I see several options for this; there is the TI's application bulletin (originally a Burr Brown App. Bull. from 1997) --- www.ti.com/lit/an/sboa068/sboa068.pdf

There's LT's LT1001-based circuit: http://cds.linear.com/image/416_circuit_1.jpg

Then, there is the simple circuit shown in Wikipedia's Precision rectifier page:


simulate this circuit – Schematic created using CircuitLab

which I can combine with the original signal (i.e., its output + Vin/2) to obtain the absolute value of the input signal:


simulate this circuit

Notice that I indicated LT6237; among other things, I need moderately high bandwidth (I want the absolute value of signals going up to 10 or 20 MHz).

I find this the simplest option, and can't see any obvious disadvantages --- are there any important advantages in the other designs with respect to this one?

All circuits are sensitive to resistors mismatch; I'm planning to use 0.1% resistors either way. But I wonder: are some of the circuits especially sensitive to resistors mismatch?

I believe that one advantage of this one is that neither of the two op-amps has the other op-amp in its feedback path (which the other two designs do, at least for a fraction of the input signal cycles), so if we have stable op-amps, then we do not need external compensation capacitors. Is my intuition correct?

With the above said: I run a simulation with LTSpice, selecting LT6237 for the op-amps, and the circuit behaves wildly different depending on the diode that I select. For the resistors, I use the ideal/generic resistor from the LTSpice tool bar. If I select the generic/ideal diode, then the circuit oscillates when the output goes positive. With 1N4148, it does not. With fast-recovery diodes (the first fast recovery from the list), the output is wildly distorted (no oscillation, but the output is quite deformed), for Schottky (e.g., 1N5817), the circuit oscillates.

Are these simulation artifacts? What should/could I expect in practice when assembling the actual circuit?


  • \$\begingroup\$ I'd be worried about propogation delays and node capacitance at 10 MHz \$\endgroup\$ – Jasen Oct 13 '17 at 3:47
  • \$\begingroup\$ The diode capacitances strongly affect the settling of virtual-ground nodes. What precision MUST you have? Or is bandwidth paramount? \$\endgroup\$ – analogsystemsrf Oct 13 '17 at 5:15
  • \$\begingroup\$ Precision is not critical --- if I could get the absolute value within a 2% or 5% margin, that would be excellent. However, my signal does have high-freq. components going up to approx. 10 MHz, maybe a bit higher, and I do want the absolute value to include those. The distorsion resulting from the filtering of the harmonics that show up when rectifying a 10MHz signal, that part I can live with (and for that matter, the crossover distorsion is not critical --- though the LT6237 has a > 50V/us slew rate). So, I guess I should put the capacitors anyway? 10 or 20 pF, maybe? \$\endgroup\$ – Cal-linux Oct 13 '17 at 17:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.