case bad channels can be masked (see below).

Assess Environment Noise

Do this by applying minimum filtering:

  1. On the IM Setup tab, un-check Filtered IM stream buffers. This causes the ShankViewer to apply only high-pass filtering.

  2. In the Graphs Window apply these Settings:

Now stdv readouts will be much larger but that's not a bad probe. It's your call how much noise pickup is OK. If the noise is greater than 15 uV you could have a poor connection in the ref or ground wire path. Check your solder joints.

This is a good time to poke the probe wires with a stick. Also poke the flex connector. The signals should be quite steady under poking. If the noise bounces when you poke then something is too loose.

Of course all environments will have their own noise sources. Very large environment noise could come from any equipment in your building. Try plugging your electronics into different wall sockets (maybe different building circuits). Note if it tracks the building air conditioning schedule. Is it worse at a certain time of day? Consider a Faraday cage.

Bad Channels

It is common that three or four channels on the probe have overly large noise or variation in amplitude. If a few channels look bad but the other channels look normal, then use the probe, but list its bad channels as such in the Each probe table on the IM Setup tab. This will mask them from use in CAR within SpikeGLX, CatGT and Jennifer Colonell's ecephys pipeline.

fin