Split control loop

Started by George Herold July 20, 2017
Hi all, (as is usual) I think I figured this out in the process of 
asking you a question.  But I welcome any input or comments. 
First a bit of background.  I made the laser diode controller ~10 years ago.
I added this section so that you could side lock to one of the absorption 
features.  Input signal is low pass filtered(single pole), DC offset, 
maybe some gain and then sent to the piezo modulation input which controls
the laser wavelength.  (In retrospect it might have been better to put the 
low pass after the DC offset, but I'm stuck with what I did.)
Here's a front panel pic.

This works fine up to a frequency of ~3 kHz, 
where there is a piezo resonance*.  
There was not much to do with the side locked laser, so no one asked about it.

Now someone wants to use the laser to make a MOT.

The locking is a not stable enough, in that talking loudly or dropping 
a screw driver on the floor can cause it to loose lock.  One fix is to 
split the error signal and send the low freq. to the piezo and the HF to 
modulate the laser current.  I made a splitter.  RC low pass with 
intrument amp across the cap.  Shown here,

Well I mucked about with this for a while, but without a clear idea 
of what I was doing.  So last night I drew the following Bode plot.
I would typically set the Low pass for 0.1 sec (~1 Hz, 2*pi ~10) 
and turn up the gain till the piezo oscillated at 3 kHz.  
(gain ~1 at this point.)  Then I want to send the HF shaded section 
to the current.  But the photodiodes have a bandwidth of 
30kHz to 300 kHz. (depending on the gain.)  So I need to 
roll off the HF too.  And if the gain in each channel (piezo and current) 
is about the same then that means I start the roll-off at about 
1 ms (~100Hz)   So that's what I did this morning.  
I use the splitter to set the 1 Hz corner for the piezo and the 
LP filter on my front panel to set the 100 Hz corner for the current.  
It's been running for ~2 hours, through door slams and conversations.

It still won't stay locked if you drop a screw driver on the optical
bread board.  Hopefully good enough.  

George H. 
(thanks for reading) 

*I've mentioned this before and mucked about some with a notch filter.  
This helped a little.. but only until the next resonance of the piezo.