Abnormal lick detection in Bpod

malda

Hi, recently I noticed that occasionally many licks are missed or have abnormal onset and offset times (for example, onset and offset of a lick signal differs by 0.1ms) on Bpod. I confirmed that this happens even when the actual outputs from the lick detector are apparently normal (which I measured by splitting the outputs using BNC T adapters and recording them on DAQ analog channels). This issue seems to randomly come and go (plugging and unplugging cables and turning things on and off sometimes fixes the problem and once it is fixed, usually the Bpod lick detection works fine for some time) and currently I always check whether this happens before and during recording. But, clearly this is not ideal. Do you have any idea why this is happening? Just to be clear, I am using a Bpod state machine r2, and 5V digital output from a lick detector is routed to S and GND sockets of the interface board. Another potentially related question is what happens when the interface board's S and GND sockets receive a voltage of ~ +- 10V. At one point, I used an apparently abnormal power supply for the lick detector and the lick detector outputted an oscillatory signal of about 10 V amplitude. I was told that such a signal may very well damage Arduino circuits which I believe the Bpod is based on. Do you think this could have damaged either the interface board or Bpod? Except for the above issue, Bpod seems to be working fine.

Josh

Hello Byungwoo,

The 'S' input doesn't directly interface with an Arduino pin. It is routed to a TI CD4050DB level converter. The level converter can handle voltages higher than 5V at the input, but only if its power supply pin has a greater voltage. The datasheet isn't clear about what happens if the input voltage is higher than the supply (e.g. in the Absolute Maximum Ratings section). It's probably worth trying a different behavior port channel, which maps to a different buffer circuit on the chip, or a second state machine if you have one.

The state machine samples its inputs and updates its outputs every 100 microseconds, so a sub-200ms voltage spike on the input line may appear as a Port1In followed 0.1ms later by Port1Out. To see this reliably on your DAQ's analog channels, you'd need the DAQ configured to sample at >20kHz.

A few questions:

Is the S pin still driven by the Janelia Flintbox detector?
What power supply are you using, and what else is connected to the state machine?
When you connected the BNC T adapter, what wiring did you use to attach it to the port interface to view the signal?

Thanks,
Josh

malda

Hi Josh,

Sorry I just noticed that you replied. I have now set up the notification so I can know when you reply.

I am pretty certain that these 0.1ms spikes and missed lick signals only occur on the Bpod side because the simultaneously recorded signal on the DAQ board (through the BNC T adapter) doesn't have the spikes and instead show beautiful lick signals. I believe that the sampling rate on the DAQ board is 50kHz so if the spikes were there as well, I should have seen them. Also note that the lick signals I see on the simultaneously recorded signal on the DAQ board were absent on the Bpod side, which couldn't be due to the sampling rate.

I have a few follow-up questions.

Is there any chance that the high voltage could have damaged the interface board? Could this issue be due to some malfunction of the interface board instead of the state machine?
If the damage has been done to the state machine due to the high voltage (but this damage must have been pretty subtle because the Bpod works most of the time now), would it affect other behavior port channels as well? I'm hesitant to buy a new state machine just for the purpose of testing because of its cost.

To answer your questions:

I am now using a custom designed detector. It outputs 5 V when there is no contact between the lickports and the animal and 0 V when there is a contact.
I am using a regular 5V power supply for the lick detector. The state machine is connected to the two interface boards which are connected to outputs of the lick detector, a Bpod HiFi Module HD, and computer.
The lick detector has a BNC output port for each of the two lickports. I attach a BNC T adapter directly to it and then the signal is split and routed to the DAQ board (through the regular BNC cable) and to the interface board (through the BNC to regular wire adapter).

Josh

Hi Byungwoo,

It's possible that damage was limited to only one channel of the CD4050DB, but equally possible that all channels were affected. The only way to know is to try another channel.

Which DAQ board are you using?

To debug, I'll need to see the schematic of your custom lick detector (and circuit board layout if it has a PCB).

Thanks,
Josh