Re: FT8-Call HF Path Tests

Ev Tupis
 

Hi Tony,

Thanks for the extra time and effort you dedicated to "deep diving" to explore the foundational reasons that the authors attribute to the technical approach taken.

To summarize, they attribute the "magic" to...

A. narrow-band modulation
B. more computing time to extract the payload (mentioned twice)
C. enhanced message-bit density
D. optimized decoding algorithm
E. abbreviated payload size/length (mentioned twice)

Taking a step back, it appears that the magic really occurs in the computer and in one more place not mentioned, in my opinion.  I'll get to that in a moment.

Items A. and E. have been applied to short-duration long-distance pathways for decades at "baseband" HSCW modulation schemes and C. as a version of data compression applied by APRS in a specialized meteor scatter mode.  In fact, they exploited the "time domain" as well...TXing for longer-than-typical periods of time to assure maximum exposure to the refracting medium.

I commend the authors for leveraging such prior art and learnings in their product.

That leaves B. and D. as the "magic".  "Augmentation through computer post-processing" appears to be where the "magic" happens.  One wonders if similar results would have been possible had B. and D. been applied to any existing keyboard mode, but I digress.

"Perception" is the other place where "magic" happens.  Andy summarized it well.  With the backing of a Nobel Laureate, these modes have captured the imagination of a critical mass of operators who are "warming the aether" and providing many, many signals to observe and communicate with.

I can't help but wonder if this gives a distorted impression of the raw effectiveness of the mode.  A question that can be answered only by placing it side-by-side with another such as PSK31 (or an even narrower version) that applies similar compression to the payload.

Anyway...thanks for the additional effort and information!  You've helped me to become significantly more clear on the approach!

Again, I commend the authors.

Cheers,
Ev, W2EV





On Tuesday, September 25, 2018, 2:53:55 PM EDT, Tony <DXDX@...> wrote:


Ev:

On 9/25/2018 5:53 AM, Ev Tupis via Groups.Io wrote: Hi Tony, as for a slow data rates allowing more computing time, I doubt it

I emailed Steven Franke, the co-developer of FT8, about the robust narrow-band modulation as being a factor in the modes ability to decode in the noise and if the code did most of the heavy lifting. He said that I identified two key elements and added a few more.

The first was that the efficiency of FT8's encoding minimizes the number of message bits conveyed by each message: this maximizes the transmitted (and received) energy-per-message-bit. The second wast that the decoding algorithms are such that they can decode with extremely noisy received frames.

He went on to say that the ham radio application is unique in that there are loads of time to dedicate to decoding each short message which allows hundreds of milliseconds of computer time to work on recovering the relatively short 77 message bits.

He also said that in the case of FT8, there is some gain had from using a decoding algorithm that would take an impractical amount of computer time if deployed in a high-date-rate communications system.

The explanation regarding computing time, data rate and the decoding algorithm is why I assumed the FEC had more time to work on decoding. At least something along those lines. 

Hope this helps.

Tony -K2MO

Hi Tony,

As for a slow data rates allowing more computing time, I doubt it.  The computing time is a function of how long the previous TX interval is allowed to continue.  Computer-processed/augmented, delayed decoding would make any data mode more robust.

It's that sort of conjecture that leads me to ask what makes the mode "Magical".

Ev, W2EV






On Monday, September 24, 2018, 9:25:25 PM EDT, Tony <DXDX@...> wrote:


Ev:

I think the use of a robust modulation type that concentrates it's energy into an efficient narrow-band signal plays a big part. Being able to detect the signal at low s/n ratios gets you so far so I suspect the heavy lifting must come from the code.

I've read that the slow data rate allows more computing time to be spent on correcting errors which helps to decode further into the noise.

There's probably a good reason why they didn't chose PSK modulation. Maybe it's sensitivity to phase distortion would restrict it's a ability to perform as well as MFSK.

Tony -K2MO  

Join main@digitalradio.groups.io to automatically receive all group messages.