field test unit no 001 "model T"

[moodz]

Excellent improvement !! No further comment required.

moodz.

Something I tried that seems to work very well is to use the auto-transformer action of the coil to generate the +ve rail: I prefer to call the +ve battery terminal "ground" and keep the -ve negative . Now when the centre tap of the coil is connected to ground and the mosfet switches one end of the coil to -ve, the other (unused) end of the coil goes up by the supply voltage above ground for the TX pulse duration. Using a fast, HV diode this can be used to juice a capacitor that holds the +ve bias for the amplification stage! I used an HER308 (3A, 1KV, 50pf, 75ns recovery) diode and it doesn't appear to have any discernible effect on flyback duration. Using the same type of diode in series with the mosfet makes a significant improvement to the flyback duration.

[AdrianM]

But was that a differential mono coil Altra? Perhaps you tried capturing some of the +ve flyback from a single ended mono - that would wreak havoc with the decay time. This scheme is quite different. During flyback, the diode is reversed biased by the full flyback voltage and only its recovery time (under 100ns) is added to the overall decay. Charging current into the reservoir capacitor also begins and ends with the TX pulse through auto-transformer action so the supply is settling throughout flyback and detection phases.

The only refinement I can think of might be to place a small value inductor in series between the diode and capacitor to control the charging current.

[Altra]

Thanks,
Way back, I tried that with a mono coil and it generated enough power. But it increased the flyback period quite a bit. I have a circuit to test your idea on. It has five dual op amps and the diode tripler that I incorporated, but it can't handle the load. I let you know.

[AdrianM]

How much load you can put on it depends on factors such as pulse TX width and coil inductance but with a low dropout, low quiescent 5V regulator (like the LM2936 I've used) there should be enough for at least three or four OPAMPS in most applications. 7805's take around 5mA just to keep 'em ticking over which I always feel is a bit greedy. The advantages of this bias generating scheme is that it is phase locked to the TX/RX cycle making free of noise when it counts and also costs next to nothing given that it consists of just a diode and capacitor

Naturally it also applies to a plain (not bifilar) coil that's centre tapped and connected in the same way as above. I'm still trying to prove to myself the merits or otherwise of the bifilar winding

[Altra]

Adrian,

Interesting idea!

Do you have any idea what kind of load it can handle? Have you tried it with a 5v regulator attached?

Regards
Mark

[AdrianM]

Thanks .... at least you wont have to run your TX pulse rate flat out to keep the bias volts up ...

Something I tried that seems to work very well is to use the auto-transformer action of the coil to generate the +ve rail: I prefer to call the +ve battery terminal "ground" and keep the -ve negative . Now when the centre tap of the coil is connected to ground and the mosfet switches one end of the coil to -ve, the other (unused) end of the coil goes up by the supply voltage above ground for the TX pulse duration. Using a fast, HV diode this can be used to juice a capacitor that holds the +ve bias for the amplification stage! I used an HER308 (3A, 1KV, 50pf, 75ns recovery) diode and it doesn't appear to have any discernible effect on flyback duration. Using the same type of diode in series with the mosfet makes a significant improvement to the flyback duration.

[moodz]

Moodz has inspired me to follow up on a project that I have been procratinating on for a long time... a separate TX and pre-amp in a small package that is isolated from the back-end so that it can be placed lower on the shaft relatively close to the TX coil to reduce the coil cable length and thus capacitance. Also it would let me position the back-end and batteries anywhere I want (under the forearm support, etc.) without regard to the coil cable length. I like the possibilities of the differential FE/Coil approach. Using a instrumentation amp such as the THAT1510 also offers some ease of interface capabilities for whatever back-end that you wish to use.

My project will be the same concept as the moodz design, but will use the traditional positive gnd and negative supply approach. This approach has no advantage over the moodz bias generator approcah other than it offers me more interface options as I will control the amplifier reference based on the reference needs of my backend (ADC, uPC, FPGA, analog). My initial backend is going to be the HH2 that I have built. That will give me a good baseline for comparison.

The TX section will use the elements of Carl's HH2 TX. The diff coil will be a centertapped 700uH total inductance (effective 175uH TX). The coil is probably going to be a flat spiral 220mm inside diam with ~18 (36 total) turns of dual wire (or twisted pair as in moodz). The target goal is to achieve a decay to reference with no taget present of >= ~11 usec. Initial modeling suggests that this is easily achieveable.

I will be using the Buur Brown INA217 @ G=56.5 instead of the THAT1500. The THAT1500 has much more BW than the INA217, but I have easy access to the INA217 (free samples from TI).

Moodz, thanks for the inspiration.

Thanks .... at least you wont have to run your TX pulse rate flat out to keep the bias volts up ...

[KingJL]

An alternate variant

Moodz has inspired me to follow up on a project that I have been procratinating on for a long time... a separate TX and pre-amp in a small package that is isolated from the back-end so that it can be placed lower on the shaft relatively close to the TX coil to reduce the coil cable length and thus capacitance. Also it would let me position the back-end and batteries anywhere I want (under the forearm support, etc.) without regard to the coil cable length. I like the possibilities of the differential FE/Coil approach. Using a instrumentation amp such as the THAT1510 also offers some ease of interface capabilities for whatever back-end that you wish to use.

My project will be the same concept as the moodz design, but will use the traditional positive gnd and negative supply approach. This approach has no advantage over the moodz bias generator approcah other than it offers me more interface options as I will control the amplifier reference based on the reference needs of my backend (ADC, uPC, FPGA, analog). My initial backend is going to be the HH2 that I have built. That will give me a good baseline for comparison.

The TX section will use the elements of Carl's HH2 TX. The diff coil will be a centertapped 700uH total inductance (effective 175uH TX). The coil is probably going to be a flat spiral 220mm inside diam with ~18 (36 total) turns of dual wire (or twisted pair as in moodz). The target goal is to achieve a decay to reference with no taget present of >= ~11 usec. Initial modeling suggests that this is easily achieveable.

I will be using the Buur Brown INA217 @ G=56.5 instead of the THAT1500. The THAT1500 has much more BW than the INA217, but I have easy access to the INA217 (free samples from TI).

Moodz, thanks for the inspiration.

[Carl-NC]

I once did a quick-n-dirty uPI design, with a PIC that I think had a 10b ADC. I only took one sample point, and did a 16-pt average with an add'l 16-pt tracking average (giving effectively 256-pt slow average). I did a simple delta for target response.

I was surprised that it worked as well as it did. It would have made at least a decent probe.

[moodz]

Hi moodz,

I recommend using several damping resistors in serial and parallel. Note: the resistors do not like high voltages and you need to serialize them. Every 200V a series resistor might work ok.

I have blown up several damping resistors with high voltages. They look even like a normal resistor but got high impedance without changing their color.

Nice work.
Aziz

Thanks Aziz ..yes I think I have seen resistor arrangments as you suggest in commercial detectors.

moodz

[moodz]

Interesting preamp, don't think I've seen that one before. Great specs. However, the 1nV/rtHz noise spec is wasted; with R5 & R8 = 1k they have 5.6nV/rtHz. You need to get them down below 100 ohms, which implies a dynamic input clamping circuit. Now there's a design challenge!

Of course, if the PIC ADC is 12 bits (I'm being lazy by not looking it up) then this level of noise is probably in the mud anyway. Might want to consider a fast SAR 18 bit ADC, I'm currently playing with a AD7982 and ADS8484. Any idea on what sample rate you are going to use?

Nice work!
- Carl

Hi Carl and thanks ...the dspic30F4012 has only 10 bits ...unfortunately ... so yes the preamp noise specs are of academic interest only in this application. The circuit is more of a "demonstrator/experimenter" however it should be able to match performance of the simpler PI designs around. The TX pulse rate is just over 5 Khz with a 100 us duty cycle. The ADC ( working at approx 750K smps ) grabs 64 samples after turn off and each sample is averaged over 1024 TX pulses. I then end up with post turn off waveform oversampled by 1024 updated at a rate of approx 5 times a second ( 5k / 1024 ) Basically then a digital version of the differential integrator is applied where late samples are subtracted from early samples and compared to an early reference sample that is stored either automatically or on press of button ... I hope to dump the averaged waveform by way of RS232 so coil speed can be measured. The traditional boxcar algorithm had to be modified for the averaging calculation as it used too much ram to average a whole waveform point by point. The current code can average and store waveforms up to 512 points over 1024 samples each or more.( there is only 4k of ram in the dspic )

moodz

[Carl-NC]

Interesting preamp, don't think I've seen that one before. Great specs. However, the 1nV/rtHz noise spec is wasted; with R5 & R8 = 1k they have 5.6nV/rtHz. You need to get them down below 100 ohms, which implies a dynamic input clamping circuit. Now there's a design challenge!

Of course, if the PIC ADC is 12 bits (I'm being lazy by not looking it up) then this level of noise is probably in the mud anyway. Might want to consider a fast SAR 18 bit ADC, I'm currently playing with a AD7982 and ADS8484. Any idea on what sample rate you are going to use?

Nice work!
- Carl

[WM6]

Congratulations for the excellent work.

Tinkerer

Thanks Tinkerer, we all have to be thankful to your excellent work. I look forward to your further succes in PI development.

[WM6]

Hi moodz,

I recommend using several damping resistors in serial and parallel. Note: the resistors do not like high voltages and you need to serialize them. Every 200V a series resistor might work ok.

I have blown up several damping resistors with high voltages. They look even like a normal resistor but got high impedance without changing their color.

Nice work.
Aziz

This happen with metal film or resisting film resistor.
Here we need to use old mass (not spiral milling film on resistor surface) pressed resistor.
Probably hard to find now, maybe in very old parts of big computer.
Most nowadays SMT resistors are the same mass pressed resistor i think.

[Aziz]

Hi moodz,

I recommend using several damping resistors in serial and parallel. Note: the resistors do not like high voltages and you need to serialize them. Every 200V a series resistor might work ok.

I have blown up several damping resistors with high voltages. They look even like a normal resistor but got high impedance without changing their color.

Nice work.
Aziz