FreeDV

Mon Jan 28 21:38:24 CST 2013

Alex,

First, that is a good link to wikipedia on crest factor.  Indeed, that 
is the same thing as peak to average ratio.

If we calculate the value for 14 or 16 tones we have to back off the 
power amplifier to a really low level to insure the waveform is never 
clipped.  A good way to check it when starting out, a good oscilloscope 
can show the transmitted waveform and as you raise the power you can see 
the clipping start.

If it was a hifi amplifier, you would want to avoid any clipping. For 
the digital voice signal, clipping will cause individual tones to cause 
trash onto the other tones.  That trash is a form of cross talk and 
quickly degrades the digital error rate.  As I said before, a 1 KW 
amplifier was found to work over HF circuits at about 110 watts average 
power.  That was with 16 tones but would be very similar for 14 tones.

Another problem is that the center tone is used for sync.  The trash 
from clipping can spoil sync so we must also take that into account.

In the end, with the 706mk2G I would run it around the same ratio which 
would be about 11% or 11 watts average power.   I want to get two FreeDV 
terminals running in a lab setup and see how this particular signal 
works with clipping and signal to noise ratio. Right now I am only able 
to speculate based on prior experience.

I did see at one point where one of the FreeDV authors said his system 
was working well at 30 watts average.  That seems high but maybe it is a 
good setting.  I want to do some tests and see a scope of the waveform 
as it goes into clipping.

On the Navy systems we designed and tested on long haul paths we 
observed the digital voice would work when we could not hear the remote 
signal coming in and I was really impressed.  We attributed it to 
processing gain from the vocoder.  FreeDV should do similarly with the 
LPC rather than the vocoder but that waits for us to get some more 
experience.

Frank K0BRA

On 1/28/13 3:47 PM, Alex Fraser wrote:
> Yikes!
>
> OK I had to look it up http://en.wikipedia.org/wiki/Crest_factor
>
>     I read the Wikipedia page and followed some links, but haven't 
> developed an intuition for this concept yet.  You seem to understand 
> it so I will, if you don't mind ask a couple of questions.
>
> - I have a Icom 706mk2g.  It runs IIRC about 100 watts all out. For 
> other digital modes on HF I have been using the "3" setting on the 
> menu which I figure gives 30 watts out.  Is it safe to use this 
> setting with freeDV?  Should I use more power?
>
> - The rig seems to give more power in the lower bands.  Would I be 
> able to put out a better DV signal on the lower bands?  If it is 
> better output wise on the lower bands, would this increase match or 
> better the increase in the noise?  Is there a sweet spot?
>
>
>
> fgentges at mindspring.com wrote:
>> The transmitted signal of FreeDV is 14 tones.  Each time you double 
>> the number of tones you double the peak to average ratio of the 
>> transmitted signal.  The result is that this type of multitone signal 
>> with its high peaks will overload the RF linear amplifier.
>>
>> Some years ago we found that with a 1 KW amplifier we needed to run 
>> about 110 watts average power.  That value still had some clipping 
>> but gained enough on the signal to noise such that it worked better.  
>> When you start with a 5 watt radio you will have to run at a really 
>> low level and not much of anyone will hear  you.  We searched for 
>> some way to improve on that and short of building an amplifier good 
>> for huge peaks, there is not much you can do.
>>
>> One fellow worked to improve the phase angles of the tones but 
>> nothing much came out of that effort.
>>
>> Frank K0BRA
>>
>> On 1/27/2013 7:33 PM, A. Maitland Bottoms wrote:
>>> Alex Fraser <beatnic at comcast.net> writes:
>>>>          I  heard my first freeDV stations on
>>>>        14236 (Saturday 4:20pm East Coast United Snakes).
>>>
>>> I wasn't active Saturday, but this (Sunday) afternoon there
>>> was more FreeDV there. I heard N9NJY's first few DV contacts.
>>>
>>> The people I am hearing have better antennas and more power
>>> than I do. The good signals come from 20m Yagis atop towers
>>> and 35 to 80 W RF power.
>>>
>>> My little Yeasu 817 and non-resonant dipole do not seem
>>> to attract attention. I took the opportunity today to
>>> swap out the CW mechanical filter for the SSB filter, but
>>> so far other than darker edges on the waterfall I don't
>>> notice a difference in decoding ability. I'm guessing
>>> my power out is about 2W (I've backed down the drive
>>> a bit in search of linearity, CW is about 5W on the rig.)
>>> How much power actually radiates is anyone's guess.
>>>
>>> I do have a good start in my PTT software mod...
>>> The interface I am using is a DMK engineering USB
>>> radio interface. Like other C-Media sound fobs, it
>>> handles the audio as a USB audio device and has
>>> GPIO pins that can be toggled with the USB HID
>>> interface. The cable from the interface plugs into
>>> the PS2-like Data connector on the 817 for a neat hookup.
>>>
>>> Good Luck!
>>>
>>> 73 de aa4hs,
>>> -Maitland
>>> _______________________________________________
>>> Tacos mailing list
>>> Tacos at amrad.org
>>> https://amrad.org/mailman/listinfo/tacos
>>>
>> _______________________________________________
>> Tacos mailing list
>> Tacos at amrad.org
>> https://amrad.org/mailman/listinfo/tacos
>>
>
>
> -- 
>     \\\\\\\\\\\\\\\\-----++++*0*++++-----//////////////////
>         No electrons were harmed in the creation of this message
>           --------------------------------------------------------
>   ~~~********************Alex Fraser********************~~~
>           --------------------------------------------------------
> [[[[[[~~^^^#___=>>>```/\/\**O**/\/\```<<<=___#^^^~~]]]]]]
>
>
> _______________________________________________
> Tacos mailing list
> Tacos at amrad.org
> https://amrad.org/mailman/listinfo/tacos

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://amrad.org/pipermail/tacos/attachments/20130128/8a72513b/attachment.html>