The Tayloe Mixer: A Low Noise Solution

[mario]

I have been doing preliminary research on one of my projects and would like your feedback regarding the issue.

After probing my detector for sources of noise, I found noise originating from the CD4066 multiplexers. These were old parts, and I decided to replace them with newer, much quieter parts from Ti.

I managed to eliminate some noise, but remained wondering if a better solution existed. I found several possibilities and became fascinated by an unconventional mixer developed by Dan Tayloe.

He invented a scheme where he takes two signals that are 90 degrees out of phase to achieve an output with less than a 1db conversion loss. He's able to recreate an output signal that is highly linear with extremely low noise properties. You may find additional information here and here.

I find it interesting that this approach has not been discussed before since it has the potential to solve one of the few remaining sources of noise. It seems our obsession with low noise has not strayed to unconventional solutions to our remaining problems.

I'm thinking about eliminating the CD4066 mixers entirely and replacing them with a Tayloe Mixer board.

I know the integrator circuit in my detector is masking the most faint of targets.

Can somebody go to this site (here) demonstrating an implementation of the mixer and advice me on whether or not I should proceed with this project?

I have a few projects to complete and need to better manage my limited time.

To the gurus:

Will I waist my time pursuing this project? Any general comments? Suggestions? Type of issues I may encounter? Benefits? Is there a better solution to my noisy CD4066 integrators?

I Thank you

[simonbaker]

I have been doing preliminary research on one of my projects and would like your feedback regarding the issue.

After probing my detector for sources of noise, I found noise originating from the CD4066 multiplexers. These were old parts, and I decided to replace them with newer, much quieter parts from Ti.

I managed to eliminate some noise, but remained wondering if a better solution existed. I found several possibilities and became fascinated by an unconventional mixer developed by Dan Tayloe.

He invented a scheme where he takes two signals that are 90 degrees out of phase to achieve an output with less than a 1db conversion loss. He's able to recreate an output signal that is highly linear with extremely low noise properties. You may find additional information here and here.

I find it interesting that this approach has not been discussed before since it has the potential to solve one of the few remaining sources of noise. It seems our obsession with low noise has not strayed to unconventional solutions to our remaining problems.

I'm thinking about eliminating the CD4066 mixers entirely and replacing them with a Tayloe Mixer board.

I know the integrator circuit in my detector is masking the most faint of targets.

Can somebody go to this site (here) demonstrating an implementation of the mixer and advice me on whether or not I should proceed with this project?

I have a few projects to complete and need to better manage my limited time.

To the gurus:

Will I waist my time pursuing this project? Any general comments? Suggestions? Type of issues I may encounter? Benefits? Is there a better solution to my noisy CD4066 integrators?

I Thank you

What kind of noise do you mean? Exactly where and due to what?

Regards,

-SB

[mario]

Good point!

There's no point explaining something which can be heard. You can hear the noise here: http://www.youtube.com/watch?v=U7mkvK-6Zos

It's really not bad if your detector is ground balanced properly. However, the chirping, background noise was a little more pronounced before replacing the stock integrators.

Before replacing them, I had experimented with minute targets and noticed the chirping noise to effectively overwhelm the signal. In deed, in some of the faint targets I would hear the chirping noise instead of the expected "wehoooo"!

I would like to eliminate the noise entirely.

[simonbaker]

Good point!

There's no point explaining something which can be heard. You can hear the noise here: http://www.youtube.com/watch?v=U7mkvK-6Zos

It's really not bad if your detector is ground balanced properly. However, the chirping, background noise was a little more pronounced before replacing the stock integrators.

Before replacing them, I had experimented with minute targets and noticed the chirping noise to effectively overwhelm the signal. In deed, in some of the faint targets I would hear the chirping noise instead of the expected "wehoooo"!

I would like to eliminate the noise entirely.

Of course I have no idea what the circuit is producing that sound, but I could imagine the first "noisy" part is due to an unstable sync pulse oscillator (or phase noise on the sync pulse oscillator), or phase noise in the detected signal (like a mechanical vibration). Is it possible it was a coincidence it changed when you substituted a new chip? Did you change back and get the bad symptoms again?

It's hard to imagine that any of the chips you mention would cause that kind of effect.

Perhaps one of the mixers is acting like a "sampler" (very short sample time) and could pick up some aliasing?

From a cursory reading of one of the references, it seemed to me the "double balanced mixer" would be as good as any; I can't see a reason why the Tayloe would be better for a MD.

However, I have no hands-on experience with that circuit so would need to really play with it and troubleshoot.

I think these kinds of mixer chips do seem quite useful in general for what we are doing with VLF MDs.

Regards,

-SB

[mario]

Of course I have no idea what the circuit is producing that sound, but I could imagine the first "noisy" part is due to an unstable sync pulse oscillator (or phase noise on the sync pulse oscillator), or phase noise in the detected signal (like a mechanical vibration). Is it possible it was a coincidence it changed when you substituted a new chip? Did you change back and get the bad symptoms again?

If I change them back, it will definitely worsen the symptom. I went through great lengths getting these newer parts to behave. At one point I changed the parts with ones that had slower pecs. It seemed to me these slower speced parts weren't sampling fast enough and did worsen the target response.

Assuming we would like to replace these mixers and employ a different mixing strategy. What would you say be a better way of getting the audio fidelity?

Is there a circuit you would recommend? Or should I say, is there a circuit that will not take up half the board?

PS, the current circuit is not that bad, but I want one with lower noise. Low noise is key.

Thanks!

[simonbaker]

If I change them back, it will definitely worsen the symptom. I went through great lengths getting these newer parts to behave. At one point I changed the parts with ones that had slower pecs. It seemed to me these slower speced parts weren't sampling fast enough and did worsen the target response.

Assuming we would like to replace these mixers and employ a different mixing strategy. What would you say be a better way of getting the audio fidelity?

Is there a circuit you would recommend? Or should I say, is there a circuit that will not take up half the board?

PS, the current circuit is not that bad, but I want one with lower noise. Low noise is key.

Thanks!

Can you show whole circuit? What kind of MD detection is it? What is the basic principle?

Regards,

-SB

[Skippy]

There are certainly better analogue switches than the 4000 series ones, more modern devices will have lower on-resistance, less charge-injection and better switch-to-switch matching. Maybe these factors contribute to the improvements you have seen.
If you build the Tayloe circuit up, it is worth considering integrated differential amplifiers, such as the AD628 or INA132 which have excellent precision.

[mario]

Hi Skippy,

Is there an improved drop-in replacement for the CD4066? My searches have not been very successful.

After reading its datasheet, it would appear that this part is a general purpose switch that serves a dual purpose as multiplexer.

There's definitely room for improvement by using true multiplexers. I seriously doubt the newer Minelab models have retained the design since these 4066s are noisy when compared to a dedicated mixer solution.

I like your idea of using the INA132 -it does have some nice specs.

To Simonbaker,

It's my old Minelab PI, you can find some of our findings here at geotech. Rather than overhaul the front end, I would like to eliminate the remaining internal noise.

At some point in the future, I would like to digitize the signal. To get a good resolution, I need the detector to be relatively quiet. Is there a proven circuit you would recommend?

[moodz]

switches

There are faster switches with lower on resistance like 74LVC4066.
regards moodz.

[Mechanic]

Hi Moodz,

That one is has a max voltage spec of 5.5v. It needs to be able to work on +-5v supply, so min of 10v is required. The likes of max 4053 or some adg switches are much better than the old 4066's but they are not drop in replacements.

The source of noise is a funny thing. I just finished off a 2000. Originally the noise at the output of the filtering was 15mv peak to peak, this noise could be masked by turning up the threshold. After replacing the power supply with a very low noise +-5v one and adding some really low esr tantallums to the main supply rails, the noise was reduced down to 5mv peak to peak. Now the threshold can be run much lower and as a consequence weaker/deeper targets will be heard that would have been lost in the 15mv noise. The next dominant source of noise seems to be from the integrators.

Mario, the noise in your video is most likely from the emi received through your coil. The effect on the detector changes with operating frequency. Try disconnecting the front end receive wire and hooking it up to the coil ground. This eliminates all external emi but still has the coil running drawing current from the system as it would normally. Any noise picked up then is internal noise. This can be viewed from the test points on your audio board, with your scope set to 5mv and sweep speed .5seconds/div. The middle testpoint is ground the ones either side are ch1 and ch2.

Cheers Mick

[mario]

Hi Mick,

good to have your feedback on this issue. I will try to take some snap shots when I get home (it's not fun prospecting under 110 degree weather, so I don't mind more benchwork).

Did Minelab solve the integrator noise issues by replacing them with less noisy 4053s?

I don't know if the Tayloe mixer is worth pursuing. Do you think I will have a hard time integrating it into the existing circuit?

[simonbaker]

Hi Skippy,

Is there an improved drop-in replacement for the CD4066? My searches have not been very successful.

After reading its datasheet, it would appear that this part is a general purpose switch that serves a dual purpose as multiplexer.

There's definitely room for improvement by using true multiplexers. I seriously doubt the newer Minelab models have retained the design since these 4066s are noisy when compared to a dedicated mixer solution.

I like your idea of using the INA132 -it does have some nice specs.

To Simonbaker,

It's my old Minelab PI, you can find some of our findings here at geotech. Rather than overhaul the front end, I would like to eliminate the remaining internal noise.

At some point in the future, I would like to digitize the signal. To get a good resolution, I need the detector to be relatively quiet. Is there a proven circuit you would recommend?

I'm not familiar enough with that circuit to recommend anything. I'm interested in general concerning noise in Synchronous Detectors, but I don't understand the source of the noise you are referring to here.

-Regards

[moodz]

Ok I'll see your switch and raise you one more ...

Hi Moodz,

That one is has a max voltage spec of 5.5v. It needs to be able to work on +-5v supply, so min of 10v is required. The likes of max 4053 or some adg switches are much better than the old 4066's but they are not drop in replacements.

The source of noise is a funny thing. I just finished off a 2000. Originally the noise at the output of the filtering was 15mv peak to peak, this noise could be masked by turning up the threshold. After replacing the power supply with a very low noise +-5v one and adding some really low esr tantallums to the main supply rails, the noise was reduced down to 5mv peak to peak. Now the threshold can be run much lower and as a consequence weaker/deeper targets will be heard that would have been lost in the 15mv noise. The next dominant source of noise seems to be from the integrators.

Mario, the noise in your video is most likely from the emi received through your coil. The effect on the detector changes with operating frequency. Try disconnecting the front end receive wire and hooking it up to the coil ground. This eliminates all external emi but still has the coil running drawing current from the system as it would normally. Any noise picked up then is internal noise. This can be viewed from the test points on your audio board, with your scope set to 5mv and sweep speed .5seconds/div. The middle testpoint is ground the ones either side are ch1 and ch2.

Cheers Mick

OK then ... the quad switch HI-201HS by Intersil .... +/- 15 volt. Not quite as fast as the LVC but quite a bit speedier than the old 4066.
http://www.intersil.com/data/fn/fn3123.pdf

cheers from moodz.

[Tepco]

Assuming we would like to replace these mixers and employ a different mixing strategy. What would you say be a better way of getting the audio fidelity?

Most probably the lowest noise analog switch for MD (or anything else having similar frequency response and switching speed requirements) you may have is not a chip, 4066, or any newer generation or type of analog switch, but just a good old, decent performance JFET switch. This may sound bit "retro" but see for example, input sampling gate in HP848x power sensors, dealing whit submicrovolt signal levels, or similar T&M low noise solutions, samplers in very low noise signal sources etc. (bit extreme and usually unnecessary for MD, but can show the point here). Tayloe mixer is interesting configuration, whit lot of pro's and con's, depending on particular application and design, but i personally don't believe it can significantly improve MD design. Also, there are many other noise sources waiting to be eradicated first, high gain input amp,then just movement of poor coil connecting cable can generate more noise than 4066 itself etc.
(Tayloe stuff also relies on low noise frontend)

[Skippy]

If analogue switch on-resistance is thought to be an issue, you could parallel up two '4066 devices, simply piggyback one on top of the other. I'm doubtful about how effective it would be, but.....
Also, as previously suggested, try power-line noise improvements. Large value SMD ceramic caps are now readily available, Sanyo's OSCON electrolytics are impressive. Also isolating the supplies to individual amplifiers/switches etc with a small series resistance, (eg. 22R) can sometimes work well.