NEW !!! Full differential PI front end.

[simonbaker]

Weeeeeeeeeeeeeeeell .... how do I explain this.

Simon ... even a single coil is 'differential' it all depends on what your point of reference is for measuring the tx or rx signal. With so called non differential designs one side of the coil is connected to ground or supply and the other side to a switch. Well by connecting one side to ground or supply you have just made your differential coil unbalanced. ie you will have to measure everything with respect to ground or supply.
Actually during TX the design above is unbalanced in a technical sense as the mosfet only drives one coil effectively ... trace DC through circuit with switch ON. However the field induced in the TX side also induces a current on the other side and the circuit balances.

Be aware also that we are not balancing mag fields here ... ie nulling out like a VLF where the coils have to be physically offset.

There is only a single coil of twisted pair used here.
First conductor is called say A - a
Second is called B - b

So when you have wound your coil you will have two ends made of two conductors .... AB and ab

The so called centre tap ( I dont like that term ... there is no 'centre tap' .. it is a single biflar coil ) is really a cross connect between pairs.

ie you take the 'a' wire at one end of your coiled pair and connect it to the 'B' wire at the other end this forms the centre tap in the schematic.

going on with the story ... during recieve you can imagine that the mosfet disappears as it is turned OFF. look at the coils are connected in additive series ie A - aB - b so any induced signal in each coil will add. If they were connected AB - ab the coils would be in parrallel so this would be the same as a single coil ... and if they were connected A -ab - B they would be opposed ... so any induced EMF in one coil would be equal and opposite thus balancing out .. but resulting in no recieved differential signal .. but you would have a common mode signal ( ie same at both ends ).

Lots of systems would not work if the differential circuit did not exist.

eg ethernet, telephone etc etc.

moodz

Ok, I was barking up the wrong tree. Would you summarize again what problems are you trying to solve with the differential design?

Regards,

-SB

[moodz]

The thing I don't get about these "differential" designs is that it seems to me that both coils will receive the target signal -- so the target signal will essentially become part of the common mode signal and be canceled out by the differential circuitry.

With PI designs, target isolation is normally done by time difference. In VLF designs, target isolation is done by "nulling" the coils.

If this differential design works for PI detectors, then wouldn't we be using it also for VLF designs to avoid all that messing nulling?

(One of the big advantage of PI designs is the ease of making the coils -- no finicky nulling.)

Can anyone explain how this center-tapped coil and differential circuitry avoids canceling out the target signal?

Regards,

-SB

P.S. I'm probably missing the point of the differential circuit - I thought it was to try to cancel out the big TX pulse that gets into the RX circuit.

Weeeeeeeeeeeeeeeell .... how do I explain this.

Simon ... even a single coil is 'differential' it all depends on what your point of reference is for measuring the tx or rx signal. With so called non differential designs one side of the coil is connected to ground or supply and the other side to a switch. Well by connecting one side to ground or supply you have just made your differential coil unbalanced. ie you will have to measure everything with respect to ground or supply.
Actually during TX the design above is unbalanced in a technical sense as the mosfet only drives one coil effectively ... trace DC through circuit with switch ON. However the field induced in the TX side also induces a current on the other side and the circuit balances.

Be aware also that we are not balancing mag fields here ... ie nulling out like a VLF where the coils have to be physically offset.

There is only a single coil of twisted pair used here.
First conductor is called say A - a
Second is called B - b

So when you have wound your coil you will have two ends made of two conductors .... AB and ab

The so called centre tap ( I dont like that term ... there is no 'centre tap' .. it is a single biflar coil ) is really a cross connect between pairs.

ie you take the 'a' wire at one end of your coiled pair and connect it to the 'B' wire at the other end this forms the centre tap in the schematic.

going on with the story ... during recieve you can imagine that the mosfet disappears as it is turned OFF. look at the coils are connected in additive series ie A - aB - b so any induced signal in each coil will add. If they were connected AB - ab the coils would be in parrallel so this would be the same as a single coil ... and if they were connected A -ab - B they would be opposed ... so any induced EMF in one coil would be equal and opposite thus balancing out .. but resulting in no recieved differential signal .. but you would have a common mode signal ( ie same at both ends ).

Lots of systems would not work if the differential circuit did not exist.

eg ethernet, telephone etc etc.

moodz

[porkluvr]

This would be so much more clear if there were dots to indicate proper coil phasing. What is it, the new convention to leave evertbody guessing? Too much trouble, or condescension?

Aww, forget it. Everybody knows what bifilar wound means so phase dots would only be redundancy.

[simonbaker]

Yes, now that's differential! I played with this a few months, got almost 1000v diff., and faster settling. Has promise.

The thing I don't get about these "differential" designs is that it seems to me that both coils will receive the target signal -- so the target signal will essentially become part of the common mode signal and be canceled out by the differential circuitry.

With PI designs, target isolation is normally done by time difference. In VLF designs, target isolation is done by "nulling" the coils.

If this differential design works for PI detectors, then wouldn't we be using it also for VLF designs to avoid all that messing nulling?

(One of the big advantage of PI designs is the ease of making the coils -- no finicky nulling.)

Can anyone explain how this center-tapped coil and differential circuitry avoids canceling out the target signal?

Regards,

-SB

P.S. I'm probably missing the point of the differential circuit - I thought it was to try to cancel out the big TX pulse that gets into the RX circuit.

[moodz]

Yes, now that's differential! I played with this a few months, got almost 1000v diff., and faster settling. Has promise.

Thanks Carl & Hobbes ...

I did some comparisons to the single ended designs I looked into and I am getting nearly twice the voltage differential as I did with single. Rather nice as the mosfet only sees half of that voltage as it sits on only one of the hot sides. 750 volts approx differential ( ie 350 + on the mosfet with respect to ground. Settling was improved but not by a factor of two. Any improvement is good right ?

Anyway I made some more changes as the previous circuit wasnt so hot ( or rather too hot ) and I blew up a couple of amps. Also got better balance by getting rid of the two damping resistors and replacing with single resistor. compare schem below to one posted previously.

BTW I am not an agent for CA3130 amps ... Its what I have most of.

moodz.

[Carl-NC]

Yes, now that's differential! I played with this a few months, got almost 1000v diff., and faster settling. Has promise.

[hobbes_lives]

Nice work. That's definitely something I'll be experimenting with at some point.

[moodz]

scope pics moodzPI differential front end v300a

Green measured at point b
White at point A

obviously there is no amplification here.

There are also more elaborate / different ways to apply the TX drive sigs.

pic 1
NO TARGET
flybacks peak at +100 and -100 volts.
note the convergance of each decay to the baseline.


pic 2
Close up of decay baseline
NO TARGET


pic 3
Close up of decay baseline
Copper target
Note the changed convergence of the +/- decay lines
Just what a diff amp likes
Note how the change is symmetrical


The diff amp section is currently under construction .. stay tuned.

moodz.

[moodz]

This is the ONE. It works. Its differential. I am happy.

[moodz]

Hi Moodz,
what about effecet of size of metal to hight of peak.
does the hight of peak change according to metal size.or size just effects width of peak?
Also what about effect of distance ( target to coil) to the hight of peak?

our variables are too much, we have to consider them all .

what do you use after this above circuitry, multiplexer and S&H caps then outer Ato C circuitry ( counter related to chip)?

Hi unreg,
the size and distance to target both effect the height of the peak 1/d squared approx. The baseline to the right (where most designs start sampling) of the peak also increases but not as much as the peak response.
There are targets that do not move the peak at all ... but do move the baseline. If I move the sampling point earlier the peak will decrease for copper targets but wont register ferrous at all.
These are all bench tests ... may not even work effectively on real ground.

My backend processing is fairly conventional .... multipoint sample and amplification fed to CPU.

moodz
I am working on a much faster monocoil ( < 5 uS delay ) and the use of novel amplifier usually employed in HF DSP radio.

Hi Moodz,
what about effecet of size of metal to hight of peak.
does the hight of peak change according to metal size.or size just effects width of peak?
Also what about effect of distance ( target to coil) to the hight of peak?

our variables are too much, we have to consider them all .

what do you use after this above circuitry, multiplexer and S&H caps then outer Ato C circuitry ( counter related to chip)?

[moodz]

Hi Moodz ,
thanks for explanationss.

I am confused about pictures.

at previous posts ,picture for copper,I choose this because it has lowest peak ,lower than also no target option.
at 10us peak is 2V
at 50us peak is 1,2V this is good.

but at your last picture
at 12us peak is 0,6V
at 15us ,it is 0, disappears.

maybe I am reading scope ,in wrong way, excuse me than.

regards/thanks for your understanding

Hi okantex ... the last picture was intended to show how the sample is taken from the falling edge of the flyback spike rather than being cut off by clamping diodes. Also the CRO set to 2 volts/div 5 uS/div rather than 1 volt / div 10uS/div in previous pics. I dont think there was a target in the coil at the time.
It is important that the fly back pulse effectively gets narrower in presence of good conducting target ( sample peak goes low ) ... wider with ferrous target ( sample pulse goes higher ) ... the base line will also lift in both cases.

moodz

[okantex]

Hi Moodz ,
thanks for explanationss.

I am confused about pictures.

at previous posts ,picture for copper,I choose this because it has lowest peak ,lower than also no target option.
at 10us peak is 2V
at 50us peak is 1,2V this is good.

but at your last picture
at 12us peak is 0,6V
at 15us ,it is 0, disappears.

maybe I am reading scope ,in wrong way, excuse me than.

regards/thanks for your understanding

[Carl-NC]

PS I have noted your comments about referencing offsite resources Carl ... thanks.

Yikes! Shrink those pics! I have auto-scaling turned off so pics come in at their full size. So you need to edit them for size before posting. A width of 800 works well:

[moodz]

Hi Moodz,

congratulations for your improvements.
I am an electronist , dealing as hobby,
I was inspection scope results.your last pictures are interesting , there is no cutting of high peak ( fly back cut-off).why there is no flyback ?
okantex

Thanks okantex ...

WHITE is the flyback pulse 175 volts peak .. you wont see the peak here obviously with 2v/div. We are looking at the base of the pulses here.

GREEN is the sample.

note the sample is not amplified .... it is o/p of sampling stage and is the input to my processing stage.

The peak you are looking at is part of the flyback. See the pic above ... the sample is taken from the falling ''wall" of the flyback. I have offset vertically the sample and flyback however you can see from the shape that the sample tracks the flyback exactly in level on the falling side. I dont use clamping diodes at all in the front end in my current scheme. The use of clamping diodes throws away the useful signals on the falling curve.

The other thing about clamping diodes is that they unclamp ( ie provide useful signal only when the forward conduction voltage is below 600mv etc )
This could occur at any time depending on the slope of the curve. However targets change the shape of the curve thus the point at which the diode unclamps ... start of sampling if you will ... is not fixed in time. My frontend sampler always begins sampling at a fixed time. ( not too early of course or I will dump high voltage into my front end ... i have not managed to do this yet.)

Note also ... some may think I am simply attenuating the flyback peak down to the sample level ... this is not correct the sampler is a switch that switches at a predetermined time after the tx pulse has ended. Approx 5 us here.

note also the peak of the sample intercepts the curve at the correct level.
( noting that I have moved the position of the flyback up slightly so that the sample trace is not obscured. )

note the 'soft' rise time on the rising edge of the sample ... this is good for me as there is less hf energy in the pulse to be processed and I can use slower op amps.

cheerz from moodz

PS I have noted your comments about referencing offsite resources Carl ... thanks.