Post Go back to editing

Lock in Amplifier assistance needed.

Hi, I need to design a lock in amplifier. The signal frequency is between 2MHz to 4 MHz. Can I use ADA2200 or AD630? If not, could you please recommend one for me? Thanks! 

Parents Reply
  • Brian, Thanks! I saw you have another post commenting using LTC6943 for lock in amplifier, would that work for this kind of frequency? LTC can work up to 5MHz? 

    Also, I completely don't understand how LTC6943 works. Each of the datasheet example only connects partial circuit. I can't really get any of those work. Tried to use the LTspice model of LTC1043, can't get it work either. For the lock in amplifier example, I can't figure out what are the proper connections for each pin. For example, on page 11 of LTC1043, the lock in amplifier example, it only shows connections for pin 12, 13, 14 and 16. Any of the other pin need to be connected (beyond V+ and V-?). I can't seem to get it work at all. Please help! Thanks!

  • LTC6943 and LTC1043 are dual parts: each contains a DPDT switch, i.e. a pair of SPDT switches. You need to connect at least the pins that are not marked specific to any one half. E.g. in the LTC6943 pin descriptions, all the xA and xB pins are specific to one half of the chip. The common pins are 3 (V+), 15 (V-) and 14 (COSC). Those must *always* be connected. The COSC pin can be driven with an external CMOS level clock, and that's what will suit you in a lock-in application.

    IMPORTANT: Do note that the "CMOS Level" here is a bit misleading. The COSC input is referenced to V-, and with +/-5V supplies, you definitely need the input to be +/-5V as well. The LT1011 comparator used in the lock-in amplifier example has an open-collector output that connects the output to pin 1 named "GND". But this pin is not connected to GND in this application: it is connected to -5V (V-) instead! So the 14 (COSC) pin sees a clock signal with high/low logic levels being essentially the bipolar supply rails.

    You can generate such signal using a sufficiently fast analog multiplexer: the clock is then connected to the "select" input of the multiplexer, and the output is switched between two inputs - connected to opposite supply rails.

    In the LTC6943 lock-in amplifier example on page 10 of the datasheet, the pin 14 (COSC) is shown connected - it's used for clock input You do need to power the part via pins 3 (V+) and 15 (V-), connecting +5V to pin 3(V+) and -5V to pin 15(V-). Of course if you reuse that application node, you must reuse it verbatim - its parts are meant to work together. So before you decry that it "doesn't work", duplicate the circuit exactly. The transformer can be easily substituted of course, as can be the op-amp as long as you're careful with matching the specs. The comparator must be the same part, or a close functional equivalent. Then once you have a working baseline, start modifying it. And don't mess with SPICE. Do it on an actual breadboard. Reality doesn't care about what SPICE says, and using even LTspice sometimes requires good understanding of the limitations of the models and the solver itself.

    While LTC6943 is specified to a maximum external clock frequency of 5MHz, it's not the only way to skin a cat. A very sensitive solution could be also obtained by using a "switch" with no charge injection: an externally clocked ADC. Take the reference signal, double its frequency with a PLL, then use the alternating ADC samples as the I and Q values. Those get dumped into separate DAC channels, and constitute the I and Q outputs. You can apply low-pass filtering there, then re-sample them at a much lower sampling frequency with an MCU, and further process the data that way. I've done that back when parallel-interface ADCs and DACs were in their heyday, and it worked quite well. It really depends on what's the use of the signal thus recovered. For example, if all you need is an "indicator", then it may be acceptable to use even a modest MCU to accumulate some ADC samples into RAM, then process them off-line: compute the average I and Q value, then the amplitude (square root of the sum of squares) as well as the phase angle (`atan2` function). Then output those, sample again, and repeat. For low-enough signal bandwidths (not related to reference clock!), you can feed those values to a pair of DACs - potentially even DAC in the MCU, or just PWM timer outputs that you then filter externally to the MCU - and provide analog outputs that way.