“Optimal Combining” Diversity ?

  • 2
  • Question
  • Updated 1 year ago
  • Answered
  • (Edited)

I ran across a couple of interesting discussions started by W5UN and K1UO on diversity reception.   Below are the links and excerpts from Steve’s (N5AC) replies:


 “The situation is this -- not only is the FLEX-6700 capable of what I would call "antenna diversity with human decoder" where you stick two receivers that are locked, one in each ear (a.k.a. poor man's diversity), the FLEX-6000 has the necessary hardware to support MRC/optimal combining.  Poor man's diversity is essentially a form of what is called selection diversity where two ears are used to actively select the best of the signal at any point in time. This is how classic voting receiver systems worked on repeaters as implemented by GE and Motorola in the day. With selective diversity, two signals with different SNRs are presented to a voting system (electrical or your ear) for the selection of best SNR at any point in time.” 

“In contrast, MRC/optimal combining is different in that DSP is used to look at the correlated signals (the ones you want) and non-correlated signals (noise) and then subtract out the noise. This nominally provides a 3dB SNR improvement on the signal you give to your brain. This is really a big difference -- it can essentially lower the noise floor by 3dB at your location. Again, this can be 3dB better than the BEST of the two signals you get off the two antennas which is not possible with selection diversity.”

“Additionally, we have a mechanism in the FLEX-6700 to take a single antenna and run it to two data converters.  This is also another form of diversity, but it aims to lower the noise floor of the data converters.“ 


 “Two of the diversity forms we discussed -- the linked slice or simple diversity method and the optimal combining method -- made the cut. The first method we plan to do in v1.3 and the second is more of what we call a science experiment. In other words, we want to do some research and testing to find out what works best before we commit to a date. When we do this, we generally do not put the item on the roadmap because we do not want to have to come back and say "well it doesn't look like it will make xyz release after all." Having said this, we have already started work on optimal combining and the quicker that work comes to conclusion, the quicker you will see it in the software.”

Assuming that the current implementation is the “simple diversity” version, is MRC/optimal combining, and/or routing a single antenna to 2 data converters to lower SNR, still in the works, even if in the science project stage? 



Photo of HCampbell  WB4IVF

HCampbell WB4IVF

  • 128 Posts
  • 32 Reply Likes

Posted 2 years ago

  • 2
Photo of Steve - N5AC

Steve - N5AC, VP Engineering

  • 1003 Posts
  • 965 Reply Likes
Official Response
Howard, we continue to work on a number of different projects like this in the background.  Discerning the best SNR in a digital signal is fairly easy since you have a quality indicator, the bit error rate (BER) and you can adjust for lowest BER.  The problem is more complex, for example, if you have a 500Hz bandwidth and there are several CW signals in that bandwidth.  Which should be optimized?  What if the optimization results in a null on one of the signals?  Having said this, we are still experimenting with algorithms in between the other research we are doing.

One thing I'd like to add to the discussion is that using the two ear method (called simple diversity before) has a unique property of spreading signals around in the sound stage.  We have heard from avid contesters that in a pileup this allows the brain to better sort through the signals, just like focusing on one person at a cocktail party in a noisy room.  You would lose this capability in a situation where the radio did the diversity combining for you, but in some cases this would be better.

I mat a cognitive researcher this last weekend and I got a chance to ask some ham radio relevant questions. I asked about the block that some folks have around CW and the inability to go past a certain speed.  I suggested that I believed there was a translation occurring in the cognitive process -- going from a sound to mental pictures of dots and dashes to a lookup table for CW letters to the letter.  She made a couple of very interesting comments.  First, she said that the brain pathways for sound are almost immediate and require very little in the way of cognitive horsepower.  If you can go from sound to meaning, you can do things much faster.  The second thing she said that caught my attention was that visual memories are more deeply lodged.  She said once you learn a visual memory, it's more difficult to shake or overwrite that.  The real lesson here is that if you want to teach someone the code, NEVER show them a chart of dots and dashes!  But in relation to this topic, it made me realize that the brain is so good at sorting out sounds in a sound stage that while simplistic, the simple diversity takes advantage of lots of horsepower in your noodle.

So we're still pursuing the optimal combining technique and hope to have something in the future, but I think based on contester experience and what I've learned about cognitive processes that the simple diversity method is useful by itself!