Speed of values being displayed may be not realistic?
nicolaskolbenschlag opened this issue · 3 comments
Actual behavior
This came up when considering the unit of the time-sweep indicator (#151). But is has to do with time-sweep only indirectly. If this is an issue, it does not come from the time-sweep itseld but more from the overall way of how data is processed and displayed in the frontend.
Expected behavior
Other oscilloscopes might (as I understood) display new data in their own speed, meaning a new div (pixel, respectively) in a fixed amount of time (unit: [time/div]). Also, Wieland said that our time-sweep should be unit time/div. But currently that is not what our oscilloscopse (and by this also our time-sweep) does.
Additional context
Changing this would require changing the whole way of how data is processed in the frontend (I guess).
Leon and me compared our ideas and agreed on how to proceed:
- Where we stand:
- The speed in which our oscilloscope plots in dependent on the frequency in which it receives packages
- This receive-speed is dendent on the sending frequency of the backend (which is the generator's frequency divided by 200)
- Hence, the display speed is determined by the backend. It is of unit "div per update" and not "time per div" (as it is supposed to be)
- Workarounds:
- Obtain the time per update and compute the time per div, as we know how many divs refer to one update (namely one pixel per update times time-sweep)
- Our ideas differed in how we obtain the time per update:
Option 1: Use a constant as we know the frequency of our generator (drawback: more unrealistic as it only works for our generator and no other "sensor")
Option 2: Compute the time per update empirically on update (drawback: computation overhead and very laggy even if only measured at every 10_000 update or so)
- Next steps:
- Go with the first workaround option
- Maybe implement second in backend after this was refactored (#165)
- Nice-2-have: fix the underlying issue: buffer received signals in backend and let the oscilloscope determine its (fixed * time-sweep) display speed
Current display speed computation:
- Expect number of updates from generator/backend per second (as constant = 10_000)
- From this, the overall width (1000 pixels) and the number of divs (=20), we can compute the expected time per div (without time sweep):
- time_per_update = 1 / 1000 s
- pixels_per_div = 1000 / 20 = 50
- time_per_div = time_per_update * pixels_per_div = 50 / 1000 = 0,005s
- This is divided by the delta variable derived at time-sweep calculation
- and finally converted into ms/div
The displayed sweep speed still not matches the actual sweep speed.
The indicators display 5ms/div (with 1 div is one grid block)
The real sweep speed feals more like 1 sec for 2 div, but that's not an correct measure, more like gut feeling