The Tayloe Mixer: A Low Noise Solution

[simonbaker]

True, the RX phase includes null residual, ground, and target. We normally run the demod outputs through filters to take out the null residual and ground, assuming the target is a transient delta. I've never worked with a Costas loop, but it looks like it might be an interesting way to combine the demod & filter functions, whereby the VCO loop filter removes long-term null & ground phase but not transient target phase.

The references I looked at indicated the Costas Loop is mainly for demodulating double-sideband suppressed carrier signals, essentially by "deriving" the carrier frequency. (It seems to require symmetric sidebands, but maybe that could be modified for the null/ground/target signal combination???) I don't yet see the applicability or advantage to the metal detector problem.

In the continuous-wave metal detector problem, we have a rock-solid reference, the TX signal. Deriving it or tracking it would seem an inferior and more noisy way to obtain it.

All the low-pass filters in the Costas Loop seem just as burdensome as LPF filtering the traditional "synchronous detector" output to remove the null/ground signal.

It's an interesting circuit (especially for communications) but I'll need to see more implementation details for MDs to judge what it can really improve for us.

-SB

[golfnut]

back on subject

Back on Marios original question,,,

Is the Tayloe worth having?



It depends what you currently use,,


1) if on simple machines like xGSL the existing preamps have a noise figure NF of more than about 9dB and you currently have a demod like 4066.....

- you could ditch the noisey preamp and high loss mixer and use a Tayloe mixer without a preamp - In line with my spreadsheet clip and get better S/N performance. The hassle is clocking the mixer at 4x.

I just got some 60kHz xtals to try at some point, to make a 15kHz machine.


2)Carls front end / full wave demod clocked at Tx rate - Id guess is the industry std for good machines. Other brands use this sort of idea.

I cant model the two schemes to separate them.

Steve

[Carl-NC]

All the low-pass filters in the Costas Loop seem just as burdensome as LPF filtering the traditional "synchronous detector" output to remove the null/ground signal.

Could be, I've done no more than a cursory glance.

Back to Tayloe mixers... I ran a sim comparing half-wave, full-wave, and Tayloe, the results were partly expected and partly surprising. Perhaps someone else would care to run such a sim, before I let the beans out of the bag.

[simonbaker]

Could be, I've done no more than a cursory glance.

Back to Tayloe mixers... I ran a sim comparing half-wave, full-wave, and Tayloe, the results were partly expected and partly surprising. Perhaps someone else would care to run such a sim, before I let the beans out of the bag.

Can you post a picture of the schematic your using for the Tayloe or a reference? There are several variations in the document I posted earlier.

I'd be interested in the schematics you're using to model the "synchronous detectors" also, just to try to compare apples to apples.

-SB

[Davor]

1... the Costas Loop is mainly for demodulating double-sideband suppressed carrier signals, essentially by "deriving" the carrier frequency...

2...In the continuous-wave metal detector problem, we have a rock-solid reference, the TX signal.

3...All the low-pass filters in the Costas Loop seem just as burdensome as LPF filtering the traditional "synchronous detector" output to remove the null/ground signal.

1... frequency, yes, but more importantly phase. You need a solid phase reference for any phase dependent demodulation, and Costas loop provides.

2...That's unfortunately just wishful thinking. Free running Tx oscillator is all but rock solid. Usual implementations are not even symmetrical, let alone amplitude stabilised. I think most of the existing designs could benefit from some real rock-solid Tx.

3... You actually need only one Costas loop for all discrimination purposes. There is room for a single target discrimination at any time, so one loop should do it. In such case it is much simpler than the existing gear with discrimination.


I think it is a worthy effort and I'll give it a try eventually. I do not expect spectacular outcome, but somewhat simpler design, and a prospect of a more sensible interface. We'll see.

[simonbaker]

1... frequency, yes, but more importantly phase. You need a solid phase reference for any phase dependent demodulation, and Costas loop provides.

2...That's unfortunately just wishful thinking. Free running Tx oscillator is all but rock solid. Usual implementations are not even symmetrical, let alone amplitude stabilised. I think most of the existing designs could benefit from some real rock-solid Tx.

I'm not following that. Whatever phase reference the Costas loop derives is just from the TX oscillator signal coming through the RX coil anyway. If you mean low-pass filtering the TX oscillator phase through a PLL to "stabilize" it, we could do that also off the TX oscillator directly, rather than using the noiser version of the TX signal (called the RX signal).

However, I think we don't want a rock-solid independent oscillator reference. If we do, then any oscillator jitter, cycle, or drift in the TX oscillator will show up in the RX signal and look like a phase modulation signal when referenced to your "rock solid Costas reference" -- in other words, like target behavior. The whole purpose of syncing to the TX oscillator is to "lock" to it and cancel out phase/frequency drift components. If you create a "smoothed" version of the TX oscillator, it seems to me you will then detect all the imperfections in the TX frequency/phase in your detector and they will look like target.

3... You actually need only one Costas loop for all discrimination purposes. There is room for a single target discrimination at any time, so one loop should do it. In such case it is much simpler than the existing gear with discrimination.

I think it is a worthy effort and I'll give it a try eventually. I do not expect spectacular outcome, but somewhat simpler design, and a prospect of a more sensible interface. We'll see.

Ok, I can't visualize it without some kind of circuit and TX, RX and target signal model to go with it. I'll be interested in what you show us.

Regards,

-SB

[golfnut]

Spill dude Im turning blue


I cant easily get cmos switches flying in LTSPICE..

Steve

[Carl-NC]

Below is the test schematic. I'm running a 10kHz pure sine, but at 20ms I switch over (via IC4) to a SFFM source which provides a 2ms phase shift, up to about 45 degrees, then back to zero. This simulates a swept target. Various demod clocks are as described before.

[golfnut]

Your enjoying this arent you !

[golfnut]

Ive got 4 crummy Jfets in spice.



Thats a fancy sim - a commercial one? - which one.

I got AWR in work but I cant sit doing this stuff there ..

S

[Davor]

Below is the test schematic.

Please have in mind that practical implementations of Tayloe go without additional resistance in series with switches. It exists though, and it is in effect attributed by source resistance plus Ron of the switches. That's why Tayloe mixer works fine in spite of the not-so-symmetrical impedance of differential amplifiers that follow the capacitors.

[golfnut]

I gotta crash, Im off site tomorrow and its a bit of a drive.

check in tomorrow for the next morcel of the future

S

[Carl-NC]

Please have in mind that practical implementations of Tayloe go without additional resistance in series with switches. It exists though, and it is in effect attributed by source resistance plus Ron of the switches. That's why Tayloe mixer works fine in spite of the not-so-symmetrical impedance of differential amplifiers that follow the capacitors.

The same circuit methods used in the Tayloe mixer can be used in the other demods; the only real differences are in the switch configuration and timing. Ergo, I used a simple RC load on all methods to give an apples-apples comparison.

[Davor]

True. However, in Tayloe you do not require additional resistors because there is no overlapping ON periods that need to be isolated. That's where you slightly deviate from apples and see some extra fruits

[Carl-NC]

There are no overlapping ON periods in the other two, either.