[Amps] Audio processing- Part 1

[Jim Thomson]

Date: Wed, 04 Jan 2017 18:47:01 +0000
From: Manfred Mornhinweg <manfred at ludens.cl>
To: amps at contesting.com
Subject: Re: [Amps] Date: Wed, 04 Jan 2017 18:47:01 +0000
From: Manfred Mornhinweg <manfred at ludens.cl>
To: amps at contesting.com
Subject: Re: [Amps] FCC Denies Expert Linears' Request for Waiver of
15 dB Rule


> Most modern transceivers include speech processing. In the pro
> world, we use both peak limiting and compression. Peak limiting being a 
> short time constant that simply reduces gain on speech peaks, and 
> compression being more of a dynamic gain-riding. Good signal processing 
> can sound very good with up to 10 dB of gain reduction, and some systems 
> are good for more than that.

The problem with that is that 10dB of control range is far too small to 
accommodate the variations in the audio level coming from the 
microphone, as the operator moves closer or farther away, and speaks up 
or speaks softly. And to maintain the 10dB compression you first need to 
have a stable audio signal. So, in order to achieve that 10dB 
compression, you need to place this compressor after an automatic gain 
control system, that has a larger control range, and has such a long 
decay time (1 second or longer) that it doesn't cause significant 
distortion. Good speech processors usually do this.

####   Unless you talk at a constant level, and use a boom-mic, the amount
of  RF clipping  or  RF compression will be a function of how loud you talk.   The easy
fix for all this to  1st use either a noise gate... or a downward expander.  A noise gate is
just a high ratio downward expander.   Then follow it up with an audio compressor.
Then follow that up with your peak limiters and or distortion cancelled audio clipper.   Add  some
eq in there to tweak the audio for a specific application.   The noise gate will  kill any back ground
noise when not talking..and just b4 vox drops out..and also between words etc.   Done right,
its all totally transparent.  Any good quality AF  compressor  will easily handle 0- 20 + DB
of compression..and then maintain a constant level output.  Constant level output is what drives the
peak limiters / clippers.  Drive to peak limiters is then set for whatever you want, say 6-10 db. 

##  The compressor can also be configured as an AGC, with a long decay time as you noted...and also a 
slower attack time.   This is normally only done when feeding program material, which may well vary, from 
one song to the next, when used for a broadcast application.   The AGC will  slowly increase / decrease the level, 
so the listener doesnt hear any abrupt level changes.      

##  wonder how  TV commercials are much louder than the program you are trying to watch ? 
Same basic process  as described above.  If you look at  TV commercials  on an audio type
spectrum analyzer,  you will see that the average power goes up exactly 6 db..at least in my 
town.   Peaks actually all drop .5 db  during commercials.   Good and loud..and squeaky clean. 
They are not using rvs connected diodes to make an audio clipper either. Darlington connected
transistors used,  one pair for the negative..and another pair for the positive.   You can easily
obtain 3-15 db of peak limiting and clipping  with such setups.   I typ use just 6-9 db most of the 
time.   Louds of clean talk power.... and either no alc, or barely any on the xcvr.     


After such an audio processor, indeed you don't strictly need ALC, as 
long as the operator always sets up the TX gain in a correct way, so 
that all amplifier stages are kept out of saturation. But with any band 
change this gain setting will be different. Often it will also change 
with frequency changes inside the same band, and what's worse, the gain 
of most amplifiers changes with temperature, so the operator will have 
to watch the output and readjust the TX gain rather frequently, to stay 
at the optimum output level. That's quite inconvenient, and so we use 
ALC to perform that task automatically.

###  IF the PO of the xcvr stays put,  I dont have to tweak anything, when
changing bands.  Drive level to the xcvr remains constant.   I bypassed the
yaesu  audio mic  jack completely..and instead feed the line level audio from the 
rack gear into a 20 db pad, then into a balanced to unbalanced jensen audio transformer,
then  coupled to analog BM  with a 220 uf panasonic  su type cap, non polarized.   Cap
has to be there, otherwise the analog BM will completely unbalance itself.    





The manufacturers build ALC into the transmitters, as a 
non-user-defeatable feature, because they have good reason to suspect 
that most hams will not properly set the TX gain by hand all the time. 
Even more so in case of radio operators in other services, who don't 
have any technical knowledge at all!

###  9 vdc through a 50 K pot..then into the ALC input on the xcvr will
tweak the RF PO of the xcvr dead on.  No need to worry about alc time constants
or developing  ALC  voltages... after the horse has left the barn.   You have already
developed your own alc voltage externally with a 9 vdc source and adjustable 50k pot.

###  The above 9 vdc + 50k pot method has one limitation.  On a normal xcvr, with NO
external audio compression, if you whisper into the mic, you will get hardly any PO.
If  you talk normal into the mic..or even scream into the mic, the PO of the xcvr will be 
fixed at what ever level you tweaked it for.  It wont budge.   If a boomset is used, its a
non issue.    






> Audio processing done entirely at baseband creates artifacts at 
> baseband, but those baseband components won't get past the TX passband 
> filter.

Yes. Very true. But the same is true for RF speech processing, as long 
as it's done before that filter!

> W4TV has noted, however, that some rigs, notably Yaesu and ICOM, 
> do part of their processing at RF, and can splatter pretty badly.

If they splatter badly, it's because of some other reason. RF speech 
processing, done before a good filter, cannot create more splatter than 
audio processing. A good RF speech processing scheme needs a first 
sideband filter, then the clipping and compressing, and then a second 
sideband filter to remove the out-of-band artifacts. The advantage of RF 
speech processing, relative to audio speech processing, is that fewer 
artifacts fall inside the passband. So, RF speech processing should 
result in a cleaner signal, having less distortion within the passband, 
and no more crud outside the passband than AF speech processing causes.

##  traditional RF clipping is used in my FT-1000D..using 2 x filters.
The MK-V has a choice of either  analog SSB..or  DSP  SSB....and 
both choices use  the same  rf compression scheme..which is typ
a fast attack fast decay.   The 1000- MK-V doesnt use RF clipping
when in analog SSB mode, nor in DSP SSB mode.

Jim  VE7RF