The Tayloe Mixer: A Low Noise Solution

[Davor]

True.
But also you can sample once, keep samples in SAH buffer and sum with the previous samples. With IB MDs you have a semi stationary situation, so Tayloe channels are hardly changing at all between samples, and Tayloe capacitors may be dimensioned for practical purposes, say 15Hz cutoff. Using a pair of instrumentation amplifiers and ADCs you have a perfect digital solution for IB.

[mario]

This guy Tayloe should have won some kind of engineering prize for his design.

The biggest issue for the lack of mass adoption, I think, is that it's a resource hog. I would love to see a manufacturer release a one chip solution in SOIC.

As things stand, you have to double the resources (and pcb real estate) for every mixed output -ouch.

[Carl-NC]

How is the Tayloe mixer different from full-wave demods (typically using the 4053) commonly found in modern VLFs?

[Davor]

Not by much, but it does it beautifully. It is a direct conversion front end with I and Q outputs.
So you have a kind of rotary switch that supply 4 large capacitors with alternating signal from Rx front end, and it can be arranged in doubly balanced fashion as well. The capacitors are charged and with 4x input impedance make a LPF. Amplification is done at the baseband instead of RF, and that's the beauty of direct conversion. With I and Q all phase information is preserved. If you remove DC by series capacitors it will still pass amplitude and phase information, but now it is equivalent to a notch at the RX frequency, and in a middle of a received band. This particular feature can be abused in IB MD for incredibly simple and effective GB and motion filtering. Besides it is not depreciating signals that are close in angle to GB, and it all happens before any amplification even starts. So whatever you really need to amplify is already filtered and there are no high amplitude residuals.

Another nifty trick is also possible with Tayloe, and that is transposition of a RX (carrier + and - baseband BW) back, or to an arbitrary frequency (carrier + and - baseband BW), thus making a sort of tracking filter (with or without arbitrary frequency transposition).

In case you are not after channel-type discrimination filter banks, your perfect VLF IB MD Rx would be just that: Tayloe mixer, DC block, Tayloe transposition to audio frequency. The output would be an easy to listen sinus with amplitude of a target, with perfectly nulled carrier and motion. A minor drawback would be a Tx that could not be a self oscillating one, but driven from a separate oscillator.

[Carl-NC]

Not by much, but it does it beautifully. It is a direct conversion front end with I and Q outputs.

Amplification is done at the baseband instead of RF, and that's the beauty of direct conversion.

Practically all VLF detectors use direct conversion (zero-IF). Doing the conversion before any amplification is almost always a losing proposition in SNR, which is probably why no one does it. But besides that, the mathematics of the Tayloe method looks to be the same as what detectors are already using.

[Davor]

More or less. The only difference is (electric) ground reference. While Tayloe achieves it by no effort, usual MD mixers must have properly ground-balanced opamps to do so.
In case of frequency transposition anti-Tayloe mixer, there is no ground balancing done anywhere, as everything is done without amplification of any kind.

Besides, the only drawback noise-wise is 1/f noise, but with signal levels sky-high it is a non-issue. There is no point amplifying a signal that is already much higher than noise.

Shortly - Tayloe is a chopper on steroids. Choppers are used prior to opamps to make them better, not after.

[golfnut]

I read conflicting arguments - the one thing that comes up repeatably ..

The conversion loss is best-in-class at less than 1dB So you hold on to more of the valuable signal.

A double balanced normal switcher loss is higher .. maybe 5dB (this would need a preamp) If the noise figure of the preamp was 4dB it would be better to use a Tayloe without a preamp!


From the patent.
"One advantage is low conversion loss. The Tayloe Product Detector can exhibit less than 1 dB of conversion loss, which is 6-7 dB improvement over the typical conversion loss of 7-8 dB in the prior art. This 6-7 dB conversion loss improvement translates into a 6-7 dB improvement in overall receiver noise figure. The noise figure improvement results in substantial receiver performance gains, in part because a pre-amplifier may become unnecessary as a result. The use of a pre-amplifier, while improving receiver noise figure by overcoming front end receiver loss, causes large signal performance to suffer due to an amplified high-level input signal overloading the input mixer. Because the Tayloe Product Detector significantly reduces front end loss, the pre-amplifier and its associated problems may become unnecessary in future direct conversion receiver designs."



This has made radio hams like Don Huff use them without a preamp (as the noise contribution from the pre amp can be larger!)

MARIO: The other thing - To sample all 4 phases of the Input - You require 4x the Tx frequency.

I just got some 60kHz xtals to run a 15kHz Tx and some chips to do the mix. Be a good while tho got to redecorate the room the desktop is in - So I may disappear for a while!!

Steve

[Davor]

The prior art mentioned in Tayloe's patent deals with diode mixers and their losses. What Carl mentions are the switching mixers using the same switchers as the Tayloe mixers, hence no loss with them either. The only big difference is how these mixers deal with electrical ground in comparison with mixers in, say, TGSL.

There is a problem of 1/f noise which is accompanied with zero IF designs, but again, we are not talking about faint Ham radio signals, but metal detectors with lots of juice at hand.

[golfnut]

I dont see those numbers in my reading.

A 4066 type switching device set up as a double balanced mixer has a practical conversion loss of 4 to 6dB

S

[simonbaker]

This guy Tayloe should have won some kind of engineering prize for his design.

The biggest issue for the lack of mass adoption, I think, is that it's a resource hog. I would love to see a manufacturer release a one chip solution in SOIC.

As things stand, you have to double the resources (and pcb real estate) for every mixed output -ouch.

Just been looking at this doc (attached) and see how the mixer is very related to the synchronous detector.

I was wondering how the "bias" network in front of the switches affects the R/C time constant of the integrators -- it looks like it would be part of it.

-SB

[simonbaker]

it can be arranged in doubly balanced fashion as well.

It seems you need this doubly-balanced arrangement to get "full-wave" detection and not throw away half the wave. It is a fair number of parts, but nice if you can do it.

-SB

[lonelyWOLF]

at the below. part of the fisher1266 schematic. it is seem tayloe mixer

[Carl-NC]

The first circuit below is a "half-wave" demod and is used in low-end VLFs. The second circuit is a "full-wave" demod used in better VLFs. Mathematically, it's equivalent to the Tayloe design. Even simpler, because the inverted input reduces the number of integration caps by half. This matters, because you want pretty good matching between the integration stages, and matching 2 stages is easier than matching 4 stages.

[Davor]

True, but again you do not require any amplification prior to the mixers, and as a bonus you don't have to mess with balancing the opamps prior to mixing. Double balanced action is achieved with centre tap on Rx coil.

[golfnut]

The first circuit below is a "half-wave" demod and is used in low-end VLFs. The second circuit is a "full-wave" demod used in better VLFs. Mathematically, it's equivalent to the Tayloe design. Even simpler, because the inverted input reduces the number of integration caps by half. This matters, because you want pretty good matching between the integration stages, and matching 2 stages is easier than matching 4 stages.

This is the Double Balanced tayloe from Dans paper. It combines 8 phases.

Dan arranges the input signals in anti phase the same as Carls drawing

I assume Carls drawing clocks at 4x the same as Dans