I have built a basic shield heater driver based on Rana’s initial suggestions using a buffering diff input, a Sallen-key active LP filter (0.15 Hz, 2nd-order) with a BUF634 in the loop at the output to provide some beef. This is presently just laid out on solderless breadboard, we can transfer easily to protoboard which is the next step.
For the LP filter I chose R1=R2=100kΩ and C1=C2=10µF. Full circuit is illustrated below. I modified this from what was linked in PSL:1747 as that was not technically a Sallen-Key.
Schmatic of full driver
I took a Transfer function of the circuit with the opamp loop closed before the BUF634. However, I have left this on a floppy disk and will need to recover it Monday. It has the expected poll at 0.159 Hz with a 1/f^2 roll off. I will attach this in a seperate post.
I also measured the voltage noise at the output of this circuit (no load), the PSD is below.
Nosie at output of LP filter stage and after the BUF634 was introduced into the network.
I then put the BUF input the final part of the loop (as illustrated) and measured the TF of the whole thing unloaded.
TF (output unloaded) with the BUF634 installed into the driver's feedback impedance network. Something's not quite right.
And with a 56Ω load (as expected from the ring heater.
Again TF this time with load with the BUF634 installed into the driver's feedback impedance network. Something's not quite right.
Both of these have a roll up above 40 Hz that I don’t really understand. I seem to recall the first transfer function I took of the LP filter stage had some higher frequency roll up but not this low. When I moved some stuff around it was improved. It is possible some additional elements need to be added to the feed back network or that the solder less breadboard has a loose connection. All speculation. This just needs to be trouble shooted by hand.
I also measured the voltage noise with the the system loaded with 56Ω with the BUF634 in the network (this load is about what would expect the shields to be). This is in the first plot of noise.
At this stage it needs to be tested under load and providing some current, for this we will need a stable voltage source and the noise can be measured by sampling the voltage drop across the load.
Data and plots attached in a zip file.
Edit: fixed linking to noise figure -awade Mon Oct 17 16:52:17 2016 |