In our case we just assume that the last PPI sample comes from current instant, and samples before that are x ms before that e.g. if we get a batch of [300,500,900] ms then they are at [t-(300+500+900), t-(500+900), t-900] where t is now

Thanks for the reply, let's assume this is not an offline recording and there is no startTime provided by the device(or SDK).
What is now in this case:

1. the time of receiving the first measurement
2. the time of sending the command to the device to start the streaming?
   In my case with OH1 there is a difference between those two points around 3-5 seconds.

I'd assume it's more likely that time of receiving the measurement is closer to reality.

I suspect the algorithm that determines the PPI values requires several cycles to establish a baseline. When switching to PPI mode you can notice that HR stream stops for a while and starts over.

@orestesgaolin is correct. I think it's even 25 seconds to receive the first sample after the start command, which contains 5 second windows of PPI samples. The algorithm inside Verity Sense doesn't calculate any timestamp for the samples and this is why it's not provided for PPI data for Verity Sense.

Some other devices such as Polar 360 do have timestamps attached to PPI data as they have more recent algorithm that provide it.