[Laser] TX Circuit for high speed data.

[James Whitfield]

From: "Chuck Hast"

> I am feeding it a square wave, and it is getting rounded off. I want to
keep it
> nice and square,
>
The rounding is an effect of low pass filtering.  It might be that the FET
transient response is the cause, but I assume not since you stated that if
you jumper past the input circuit you get 10 MHz response.  That leads to
the conclusion that the input circuit is shaping the driving waveform with a
low pass filter by the time it reaches the gate of the FET.

I know that the schematic of the input circuit does not look like a low pass
filter, but the way you describe the way the circuit behaves, some thing is
making it act like a low pass filter.  I suggest this is what is going on:
The square input passes through the first 1K resistor to the the capacitance
that is the composite of the two back biased 1N914 with the 100K bleeder (
first stage low pass filter ) then through the second 1K resistor to the
stray capacitance of the gate of the FET.

Now the trick is to figure out how to make the circuit act less like a low
pass filter, or at least raise the effective frequency of the filter so that
it is not a problem.  You can perhaps reduce the capacitance of the two
signal diodes if you substitute different components.  Reducing the 100K
resistor, as suggested by Yves, should work too, but I think that it will
also shift the DC bias conditions on the FET gate.  If the result of that
action is undesirable, you may need to adjust other components, such as the
22 ohm source resistor.

My suggestion will leave the DC bias of the circuit unchanged, but will
provide a high pass signal path to the FET gate:   If you capacitively
couple the signal from the input to the gate of the FET, it will be a high
pass filter that will restore the shape lost by the low pass filter.  My
initial estimate of the capacitance needed may have been too high.  Perhaps
it should be on the order of the stray capacitance of the FET's gate, or
perhaps as much as ten times that high.

Also, you need to be aware that when you substitute the laser diode for the
LEDs that you have been testing with, the non-linearity of the LD will tend
to "square up" the edges of the emitted light.  LEDs are linear and will
reproduce any "squishiness" the circuit will put into the current feed to
the light emitting device.  LD will be linear above the laser threshhold,
but  when the supply current drops below the threshhold, the light intensity
will quench very rapidly.  I cannot predict how the difference will affect
your transmitter, nor suggest how you would measure the difference for the
data rate you are trying to use.


> I was thinking, why not replace the IRF-510 with say a 74HC07 open drain
> buffer, I do not need anywhere near the power the MOSFET is designed for.
>

An interesting idea.  I suggest that you include a resistor on each drain to
balance the current to each buffer.  A 4.7 ohm resistor would add 94
millivolts drop, whereas a 1.0 ohm would add only 20 millivolts.  Anything
in that range should provide the load balance needed without affecting the
low voltage drop needed at the current draw for quenching the laser diode.
It should not slow down the response time either.

I don't think you said how much current your circuit is designed for.  I was
using calculations based on 47 ohms for R1, but that would limit your LD to
about 25 ma. ( assuming that you need 3.6 volts for the LD ).


Best Wishes

James
 n5gui