NEW !!! Full differential PI front end.

[simonbaker]

Question

I think this has been asked before, but not sure...

With differential design, we put current through half the coil, but use whole coil to detect target signal. Advantage is less flyback voltage across MOSFET, right?

But.... what about magnetic field? Is it half of that with non-differential coil? Or do we try to pump twice the current into it?

What I'm asking is: is our magnetic field weaker with differential design, so we get less target stimulus?

You might say: yes it is weaker, but with non-differential design it was wasted anyway because MOSFET breakdown.

Reply to that is -- use slower cut-off to avoid MOSFET breakdown, and we'll get longer (therefore stronger) target stimulus -- at least for targets that are not extremely "fast".

So does differential design have weaker transmitted magnetic field pulse?

Regards,

-SB

[moodz]

Hi Paul,

I have some true instrumentation amplifier chips at home (INAxxx). I would like to try out these first.

Aziz

Cool ... you will be able to give them a good work out with the new CRO

moodz

[Aziz]

Hi Paul,

I have some true instrumentation amplifier chips at home (INAxxx). I would like to try out these first.

Aziz

[Aziz]

Hi Paul,

Aziz .... I am waiting for that amplifier design ha ha

moodz

It will take some time. I am not ready at the moment due to lack of time and other things to do.

Why not operating the differential amplifier with bipolar power supply? This would increase the performance.

Aziz

[moodz]

Paul,

This kind of circuit often works extremely well in sims but not in real life. Ferinstance, you would have to run the sim a long, long time to see the effect of leakage currents on the floating node. Also, the fact that it comes up to a decent DC value means little unless you run a transient sim that includes parasitics and ramping up the power supplies, and ramping them up at different rates. That floating node can easily end up railed out.

Are you actually running a real circuit this way? I would suggest a high-resistance voltage divider on the + input to guarantee a correct bias point.

- Carl

Hi Carl ... you are right ... the amplifier does not work in practice ... I reboarded the circuit on the v5 schematic and the opamp refused to bias at all on the new layout. Then I was thinking about my previous comment regarding the protective diodes on this type of amp that are present at the inputs to protect the sensitive mos gate. With previous layouts the first amp was very close to the switching mosfet ... and it's the high voltage transient spikes that were getting into the front end of the opamp causing the protective diodes to avalanche and basically conduct to V+ and charge the cap thus provide a sort of 'switched' bias. I never saw the transients beacause they appeared as common mode and were too fast for the amp to respond to. With version 5 the high voltage no longer gets near the reciever because of the separate shielded pair pickup from the clamping and damping which are now located at the coil. So there are no transients to avalanche the protective diodes. So the amp never biases. To test this theory I connected a 1M resistor from the hot side of Q1 to the non inverting input and sure enough the bias cap charged in couple of seconds ( not via the 1M but via the avalanche diodes ) and the amp worked like a charm. It keeps working after the resistor is removed because the offset current in these amps is 5 pico amps so the cap would take maybe some hours to discharge ... and I only work the circuit for an hour or two.
Well now I know three things ...

1. Carl knows what he is talking about.
2. My old layout suffered from transients.
3. Must ring the patent office and put a hold on that Novel OPAMP biasing technique.

... its nice to know why it was happening.

Aziz .... I am waiting for that amplifier design ha ha

moodz

[moodz]

Hi Paul,

excellent work! I also would like to try out some full differential PI front end's soon. I will do this with bipolar power supplies due to its simplicity (biasing) and increased noise immunity performances. But it will take a few weeks to start with.

The differential stage could also be done with discrete solution with highly matched transistor/FET pairs (see SD2000 schematics). This should lower the noise further.

You did really a good job. Thanks for sharing your work.

Regards,
Aziz

Thanks Aziz .... can you work on the amp ... I need one

Regards,

moodz.

[moodz]

Hi Moodz:

Which "non-differential" circuit would be the closest to compare to help understand the benefits and workings of your differential design? Can you provide a schematic or link to one?

Regards,

-SB

Probably the Goldpic by Trevor Hill .... however on second thoughts having looked at the MineLab SD2000 ... you could compare it with that ... sort of a 'gold standard' ( no pun intended ) to aim for. I would want at least that level of sensitivity or better. I think you could easily google detail of both.
In terms of SNR the differential circuit is at least twice as good as a single ended design ... have a look at pulse amps used on disk drive heads etc. Now that I have split the Tx path from the Rx path by moving the "clamping and damping" ... hey Im a poet ... to the coil head .. this means the TX and RX circuits can be well isolated from each other and the design of each optimised. I was tempted to move the RX actives to the coil as well ... but there is the law of diminishing gains ( no pun intended ha ha ) .. and I would like the RX in the box so to speak.

Regards,

moodz

[moodz]

Hi Carl ... I never as a rule rely on spice sims ... when I ran the sims above I fully expected it not to work as I was expecting the results I was getting on the real circuit to be dependent on something in the front end of CA30XX devices ... like the protective diodes on the mosfet gates at the +/- inputs. However it does seem to work ... and I run the prototype circuits for as long as my battery pack has charge .. 2 - 4 hours while I poke around with the CRO and burn my hand on the soldering iron. There is no noticeable drift in the bias point. None of the schems I have published are based on spice or cut and paste circuits ... I have built all of them ... I could publish 10 times as fast if I just spiced everything.
I ran some sims on older devices like the LM740 and these work but very poorly ... bias point near zero.

I am mainly concerned with getting the coil bit right ... the diff amp is not rocket science .. there are much better amp designs ... Aziz ?? If I can contribute to improving the front end then I will be happy.

Regards,

moodz.

[Carl-NC]

Paul,

This kind of circuit often works extremely well in sims but not in real life. Ferinstance, you would have to run the sim a long, long time to see the effect of leakage currents on the floating node. Also, the fact that it comes up to a decent DC value means little unless you run a transient sim that includes parasitics and ramping up the power supplies, and ramping them up at different rates. That floating node can easily end up railed out.

Are you actually running a real circuit this way? I would suggest a high-resistance voltage divider on the + input to guarantee a correct bias point.

- Carl

[simonbaker]

Hi Moodz:

Which "non-differential" circuit would be the closest to compare to help understand the benefits and workings of your differential design? Can you provide a schematic or link to one?

Regards,

-SB

[Aziz]

Hi Paul,

excellent work! I also would like to try out some full differential PI front end's soon. I will do this with bipolar power supplies due to its simplicity (biasing) and increased noise immunity performances. But it will take a few weeks to start with.

The differential stage could also be done with discrete solution with highly matched transistor/FET pairs (see SD2000 schematics). This should lower the noise further.

You did really a good job. Thanks for sharing your work.

Regards,
Aziz

[moodz]

amp with ideal op amp

.. amp simulation with single supply, cap on bias leg and IDEAL opamp. Notice that the bias point is EXACTLY vsupply/2.

Supply = 10 volts.
Bias point = 5 volts.
Pulse input = 100uS duty cycle
Period = 200uS
Pulse amplitude =100 mv pp

I am happy enough that this self biasing is a valid design technique for AC amplifiers.

real opamps will exhibit a slight DC offset but if everything is AC coupled ... no problem.

Any comments welcome.

moodz

[moodz]

v5 errata 2.

...eek .. should be a 1k resistor from top of Q1 to ground. Play with the value a bit ... depends on your MOSFET capacitance ... will improve the decay some.
Will publish in next schem update.

moodz.

D1, 2, 4, and 5 are 1n4148 ... not schottky as illustrated.

moodz

[moodz]

v5 ommission

...eek .. should be a 1k resistor from top of Q1 to ground. Play with the value a bit ... depends on your MOSFET capacitance ... will improve the decay some.
Will publish in next schem update.

moodz.

[moodz]

... and now with single supply 10 volts ... and cap .... and the o/p self biases to approx V supply ( ie 5 volts ) and the pulse train looks OK.

... any comments welcome.

regards

moodz.

That should read ... and the o/p self biases to approx HALF V supply ( ie 5 volts ) ...

moodz