[RTTY] 300hz or 500hz IF filter?

[Kok Chen]

On Aug 25, 2013, at 6:16 PM, Joe Subich, W4TV wrote:

> Chen, it would help if we knew how much group delay can be tolerated
> in the filters.  

Joe, you already have seen those.  Perhaps, you just don't realize it.

Just go take a look at the flutter curves at VE3NEA's site.   You can then relate the path delays back to filter group delays.

http://www.dxatlas.com/rttycompare/

Do you notice that some modems asymptotes out (i.e., error rate stops improving as SNR improves)?  That is the effect of ISI caused by multipath (relative delays between Mark and Space bits).

The irony is that MMTTY does not asymptote like TrueTTY and MixW.

The explanation is simple -- MMTTY does not use any filter that is close to either a Raised Cosine or a Matched Filter.  It is wide as a barn door in DSP terms.  So basically, while MMTTY filters have terrible SNR characteristics (as shown in the AWGN curves), it also does not ring as much because it is so wide (but not infinite HI HI).

Remember what I wrote about Raised Cosine and ringing?  While MMTTY does not place exact zeros at the mid-bit locations, the ringing is low, so that while the ISI may be non-zero, it is also much lower than multipath through a Raised Cosine.

How much multipath is in VE3NEA's plots?  Alex had used the CCIR Flutter Curve (the HF model comes from the CCIR 520 document that was the precursor of the ITU 1487 document that has much more comprehensive HF Channel models).

Those few curves in VE3NEA tell a whole lot of story -- if you know how to read them.

The delay between two propagation rays in the CCIR Flutter curve is only 0.5 ms (with a Doppler spreading of 10 Hz though; so that also required a wider filter to pass the fluttered signal).  And look at how destructive that profile is already to demodulating RTTY -- it pretty much limited TrueTTY to about about 1 error out of 200, and limited the performance of MixW to one error in 50 characters *no matter how loud the station is* over the noise floor!! 

MMTTY on the other hand is capable of producing 1 error out of 1000 if you give it a strong signal with good enough SNR.

I agree, Joe, that the proper thing to do is to substitute a model of a crystal filter in place of a model of the HF channel (it is really quite simple to do).  I probably will never do it though, since I have moved away from using superhets, and have no need to use any crystal filter at all in my receiver chain.  Perhaps we can persuade Andy K0SM to do it?

73
Chen, W7AY