1
\$\begingroup\$

I'd love to hear any thoughts about a good break-before-make analog switch with low charge injection or even even better any thoughts on generating precision amplitude square waves in general. As to why ...

I have a 5V±5%/GND powered op amp (AD8648), a precision 2.5V(±0.2%) low impedance voltage, and an external 3.3V±5%/GND clock input that runs at about 10 KHz. I need to generate precision amplitude 10 KHz square waves that are symmetric about 2.5V. The amplitude will be controlled with a 10K digital potentiometer. The potentiometer resistance is known with 0.1% precision. The circuit needs to be as low cost as possible. Ideally, I need a precision of about 1mv for an amplitude of up to 2.5V±~1V. The resolution is not an issue as much as the precision. I plan to use the AD8528 64 tap potentiometer, since it comes calibrated with the 0.1% accurate full scale value stored in internal memory.

One possibility that I am considering is to scale the voltage across a precision 1V reference with the potentiometer and then use analog switches to switch a capacitor above and below the 2.5V common mode voltage. I have been looking at various analog switch possibilities and the best seems to be the LTC6943 switched capacitor controller, but it costs about $7 a pop. I could potentially use other analog switches, but I am worried about the voltage offset you can get from charge injection from a switch not made specifically to balance it out, like the LTC6943.

Below is the modified LTC6943 voltage inverter application circuit. It uses an inverting and non-inverting version of this circuit to drive the op amp input above and below the 2.5V reference by Vin, which will come from a digital pot divided version of the 1V reference.

Basic block diagram

So, to repeat the question, any thoughts about a good break-before-make analog switch with low charge injection or even any thoughts on generating precision amplitude square waves in general?

\$\endgroup\$
7
  • 1
    \$\begingroup\$ Some schematics or block diagram. \$\endgroup\$ Commented Feb 15, 2018 at 21:15
  • \$\begingroup\$ You can't just AC-couple your signal? \$\endgroup\$
    – The Photon
    Commented Feb 15, 2018 at 21:38
  • \$\begingroup\$ I want to avoid the uncertainty of the offset on the op amp between the potentiometer and the next op amp. I suppose that I could just add a really low offset op amp. Unfortunately the AD8648 has ~2.5mv input offset voltage. A low offset op amp might be cheaper than an good analog switch. Definitely something to consider. \$\endgroup\$
    – crj11
    Commented Feb 15, 2018 at 21:49
  • \$\begingroup\$ Is this what you are proposing? i.imgur.com/ARJQwka.png \$\endgroup\$
    – jms
    Commented Feb 15, 2018 at 22:05
  • \$\begingroup\$ Not exactly. I updated the block diagram to perhaps make it a bit more clear \$\endgroup\$
    – crj11
    Commented Feb 15, 2018 at 22:12

1 Answer 1

1
\$\begingroup\$

I did manage to find a good low charge injection analog switch, and it is about 1/3 the cost of the Linear Tech part. The LTC part probably has better specs, but they aren't really needed in my design. Analog Devices has a good article that discusses charge injection on page 22 page of this issue of Analog Dialogue. It turns out that a key indicator of low charge injection is high switch resistance. It makes sense because charge injection is proportional to gate area, and lower resistance requires more gate area. My design can tolerate the higher switch resistance in order to avoid the spikes caused by the charge injection.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.