[TenTec] RF speech processors

Merle Bone merlebone at charter.net
Tue Feb 3 19:41:25 EST 2009

I ran tests on the Speech Processing effectiveness of the V1 and V2 code from the Orion on 29 May 2007. The results are below. It will be interesting to see how the results could be any better with an external compressor - of course you can turn down/off the internal Orion compression, greatly reduce the mike gain or even change the speech compression code.

Doug Smith, the major author of the V1 DSP code, wrote the following about the speech compression in the Orion:
"It might seem funny but in a DSP transmitter, it is relatively easy to compute the transmitter's RF envelope before the modulation is even performed! That, in turn, makes it possible to preprocess the audio applied to the modulator to get exactly the same effect as that produced by an RF compressor. See Fig 5. Post-modulation band-pass filtering maintains the desired occupied bandwidth. As the decay time of the compressor is decreased, an RF compressor approaches the behavior of an RF clipper, long known to be the most effective form of speech processing for SSB. In combination with the Orion's transmit equalizer, results are quite dramatic.
The reason RF compressors and clippers are so effective is that they increase the average power transmitted. Human voices tend to have high peak-to-average power ratios-- as high as 15 dB. That means a transmitter whose peak envelope power is limited to 100 W may achieve an average power of as little as 3 W. Under such conditions, the Orion's RF compressor may add 10 dB or more to the average power, equivalent to a ten-fold increase in output power. The speech processor achieves this goal without introducing the kind of distortion that harms intelligibility, such as that created by heavy audio compression or clipping.

Text of Previous message on Orion SP testing:
Text from previous note  http://lists.contesting.com/archives//html/TenTec/2007-05/msg00478.html :
Although the Orion V2 code still has some problems/issues with performance, one 
area that appears to have been addressed in the V2.06X releases is the speech 
processing performance. Earlier versions of the V2 code suffered from a 
combination of audio distortion and weak, if any, compression. The V2.06X 
releases clearly sound like the distortion has been addressed and that the 
compression is much improved. I recently bought an LP-100 Digital Vector RF 
Wattmeter from Larry (N8LP) http://www.telepostinc.com/lp100.html . Among the 
many functions provided in the LP-100, Larry recently added new code to allow 
the user to measure the "effective audio compression" of a transceiver. The 
technique uses an audio test tone pattern ( the one I used had a 1KHz tone with 
alternate 50MS time segments with the levels 20db apart).
The RF power meter samples the power level 40,000/second to generate average 
value. I used an "audio sound power meter" to measure the audio level of my 
voice calling DX and then used the sound power meter to set the test tone to 
the same level (Using a single level tone) into the microphone.

I tested the V2.062a code and got the following results:

Proc Setting - Audio Result - RF Result Average Power
Off          Slight Expansion           ~8W                   
1-4          Close to no compression    ~9-10W
5            ~5dB Compression           ~12W               
6            ~5-10dB Compression        ~13W
7            ~10dB Compression          ~16W                      
8            ~10-15dB Compression       ~20W
9            ~15-20dB Compression       ~26W 
I tested the V1.373b5 code and got the following results:

Proc Setting - Audio Result - RF Result Average Power 
Off          Slight Expansion            ~8W                        
1 - 4        Close to No Compression     ~9-10W
5            ~5dB Compression            ~12W              
6            ~5-10dB Compression         ~14W 
7            ~10dB Compression           ~17W                       
8            ~10-15dB Compression        ~24W                  
9            ~15-20dB Compression        ~31W                 

In all cases the peak power stayed very close to 39W - which is what I use to 
drive my Alpha 87A to 1500W peak output.

As you can see from the results they track very closely - so closely that they 
almost look like the same algorithm! There is a slight difference at the 9th 
level on the Speech Processing but it is very minor (And might be altered if 
the mike gain were increased). At a Speech processing level of 7 the talk power 
is doubled from the level with the Speech Processing off. At the 9th level 
there is roughly a 6dB (1 S-Unit) improvement in the talk power of the rig. I 
usually stop at eight but there is no audio "background noise level increase 
even at 9 (Maybe a very little hum). The audio sounds very good in the monitor 
and the quality was verified by a friend. The talk power improvement is very 
important in trying to work SSB DX.

I think the Orion V1 code has the best Speech Compression I have ever seen in 
an HF radio and now it appears that the V2 code matches it! Thanks Ten-Tec!
73, Merle - W0EWM

More information about the TenTec mailing list