field test unit no 001 "model T"

[moodz]

Cool.

I might wait until you have the diff amp section done and have posted a revised schematic, then I'll add the diff amp and see what happens.



Mick

Mick ... I have been able to reliably recreate the fault and a fix is in place ( very simple ). I will do some more testing and include in next drawing. The problem was to do with the damping mosfet latching up during turn on and never being within control range of the damping control voltage It also causes the flyback voltage to reach excessive voltages which can damage the blocking fets.

moodz

[mickstv]

Mick ... I thought that might work ... the CRO should not be the issue ... though a multichannel digital CRO is more about saving time than resolving issues.

It appears that the damping is very critical to the descrimination process .... variable damping is required to set the "balance point" .... PI detectors to date have ignored this ... most designs just set the damping to some value to try and optimise the coil decay ... however I have come to realise that it is absolutely critical to obtaining the correct operating point ( well for my design anyway )

The second factor is that the blocking FETS source connection is effectively floating with respect to the gate and so can be easily upset by any voltage imbalance that occurs across these devices ... this will either cause the fets to stop blocking and switch high voltage through to the demod or punch through the FET gate. The other factor to consider is that any imbalance will upset the damping value which is extremely sensitive when at the balance point.
Even the high impedance probes on a CRO can upset this balance point by loading a very small current through the blocking FETs ....

I am looking into a more robust set of balanced FETS with 500 volts of isolation.

Below is a shot of the current circuit being probed on the bench.

I have not drawn the circuit yet ...as I only got it up this morning .... I will try an complete the DC amp this weekend .. unless I have to get some parts

moodz

[ATTACH]17281[/ATTACH]

Cool.

I might wait until you have the diff amp section done and have posted a revised schematic, then I'll add the diff amp and see what happens.



Mick

[moodz]

Using battery now looks like it fixed the waveforms, also removed the damping circuit and relaced with a 1k wirewound pot. CRO isn't able to show any difference between ferrous and non ferrous metals at this stage, but having said that I have never been able to replicate the results you have shown Moodz. Maybe part of my problem is still to do with the Analog CRO ?



Mick

Mick ... I thought that might work ... the CRO should not be the issue ... though a multichannel digital CRO is more about saving time than resolving issues.

It appears that the damping is very critical to the descrimination process .... variable damping is required to set the "balance point" .... PI detectors to date have ignored this ... most designs just set the damping to some value to try and optimise the coil decay ... however I have come to realise that it is absolutely critical to obtaining the correct operating point ( well for my design anyway )

The second factor is that the blocking FETS source connection is effectively floating with respect to the gate and so can be easily upset by any voltage imbalance that occurs across these devices ... this will either cause the fets to stop blocking and switch high voltage through to the demod or punch through the FET gate. The other factor to consider is that any imbalance will upset the damping value which is extremely sensitive when at the balance point.
Even the high impedance probes on a CRO can upset this balance point by loading a very small current through the blocking FETs ....

I am looking into a more robust set of balanced FETS with 500 volts of isolation.

Below is a shot of the current circuit being probed on the bench.

I have not drawn the circuit yet ...as I only got it up this morning .... I will try an complete the DC amp this weekend .. unless I have to get some parts

moodz

[moodz]

DIFFERENTIAL DEMODULATOR SUCCESS

After a fair amount of stuffing around ... the differential demodulator frontend is now operational ..... this includes the blocking fets and damping circuit ( I thought the damping circuit might have been dodgy ... but it isnt .... its all about critical balance )

The gain of this frontend is only 1 .... the design and timing removes the need for low noise high speed fast recovery amps in the frontend whilst still offering sub microvolt accuracy.

The output is differential ...so now we need a good DC differential amp ( )

Output noise is extremely low and narrowband ... no resistors used in frontend at all ... ( you once challenged me to do this Carl .... so its done now he he )

The descrimination is excellent for ferrous / non-ferrous ..... see pics below ... note polarity&shift of demod ouput voltages.

The next task is to get the precision DC amplifier going ( chop chop )

After the DC amp task is complete we will have a complete detector signal chain ... sigh.

moodz

[mickstv]

Using battery now looks like it fixed the waveforms, also removed the damping circuit and relaced with a 1k wirewound pot. CRO isn't able to show any difference between ferrous and non ferrous metals at this stage, but having said that I have never been able to replicate the results you have shown Moodz. Maybe part of my problem is still to do with the Analog CRO ?



Mick

[moodz]

It's a 12 volt smps. Sounds like I had better charge up a battery pack and see what happens.

Another thing all my tests have been done at 50us TX pulse. Are you able to check yours running at 50us and see how the descrim work.



Mick

Mick ... I was able to replicate your results ... the damping circuit is flakier than I thought .... however I blew the CPU somehow when I was testing so I will change the chip over and rehash the damping ( you could always try a simple 500 ohm resistor ... however the variable damping is essential to balance the detector ).

The descrim works at 50 , 100 , 150 us ... stronger response for longer pulse lengths ... I have not tested but would probably find that the pulse length will change response for some types of targets.

moodz

moodz

[mickstv]

Mick ... are you using a battery or a 240 volt supply to power the circuit .... the ground on your supply may be "floating" and dragging the ground down / up after flyback thus charging the gate and turning the TX mosfet back on causing it to conduct again.

moodz.

It's a 12 volt smps. Sounds like I had better charge up a battery pack and see what happens.

Another thing all my tests have been done at 50us TX pulse. Are you able to check yours running at 50us and see how the descrim work.



Mick

[moodz]

Mick ... are you using a battery or a 240 volt supply to power the circuit .... the ground on your supply may be "floating" and dragging the ground down / up after flyback thus charging the gate and turning the TX mosfet back on causing it to conduct again.

moodz.

[moodz]

Maybe my coil just isn't fast enough. I'll wind another tomorrow and instead of tight bound I'll wind a spider and see if I can reduce the capacitance.

Although I do remember Moodz saying he also tried a cat5 wound coil and it worked better than the coil he normally uses ?



Mick

I saw the photo of your coil and it is fine ... identical to the one I made up except it is the orange pair ... it seems that the something is conducting the current after the mosfet has turned off.

moodz

[mickstv]

Maybe my coil just isn't fast enough. I'll wind another tomorrow and instead of tight bound I'll wind a spider and see if I can reduce the capacitance.

Although I do remember Moodz saying he also tried a cat5 wound coil and it worked better than the coil he normally uses ?



Mick

[Midas]

But hang on, haven't we already ruled out the blocking fets as the cause when Mick disconnected them completely and still got the same result in back in post #81 ?

[mickstv]

Hi Paul, do you have an old Analog CRO that you could use to compare my pictures ?

The PHC2300 packs have protection diodes between drain and source maybe this is the problem ?



Mick

[moodz]

Hi Moodz,
Yeah right, sorry it was rude of me to assume that it wasn't a design decision. I understand you're trying to keep it as simple as possible.
But I definitely think the mosfet driver is worth the extra complexity. Even your implementation may be perilously close to unwanted mosfet turn on possible only needing a IRF740 (they have a wide gate threshold variation 2-4v) from a different batch or longer traces to tip the balance. The other issue with direct drive is that the IRF740 is far from being fully turned on at 5v either which will reduce maximum coil current capability and power efficiency. Then of course switching times are increased also extending the time the fet spends in its linear region wasting more power.

No ... no offence taken .... it could be what you mentioned. ... there isn't much else to try out unless I actually send Mick my working example ... I will be trying out Micks setup on the weekend ... I have previously been using out of production mosfets for the blocking function. ... maybe there is a "feature" on the NP mosfet pack that Mick is using.

moodz

[mickstv]

Hi all, I've pretty much tried everything including different coils, I also tried running it with a TC4422a driver IC made no difference to the outcome.

All other components bar the blocking fets are exactly the same values as Paul's second design. So unless the frontend schematic has a mistake in it I have no idea why the circuit isn't working correctly for me, well unless the analogue CRO isn't capable of displaying the responses correctly.

Maybe I should wait until Paul has the frontend finished including the diff amp then try again.




Mick

[Midas]

Hi Moodz,
Yeah right, sorry it was rude of me to assume that it wasn't a design decision. I understand you're trying to keep it as simple as possible.
But I definitely think the mosfet driver is worth the extra complexity. Even your implementation may be perilously close to unwanted mosfet turn on possible only needing a IRF740 (they have a wide gate threshold variation 2-4v) from a different batch or longer traces to tip the balance. The other issue with direct drive is that the IRF740 is far from being fully turned on at 5v either which will reduce maximum coil current capability and power efficiency. Then of course switching times are increased also extending the time the fet spends in its linear region wasting more power.