The Tayloe detector has become a mainstay for quadrature front-ends in SDR radios. It makes developing I and Q audio paths easy and it has impressive conversion loss and strong signal handling. Here is an example of the Tayloe detector in its native form:
The incoming signal is sampled by a CMOS multiplexer at four 90 deg intervals over its cycle. This results in detected audio samples of 0 deg, 90 deg, 180 deg and 270 deg on the sampling capacitors. Op-amp difference amplifiers subtract the 0/180 samples and the 90/270 samples to produce 2 audio samples that are 90 deg spaced. Any strong off-channel signals that blast through the sampler appear at equal phases and amplitudes on the 4 sampling caps. The difference amplifiers subtract these equal-amplitude / equal phase spurious signals to produce a net value of zero at the I and Q outputs. Notice that the sampling caps combined with the input signal source impedance form lowpass filters which can be used to advantage if the RC time constant is set just a little above the desired audio passband. However, when set this close to the desired passband, their values need to be matched, otherwise the phase relationships of the four detection phases will be skewed.
Now look at the version of the Tayloe Detector used in my QSD receiver It’s shown in the bottom half of the schematic:
I lifted this from the Soft-Rock receiver design because there’s a large body of field experience with it and it looked like it might offer better performance. In this version, instead of a single-ended input driving a sampler with 4 output phases, a transformer produces two 180 deg phases to drive the sampler, which is run “backwards” from the usual configuration. This produces just 2 sampling phases, spaced 90 deg. At first glance, it looks like a good idea because it saves two matched caps at the expense of an input transformer. It also appears to have signal balance which might offer cancellation of odd-order products at the I and Q outputs, improving LO leakage back to the input, better strong signal handling and less unwanted signal bleed-through.
Upon closer inspection, however, it fails to deliver. Because the LO phases are not 180 deg apart, there is little or no LO cancellation at the input. Also, strong AM modulated off-channel signals blast right through to the sampling points, without the cancellation offered by the difference amplifiers in the original Tayloe. So, the bottom line is it appears to reduce the capacitor count at the output, but at the expense of a transformer and degraded off-channel signal feed-through.
My measurements indicate that the LO isolation is little or no better than the single-ended version.
Also, at the input transformer, there is significant audio frequency AC current at the center tap where the bias is injected. This results in a need for a very”stiff” bias source which, ideally, should be an op-amp voltage source. I stiffened it with large electrolytic capacitors. The single-ended Tayloe does not have this problem.
I think I’m going to redesign the QSD board to incorporate the original Tayloe design. Doing this could free up one of the two sets of sampling switches in the multiplexer, allowing it to be used as a transmit modulator.
W3JDR Bright Ideas 