Moodz' Awesome Gold Pulse Induction Version 3 - MAGPI V3 Project

[moodz]

Looks great and a completed project.

Can you please post Timing of all the control signals and then a flow of the signal processing.

Also, the PIC32 code.

thanks

Thanks waltr ....

Attached is the source code ....

Theres stuff in there to do LCD / Serial and Rotary Encoder .... WIP.

The timing is set up for a 30us TX followed by a 30us RX. Why 30us ? ... because otherwise the TX fet will approach avalanche and the linear 5 volt reg will current limit and/or overheat.
The damping method will achieve the same timings with higher TX currents / voltages ... but obviously with a TX circuit and damping that can handle the higher currents and voltages.

.... built with NMOS / or SIC is the bees knees.

There are 2 feedback loops

Loop 1 ... the damping loop
The BSP230 mosfet performs the active damping and has a capacitor at its gate that stores the bias level fed by a sampling feedback loop. The sampling feedback loop takes a sample from the flyback error amplified by the preamp LM6171 ... this is the X sample.
Typically the X sample is taken just after the optimal flyback damping should hit 0 volts and 0 amps in the TX coil this the zero damping point ZDP ( use the SFR of the coil to get this ). The feedback loop sampling either can use a simple resistor ( eg 1k ) OR a peak diode detector & resistor ( eg 1k ) and leaky resistor ( eg 100k ).
If a simple resistor is used then the X sample can be set to say 100 nanoseconds at around at about 100 - 200 nanoseconds past the ZDP. If the diode peak sampling is used then the X sample window can be quite wide ( eg 10 us ) as the diode optimally samples the error peak.

Loop2 ... noise and offset reduction loop
The TX pulse is used as a control sample DG411 during the RX period to feed a noise / DC offset error correction via the NE5534 to the "ground" reference of the LM6171. This removes substantially all DC offset errors and low frequency mains / magnetic field influences etc in the signal.

Target Detection ...
Two samples DG411 drive the integrators feeding each input of the differential amp INA125 the output going the ADC on PIC32 via a simple low pass filter. As a baseline setting ... the RX time ( in this case 30 us ) is simply divided in half so that the first sample A is taken for 15 us .. then the second sample B is taken for 15 us where the total sample time is 30 us.
Surprisingly ... for a PI this "baseline" gives very good performance. This is the baseline setting for V3 with no controls fitted. ie basic detector. There is code in there to operate an LCD / Serial and Rot Enc.

Sound Generation ...
A single cycle of sine waveform is calculated and stored in an array at a particular volume. ( amplitude ) when the code boots. The cycle is played of an OC port pin as a PWM waveform by DMA using the repeating the single cycle sine over and over. The OC PWM timing is driven by clock timer 4. A DMA complete interrrupt is generated at the end every sine cycle "played" out.
The DMA interrupt handler basically reads the current averaged ADC target value and sets the frequency of timer 4 according to the value of the average target value. In this way a variable sine wave VCO is generated with relatively little CPU overhead. The volume can be adjusted by recalculating the amplitude of values stored in the single cycle array.

thats about it ..

PS ... the attached code is for a PIC32MX150 not 250 variant chip ... the MX150 is a little cheaper ( no USB ) ... but I dont use USB ... so etc .... ordered by mistake ... works the same otherwise.

PPS .. I use the v2.15 compiler ;-)

moodz

MAGPI008_2023_01_28_MX150.zip

[lucifer]

Loop2 ... noise and offset reduction loop
The TX pulse is used as a control sample DG411 during the RX period to feed a noise / DC offset error correction via the NE5534 to the "ground" reference of the LM6171. This removes substantially all DC offset errors and low frequency mains / magnetic field influences etc in the signal.

[ATTACH]n409755[/ATTACH]

Hi moodz,
Is that loop worth it? Is it really reducing noise? Isn't more noise (generated by the DG411 switching + noise generated by NE5534) injected into the LM6171? I think I've seen this in other PI circuits and always wondered if it really does any good.

[moodz]

Hi moodz,
Is that loop worth it? Is it really reducing noise? Isn't more noise (generated by the DG411 switching + noise generated by NE5534) injected into the LM6171? I think I've seen this in other PI circuits and always wondered if it really does any good.

Hiya Lucifer ... refer the two circuits below ... identical except one amp is only referred to ground. There are white noise sources in each coil.

The top blue trace is the grounded amp. :-)

moodz

[lucifer]

Thanks. Is that what happens in practice too? I don't always trust simulations. Is your preamp's output nice and clean?

[moodz]

see this post .. the actual circuit ... https://www.geotech1.com/forums/foru...560#post408560
post #11

[Carl-NC]

The sampled feedback forms a comb filter with the response

H(f) = 2 |sin(π f T_s)|

where T_s is the sample period. The response looks like:



You get a perfect null at DC, plus nulls at every 1/T_s. You can calculate the exact amount of EMI suppression using the equation. For example, at a 5kHz pulse rate the 50 Hz mains is attenuated by

2 |sin(π*50Hz*200us)| = 0.0628 = -24dB

It's not a complete panacea because some EMI frequencies also get boosted; this is where a variable pulse rate is valuable. If the sample is taken after the target signal has decayed to zero then it has no effect on the target response. But for high conductors that have not fully decayed there is some attenuation which depends on the delays of the target sample and the feedback sample.

This technique has exactly the same effect as the classical EFE subtraction sample. The advantage here is that it's applied to the preamp so that you never have to worry about offset headroom problems. The disadvantage is that the nulls are set by the TX pulse rate, and with EFE subtraction they are set by the difference of two sample times which gives you more flexibility in placing the nulls. You could combine the two methods and get the best of both.

[KingJL]

... here are the kicad files.

Great work moodz!
After studying the Kicad files, I do have some questions:
(1) Can you provide "BUILD NOTES FOR R19,C18,R1,R20,D8"?
(2) Is "INSTALL 9 VOLT ZENER AT R19 ANODE TO GROUND" still valid?
(3) Is "USE 0.1u FILM for C1" still valid?

Do you think this would make a good pinpointer if implemented in SMD? I personally think it would make a great one.

[moodz]

Great work moodz!
After studying the Kicad files, I do have some questions:
(1) Can you provide "BUILD NOTES FOR R19,C18,R1,R20,D8"?
(2) Is "INSTALL 9 VOLT ZENER AT R19 ANODE TO GROUND" still valid?
(3) Is "USE 0.1u FILM for C1" still valid?

Do you think this would make a good pinpointer if implemented in SMD? I personally think it would make a great one.

Thanks Jim .. the circuit is simple enough and could be further trimmed to make a compact PP. I have used a 300 uH ferrite inductor and it works great.

Here is a modified schematic to clarify ... the PCB still works ok. ( ie not modified ). The diodes are rearranged from a inverter to an inverting doubler to cater for lower lithium voltage input ( 8V instead of 12 volts ).

[moodz]

(2) Is "INSTALL 9 VOLT ZENER AT R19 ANODE TO GROUND" still valid?

The 9.1 volt zener is not strictly necessary ... its a safety net ... if the bias on the gate of the BSP230 mosfet goes too negative or floats the flyback / voltage / current can be conducted into the preamp or blow the 1N1418 or the mosfet itself ... very annoying an usually no warning.

Usually happens when probing around the circuit

So if you feel lucky dont fit it.
moodz

[KingJL]

Thanks Jim .. the circuit is simple enough and could be further trimmed to make a compact PP. I have used a 300 uH ferrite inductor and it works great.

Here is a modified schematic to clarify ... the PCB still works ok. ( ie not modified ). The diodes are rearranged from a inverter to an inverting doubler to cater for lower lithium voltage input ( 8V instead of 12 volts ).

Paul, thank you for your answers... the rearranged diodes a great change. I think I will layout a compact pin pointer. is a great

[moodz]

Thanks Jim.

[moodz]

..heres the HEX file for the CPU.

MAGPI008.X.production.V3HW.zip

[lucifer]

The sampled feedback forms a comb filter with the response

H(f) = 2 |sin(π f T_s)|

where T_s is the sample period. The response looks like:



You get a perfect null at DC, plus nulls at every 1/T_s. You can calculate the exact amount of EMI suppression using the equation. For example, at a 5kHz pulse rate the 50 Hz mains is attenuated by

2 |sin(π*50Hz*200us)| = 0.0628 = -24dB

It's not a complete panacea because some EMI frequencies also get boosted; this is where a variable pulse rate is valuable. If the sample is taken after the target signal has decayed to zero then it has no effect on the target response. But for high conductors that have not fully decayed there is some attenuation which depends on the delays of the target sample and the feedback sample.

This technique has exactly the same effect as the classical EFE subtraction sample. The advantage here is that it's applied to the preamp so that you never have to worry about offset headroom problems. The disadvantage is that the nulls are set by the TX pulse rate, and with EFE subtraction they are set by the difference of two sample times which gives you more flexibility in placing the nulls. You could combine the two methods and get the best of both.

Thank you, Carl. That is a great explanation!

[Dean Sarelius]

[Dean Sarelius]

..heres the HEX file for the CPU.

[ATTACH]n409851[/ATTACH]

Hi Moodz,
Just curiours, which LCD and rotary encoder did you use, or were you considering to use..?