Author Topic: Hackaday Superconference 2017 - Mike Harrison  (Read 859 times)

0 Members and 1 Guest are viewing this topic.

Offline Jope

  • Regular Contributor
  • *
  • Posts: 75
  • Country: de
Hackaday Superconference 2017 - Mike Harrison
« on: November 12, 2017, 03:39:22 am »

Offline ruairi

  • Frequent Contributor
  • **
  • Posts: 298
  • Country: us
Re: Hackaday Superconference 2017 - Mike Harrison
« Reply #1 on: November 12, 2017, 04:41:13 am »
I was there today and really enjoyed Mike's presentation.

My other favorite of the day was Kristin Paget's "IoT Security: A Study of Failure".

Offline TMM

  • Frequent Contributor
  • **
  • Posts: 437
  • Country: au
Re: Hackaday Superconference 2017 - Mike Harrison
« Reply #2 on: November 12, 2017, 10:02:13 am »
Manchester code only has two frequency components huh? Manchester has no DC, but the spectrum still extends down to DC. If your data is repeating at say 100Hz, then bam you have a 100Hz component in your signal. It will be attenuated compared to the NRZ signal but it's still there. You can add filtering after the manchester encoder if you need to bandwidth limit but if you put a 20kHz brickwall highpass filter on 20kbaud manchester you will destroy the data integrity as there will be strong repetition at 10kHz (2bits), 6.67k (3bits), 5k (4bits) etc. The phases of these 10k, 6.67k, 5k etc components are constantly changing with the data so instead of having peaks at discrete frequencies you end up with a broad frequency spectrum.
For that reason you'd probably use >100kbaud data rate and then implement a steep 20kHz high pass filter to avoid interference with <20kHz systems. When you look at the data in the time domain after the brickwall filter it will seem to have some low frequencies added to it - this is actually the low frequencies that the brickwall filter removed. This can be dealt with by a simple circuit that recovers the 'DC' level (mid point between high and low) so you can stick the bandwidth limited signal into a comparator and 'recover' it to a logic level signal. If you try to highpass 50kbaud manchester at 20khz the effect will be bad. If you highpass 10Mbaud at 20k it will be barely noticeable and you can probably use the signal without needing to 'recover' the lost components. Ultimately you are still throwing away data by highpassing so you'll have increased bit error rate. It's up to the designer to decide where to draw the line between BER and limiting your transmission bandwidth - or in the case that your system must have a limited bandwidth, where to draw the line between data rate and BER.
« Last Edit: November 12, 2017, 10:34:18 am by TMM »

Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo