[Elecraft] optimizing recorded audio

[Jim Brown]

On 4/22/2011 12:40 AM, Ian White GM3SEK wrote:
> Most people can improve articulation dramatically by slowing down only
> 10-20%, so it only requires a modest increase in the tempo setting to
> restore a normal brisk speed. Time compression is a re-sampling
> technique and it does introduce some artefacts, but these are minor
> compared with everything else that happens to a SSB voice signal.

You're right, Ian.  My advice is really directed at users who are not 
skilled in audio editing, and is part of a KISS (Keep It Simple, Stupid) 
philosophy.  I do a few things with editing that I wouldn't dream of 
recommending that others try (and I won't even mention them) because 
they are so complex and easy to screw up.  My experience with time 
compression goes back to the Lexicon D224 hardware product, a pro 
product that sold for about $7,000 in the early 1980s. I demoed and sold 
them to studios for the purpose of shortening radio spots (commercials). 
I heard them on a lot of material, all with very expensive voices. 5% 
compression sounded great, 10% was OK, but more than that was artificial 
sounding.

Don suggested keeping a copy of the original file. Yes, a good idea, but 
all of the editing software mentioned has an undo function, so if you 
listen to each step as you go along, you can get away without that. And, 
of course, you can always re-record the message, which I do occasionally 
because I don't like the first attempt.  In fact, I often record a 
message a half dozen time (or more) before I start editing it.

A few other suggestions.

When recording, make sure your shack is quiet -- close the door, turn 
off all the fans and air conditioners.

Work with the mic not too close to your mouth so that you don't get 
breath pops and low end boost, and make sure that the audio levels are 
right as shown on the editing software's meter and waveform display.  
You should NEVER see any overload, and it's best to keep the peaks of 
the waveform at least 3dB below max (0dB on the display).  If you do, 
throw out that recording and start over.  You CANNOT fix it by turning 
it down after it's been recorded.

After you've finished editing, use the EQ function to roll off the low 
end at about 100 Hz, and to roll off the high end at about 6 kHz.

If you like to use VOX (I do), record a click at the beginning of each 
CQ to activate the VOX a few milliseconds before the message starts. 
This prevents losing the first syllable of the recording. Adjust the 
peak level of the click to be 15-20 dB below the peak level of the 
message. Use this click only on messages that will transmitted alone, 
like your call, a CQ, and the Thanks message at the end of QSO.  Do NOT 
use it on an exchange -- you should activate the VOX with the live mic 
when you say the other guy's call. When all this is working well the 
click should not be transmitted.  To get the timing and level right, 
play the track through the rig and listen to the result.

Setting levels is VERY important.  I like to keep the highest peaks of 
the final recording between about -6dB and -3dB as indicated on the 
Audacity waveform display. It's also important not to set the output 
gain of the computer too high. Most sound cards have greatly increased 
distortion when they get close to full output, so it's best to run their 
output a bit lower to keep that distortion low.  You don't need to 
reduce it a lot -- 3-6 dB is enough.

When setting levels at the K3, remember that you want to match the level 
of the live mic going straight into the K3 with the level of the 
playback from the computer. We use the Line Input control to set the 
level of the playback audio, and to do that, we must temporarily set the 
K3 for Line Input and use the front panel Mic Gain. I usually set the 
Line In gain so that I get the same indicated ALC and COMP indications 
on the K3 meter display with playback as I do with the live mic (about 
10dB of COMP on the hottest voice peaks).

73, Jim K9YC