[Laser] PWM of LED in QST

[C. Turner]

In talking about the Ramsey kit, I note that their manual - freely 
available online - has excised from it the schematic.  I'm not sure what 
this means, but I (personally) would be reluctant to post it online for 
all to see.

Having said that, however, I was provided with scans of the schematic 
after asking some questions and it has come in handy for other questions 
that have been asked of me about it, so here goes...

* * *

The Ramsey kit uses a Motorola processor and digitizes the audio, 
producing a PWM waveform with a clock rate (according to the manual) of 
around 18 kHz, having been lowpass filtered with what looks like 4 poles 
(two LPF's - and 2 more poles if you count the integrators in the mic 
amp and the PGA.)  Using the old trick of switching some pins on the 
processor between "outputs" (grounded) and "inputs" (open) a sort of AGC 
circuit is formed with the integrator stage providing the function of a 
PGA (Programmable Gain Amplifier.)  A jumper is provided that switches 
it between a "tone" mode and "audio" mode for normal use.

Now, the complaints/questions that I have received about the Ramsey kit 
had mostly to do with the AGC in that it seemed to be too aggressive or 
too "steppy".  Not having used the LCB6K myself (I can't quite get 
myself to spend the $50 or so...) I've not heard it (but a recording 
from the receive end would be interesting) but I did spot a problem 
right away:  Unfortunately, the resistors used to set the PGA were 
ill-chosen for the task, with massively varying steps in gain - some 
ineffectual, and others annoyingly large.  It would appear that most 
people simply replace the kit's CPU-operated AGC with a potentiometer - 
that is, a 100k pot to ground at the "CPU side" of C15.

Having already designed a PWM-based LED/Laser modulator years before 
(see:  
http://modulatedlight.com/optical_comms/LED_pulse_width_modulator.html ) 
with AGC (although it went about doing the gain control in a very 
different way) I decided to try a crack at doing a similar scheme, the 
result being here:  
http://modulatedlight.com/optical_comms/simpler_pulse_width_modulator.html  
(Yes, it's THAT link again...)

By crunching numbers and properly selecting resistor values, I was able 
to obtain step sizes of 4dB or less, yielding an AGC range of about 
13dB.  Since this wasn't very much, I did some simple DSP and gained 
another 12dB of AGC range, providing a respectable AGC range of about 
25dB - enough to accommodate a person at the microphone or, when the 
speaker stops talking, the voices of those within several feet.

The correspondence also implied another problem with the LCB6K:  The AGC 
algorithm was "wrong" somehow.  (Again, without a recording I don't know 
what the problem was, exactly.)  In writing the code for this unit, most 
of the time that I spent (by far) was in the creating of a 
semi-intelligent and decent-sounding AGC algorithm that minimized 
"pumping" (that is, the annoying, sudden increase of noise between 
words) and "popping" - the inability of the AGC to react quickly enough 
to prevent brief, severe clipping.  With such a simple system it's not 
possible to avoid the two entirely, but it's practical to avoid 
exacerbating the problem excessively.

***

At the moment, the "simple" PWM system that was devised for the web page 
has languished for more than a year and a half, but it is a working 
system.  The original prototype was "over-built" and could be simplified 
somewhat and there are a few other "features" that I mean to add to it 
(different gain control mechanism, etc.)  I suppose that if someone were 
*really* interested, I could add a few extra features and make some 
chips available for cheap ($5 or so to cover the chip, mailing and 
"hassle factor") if they were interested in experimenting with it...  If 
nothing else, it makes for a cool-sounding digital audio sine wave 
generator!

73,

Clint
KA7OEI