Download data from FI5 to do comparative analyses

[pattacini]

We need to download experimental data from FI5 in order to be able to carry out extensive analyses to validate the device and the simulation predictions.

@marcoaccame can this be done relying merely on the debugger or do we need to design some sort of rudimental acquisition layer?

[pattacini]

@marcoaccame reported the following:

The serial bus works at 112500 baud and thus can deliver roughly 12 bytes at 1 kHz. As of now, the serial is working in blocking mode; as a result, we cannot acquire lots of signals in each cycle. Need to think about that.

Debugger throughput is higher and we already have a way to print. Anyway, testing is required.

[pattacini]

12 bytes at 1 kHz means that we should have 120/4 = 30 float that we might be able to download at 100 Hz. Very likely, we'll need to acquire fewer data.

That said, we may consider also storing data onboard to then retrieve them later. In fact, we don't require to dump data necessarily in real-time.

@marcoaccame what is the available onboard RAM?

[marcoaccame]

Hi @pattacini, we don't need to buffer data in RAM.

We have two solutions:

• use the trace port, which I have already used to print data at every 10 ms. it is textual data which i can format as i like.
• use the usart with an increased sped. i have just tested a speed of 20x=112500 and I read correctly w/ putty.

In following days I will write and test a proper data dumper.

[marcoaccame]

just committed code for performing data log: https://github.com/icub-tech-iit/ventilator-FI5/commit/979b7d7a203667c4bb624f9fec48b8f96b6cc722. it is possible to use either trace port of usart @ 4mbps. tests + data collection: tomorrow.

[pattacini]

Analyses carried out in #60 confirm that this acquisition methodology is effective and up to the task.

@marcoaccame we can close this, I reckon.