Author Topic: Gaussian Shaping  (Read 1160 times)

jrdoner

  • Newbie
  • *
  • Posts: 34
Gaussian Shaping
« on: November 26, 2015, 10:16:55 PM »
I am happy to say that I have waded through the process of getting Moteino's to talk to each other in a friendly manner.  But I have some questions about parameters.

I dumped the registers, and it appears the radio is set at 100 Kbs, using FSK, no Gaussian shaping. 

1. Is this a correct reading of the registers?
2. Is there any disadvantage to using a higher bit rate?
3. The registers are apparently non-volatile?
4. Gaussian shaping can narrow the bandwidth of the transmissions:  is there any downside to turning it on?

Thanks in advance for your response.

Felix

  • Administrator
  • Hero Member
  • *****
  • Posts: 5905
  • Country: us
    • LowPowerLab
Re: Gaussian Shaping
« Reply #1 on: November 26, 2015, 10:28:23 PM »
The registers are volatile, but have power up default values, i believe the bitrate is 2.4Kbps. See the datasheet for more info on registers.
If you're using my library, I set the defaults to 55.5Kbps. I have some sample code of reading the registers, see here.

Higher bitrate means lower range as an end result. It's like talking really fast, you have to be pretty close to get a better chance to understand what's spoken.
Not sure what to advise about Gaussian shaping, I am not using it in my lib.

jrdoner

  • Newbie
  • *
  • Posts: 34
Re: Gaussian Shaping
« Reply #2 on: November 26, 2015, 11:07:50 PM »
Thanks, Felix,

I appreciate the info.   Just for grins, I'll do some range testing with shaping turned off, and shaping turned on, and get back to you. 

emjay

  • Full Member
  • ***
  • Posts: 120
  • Country: nl
Re: Gaussian Shaping
« Reply #3 on: November 27, 2015, 12:44:15 AM »
@jrdoner,

Shaping is useful when you are using a low modulation index (m<2) and wish the minimise the transmitted band occupancy.  Neither is true for the typical Moteino set up.
Nothing is for free, aggresive shaping introduces inter-symbol smearing, so in general you need a higher S/N ratio for the same bit error rate - which maps directly into reduced range for the same RxBw setting.

Note that if you are trying to test this out, you need to get the Rx signal way down close to the noise floor - if not, the chip AGC will adjust the LNA gain to fit what the demodulator section wants, as a side effect, also changing the effective S/N ratio and confusind the observations.