AGD25.1 PI Detector Development Work

[Auto-Mation-Assist]

This the start of my work on my AGD25.1 PI development project.

I have been doing quite a bit of simulation work for improvements to my prior versions of my AGD23 and AGD24 PI detectors all of which has had a number of variants. Some more sensitive than others with the best able to detect 0.015 grams of gold at about two inches in air tests which I think is pretty good considering a somewhat noisy environment.

But I want to make some changes to the design that I think would be beneficial and thus I start the AGD25.1 design project. I will post all development info here for all sections of the proposed AGD25.1 PI detector. This includes the Qspice simulation files, pictures of the schematics, graphical plots and pictures to help explain circuit behavior.
Over the years I have spent a good deal of time looking at coil decay times and TX/RX switching methods. As a result, I still I prefer those methods do not require any sort of digital pulses for switching from transmit to receive mode at the coil in a PI detector.

This choice only leaves the resistive method and clamping diode combination, with or without a blocking diode, or use only diode switching with clamping diodes. With the current availability of super high-speed high voltage diodes using only diodes to handle the switch from transmit to receive mode makes a lot of sense. Thus, my DIY AGD25.1 project will use this method But I will also provide a combo diagram which shows how to potentially add a blocking diode to an existing detector.

Diode transmit receive switches have been used for decades but mostly for RF use. But in a PI metal detector the switch becomes more of a high voltage blocking device. Unfortunately to turn on such a switch to on during the receive cycle a voltage must be supplied to make the diode conductive. This turn on voltage creates a big DC offset problem and asking a capacitor to block it at this point in a circuit is far from the best solution.

Back on July 3, 2017 David Emery filed for a Patent for his “Dual Polarity High Voltage Blocking Circuit for Pulse Induction Metal Detector" and was granted a Patent for it on Jan 15, 2019, Patent US 10.181.720 B1.

What makes the diode TX/RX switch attractive now for use in metal detectors has been the recent developments of super fast 1amp rated diodes with voltage ratings of over 600 volts and recovery time of less than 30ns and simplicity of the basic circuit and that it is automatic. My own adaptation of this transmit and receive switch circuit can achieve dc offsets below 10uv with only minor trimming and 1.11mV with no trimming at all according to all my Qspice simulations.

Let’s begin with PART 1: The Signal source simulation and processing

This part consists of:
1. The test signal source to be used for all development work.
2. The comparator used to determine a specific point in the signal source decay waveform.
3. The TX/RX switch circuit using two methods.
4. And my circuit to eliminate the TX/RX switch DC offset.

The test signal source that I use in the simulations should be fairly close to a CoilTek Elite nine-inch flat wound mono coil. Most values were measured after disassembly of one of my coils, some are estimated and the carbon covered paper shield was ignored. The schematics may have some comments that I felt were important to remember while running the simulations. The plan is to progress to the demodulated output signals.

Here is a picture of the outputs of the part 1 signal source using a 100% diode based transmit - receive switch. It will be the circuit that I plan to use. I will post schematic shortly.




Here is a picture of the outputs of the part 1 signal source using a blocking diode and a resistive network and common clamping diodes along with my DC offset control circuit. The series resistor used is 2.5K and would be a none inductive wire wound resistor to minimize resistor noise.




Here is an additional picture of the circuit with a 100 % diode transmit - receive switch that I plan to use but at the micro-volt level to show the DC offset of the output of the switch. It contains my sine wave test signal as shown.





The below picture shows the currents through the diodes during the high voltage part of the decay curve for the signal used.




The below picture shows the currents through the diodes after 32us for the same.



I will post diagrams of both circuit options in my next post.

[Auto-Mation-Assist]

The NET names on the prior pictures refer back to the NET names on the diagrams below.


The first schematic below is for my proposed diode transmit - receive switch, decay complete detector and the switches automatic DC offset control X9.

In the Patent mentioned before there is a resistor to ground so that switch has a very large DC offset on its output. This DC offset will not allow clamping diodes to be used at that point because the large DC offset will drive them into conduction. I will use my DC offset control, X9 instead to eliminate that problem at the switch itself.

All resistors should be of high quality and if a circuit board is developed it can be made to build either the circuit or the second. Only two parts to change out


See notes on the diagrams.





The second schematic below shows a how a older detector with a resistive type clamping circuit could be updated to add a high voltage blocking diode.

[Auto-Mation-Assist]

In the diagrams posted before, comments are in blue and spice commands are in black.

Here is a picture showing the diode switch signal loss. Converting percent to dB indicates it is about 1.6dB.

I have not been able to upload the actual Qspice simulation files as a .RAR file. It loads but when I try to access after saving the post it comes back with a invalid link error. So I think need to upload the files somewhere else. Will work on that tomorrow.

[eduardo1979]

Very good work as always.
I am surprised with the sensitivity of your detector. Detecting a 0.015g nugget from 2 inches away, that's really impressive. I have never been able to make such a sensitive PI detector. My PI detector detects a 0.15g nugget 1 inch away.

[Auto-Mation-Assist]

Since I needed a place to upload the actual Qspice file that I used for development work to allow sharing I decided just to add this project to my github account. I uploaded all the files and pictures that I posted here and the actual Qspice files for Part-1 of this project to:

https://github.com/auto-mation-assis...velopment-Work

I verified that the link below will take you directly to the actual Part-1 Qspice files

https://github.com/auto-mation-assis...Qspice%20Files

The next addition to Part-1 will be Part-2 which adds the first gain stage and its required surrounding circuitry.

The first gain stage will use the LM7171A low noise high speed op-amp. I don't care much for the output stage of this op-amp but it can supply up to 100ma of drive but is not rail to rail and can have substantial DC offset at its output. Its DC offset will be brought as close as possible to zero using a time slot 100us sample from the end for the receive signal and 2us before it ends. To insure that the LM7171A's output does not become sloped from start to finish there will be no capacitors used to attempt to reduce its gain at lower frequencies.

The first gain stage will be followed by a AD1212 analog switch which will divide the receive signal into 3 time slots. The 100us wide one for front end offset control and a time slot for the leading fast channel followed by a wider time slot for the slow channel. In my prior AGD23 and AGD24 designs I used the fast channel signal to control the front end offset to center the receive signal to maximize maximum dynamic range of later gain stages but that also effected the slow channels DC offset and was compensated for later on. It was workable but possibly not the best solution.

With my proposed changes DC offsets will be the totally independent for both the fast and slow signal channels. Thus the 100us added wide channel is the master to set all DC zero volts reference points in all gain stages and should also adjust itself for various ground conditions during actual use.

[Auto-Mation-Assist]

Very good work as always.
I am surprised with the sensitivity of your detector. Detecting a 0.015g nugget from 2 inches away, that's really impressive. I have never been able to make such a sensitive PI detector. My PI detector detects a 0.15g nugget 1 inch away.

Good sensitivity can be hard to obtain because it is easy to make timing and compare reference levels not stable due to noise either from external sources or circuit board layout. My own best as a version where most all of the transmit circuitry was on its own circuit board which likely reduced ground plane induced noise. That version used what I called at the time by the waveform corrector which allowed fine tuning of coils decay damping to remove most of the voltage spike that were caused by the clamping diodes to stop conducting. This was adjustable via a trim pot and used an active feedback loop to the coil before the first gain stage. I described the circuit used for this in my AGD24 version posts. Removing this spike greatly reduced the signal level that had to be handled and thus overload of later gain stages.

When trying to detect real small gold my fast channel usually had a gate width of about 1.8us and small gold will only show up in the very first .2us or so at the most. When you average out the ratio of this then the desired signal is only about eleven percent of the generated average voltage to compare to. At this point if is just a very small change in what is basically a DC voltage in the demod section. The slope of the coils decay curve is also very fast and steep at this point and the demod processing has to be fairly fast to be able to detect the very small voltage change. If the detector has to much internally generated noise than it may be impossible to detect such small signal level changes. I know that when I use a TX pulse width of 36us that I need to start sampling no later than about 3.7us later. I think I documented that is some of my AGD24 version post via the timing charts.

[eduardo1979]

Thank you very much for the clarification.

[Auto-Mation-Assist]

I had been running a lot of simulations on the various sections of my proposed AGD25.1 analog PI detector and got some decent results especially when it can to choosing some of the OpAmps. As a result of the simulations I started initial work on laying out the circuit board that contains most all of the analog sections. A picture of the first prototype version of the analog board is show below.



The transmit circuitry is on the left while the receive is on the right. an going up from bottom to top. At the top and heading left are the DEMOD low pass filters. After the PreAmp there are two identical channels which I label fast and slow. This board has the TX-RX switch circuitry that I described before, and also has the the ability to use a single diode and conjunction with a 2.5K none inductive wire wound resistor. Proper TX coil damping is very important to me so the board has some large through hole type resistors that provide low noise levels for circuitry that is either across the TX coil or in series with it.

The board has a isolated 12v to 15v DC to CD converter to generate a +15VDC form the -14.4Vdc from the battery. The on board low noise low dropout voltage regulators provide plus and minus 12VDC power for the circuitry. The board has eight analog gates that will be controlled by five digital pulses. During the TX coil's decay a pulse is sent to logic to start the pulse sequence required to properly demodulate the receive signal.

It is time to develop a second board that this board will plug into and will contain the digital parts.I have updated STMCubeMX and the STM32CubeID, but It has been a while since I looked at my STM32 development hardware so it will be like learning how to use it again.

[Auto-Mation-Assist]

I was not happy with this circuit board layout and discarded it. Over the last several I developed new circuit board layouts after doing most of the programming work for its processor and after getting tired of running simulations. One of the things that I did not like on my original circuit board is the amount of space the coils damping circuitry was taking up. The reason for so many resistors is to have the flexibility to parallel resistors to come very close to the proper value and then trim the small remainder with some one watt trimmer pots. The second reason is to reduce heat generation and thus provide improved thermal stability. To reduce resistor noise levels one inductive wire wound resistors are used which unfortunately do not have a lot of standard values and this again requires a greater number of parallel resistors to come close to the required value. The final required value is not calculated but determined by looking the waveform entering the sample and hold circuitry. Here you can easily see how even a one ohm change damping resistance effects the waveform of the received gated signal. I pay a lot of attention to coil damping because it is very important for optimum performance. In the new circuit board layouts all the coils damping resistors, DD coil switch and the GX16-5 coil connector have been moved to their own back panel circuit board. There is also high voltage TX/RX switching diode installed on this board.

One of my requirements for the AGD25.1 was that it has to fit into my existing enclosures and with improved access to the connector that route from the circuit boards to the rear panel. This requires that the upper transmit board be shorter than the lower receive board. Most of the coding for the processor done and what is left to do will be done once the first AGD25.1 is assembled its circuitry tested towards the end of January 2025. Right now I'm busy with the final verification of the four circuit board layouts and insuring that all the required parts will be on hand.

One of my checks and to eliminate any confusion caused me to make a block diagram of how all the circuit boards will interconnect with each other. The two main circuit boards will again plug in back to back while the front and back panel boards will have wires soldered to their connection points and have plugs on their far end that plug into the TX or RX board. Here is a picture of how these will interconnect.



Since I started this post mostly with the coil damping I drew in the parts that I think will end up populating that board. It gets soldered to the coil connector after the coil connector is installed and help place by that connector and the DD coil switch. The coil connector will need some solid wire soldered to its pins for connecting to the circuit board and perhaps have the pins in the back file'd down a bit before the solid wires are soldered to the pins. The two fine adjustment trimmers are actually round and I will likely make them reachable through screw holes in the rear metal panel so adjustment can be made while the detector is fully assembled.

A PDF file of the picture is also attached.

[Carl-NC]

Temperature changes can cause a greater damping shift than 1 ohm of damping resistor change. For lowest flicker noise, use metal film instead of wirewound. Be careful using pots in the damping resistor, prolonged exposure to high flybacks can burn the sliding contact.

[Detectorist#1]

Congratulations AMA! Your record of 0.015g from 2" must to be written is "Guinness World Records" book! This allows finding of gold with price of 1USD! But what is the real benefit for members of MD community from your efforts? Maybe it will be more useful for us if your very good skills in electronics are pointed in other hard to avoid problems in PI MD - discrimination and ground balance. Thank you in advance!

[ivconic]

As before; such efforts and work are worthy of all respect. We haven't had a member like this on the forum for a long time, who explains his ideas and presents his work so meticulously and in detail.
However, unfortunately, as in previous cases; something "tangible" in the form of shorter video presentations is missing here.
Don't get me wrong; I'm the first to criticize nonsensical video demonstrations that prove nothing.
But this is a special and isolated case when it is clear from the beginning that this is a man who does his job perfectly correctly and with a lot of knowledge.
The main reason for my insistence on video is because it uses very expensive components, which unfortunately I can't get easily.
It's not just about money, but also about the literal impossibility of easily getting some opamps and some very precise components.
A mere inspection of the schematic does not reveal many new details, mostly it is known technology and we have had most of these solutions on the forum before, partially or in whole.
It is easy and simple for someone to engage in self-construction of this, and for someone it is almost impossible, for the reasons mentioned.
That's why I still think that one or two well-made videos showing how pre-made detectors work would do; raised interest in this topic by 500% and more!
And certainly there would be many more forum members who would be happy to join in the whole story.
You have to understand that detectorists want as many documents and details as possible in order to decide to engage in these kinds of stories.
Let's be real, this project isn't easy and it's not for beginners. This project requires a serious hobbyist with a lot of skills and experience.
In order for this topic to live and develop; I am still of the opinion that it is necessary to record and give at least a couple of good videos.
My 2 cents...

[Auto-Mation-Assist]

As before; such efforts and work are worthy of all respect. We haven't had a member like this on the forum for a long time, who explains his ideas and presents his work so meticulously and in detail.
However, unfortunately, as in previous cases; something "tangible" in the form of shorter video presentations is missing here.
Don't get me wrong; I'm the first to criticize nonsensical video demonstrations that prove nothing.
But this is a special and isolated case when it is clear from the beginning that this is a man who does his job perfectly correctly and with a lot of knowledge.
The main reason for my insistence on video is because it uses very expensive components, which unfortunately I can't get easily.
It's not just about money, but also about the literal impossibility of easily getting some opamps and some very precise components.
A mere inspection of the schematic does not reveal many new details, mostly it is known technology and we have had most of these solutions on the forum before, partially or in whole.
It is easy and simple for someone to engage in self-construction of this, and for someone it is almost impossible, for the reasons mentioned.
That's why I still think that one or two well-made videos showing how pre-made detectors work would do; raised interest in this topic by 500% and more!
And certainly there would be many more forum members who would be happy to join in the whole story.
You have to understand that detectorists want as many documents and details as possible in order to decide to engage in these kinds of stories.
Let's be real, this project isn't easy and it's not for beginners. This project requires a serious hobbyist with a lot of skills and experience.
In order for this topic to live and develop; I am still of the opinion that it is necessary to record and give at least a couple of good videos.
My 2 cents...

ivconic,

This project is for individual who are interested in working mostly with analog circuitry, which by nature is more complex and expensive and likely only of interest by a very small percentage of individuals.

If you depend on videos to make decisions you are different then me. I base my decision on circuit design and analysis, and perhaps a few words of what a developer is trying to accomplish is more than sufficient.

I have no problems getting the parts from Mouser or DigiKey, and when you want the use high quality parts the expense can be substantial and it should be clear that the development work for my detector has been done at considerable costs and time.

I think that if you build one of these you may even find a way to make improvements that I have overlooked or considered. Then as you test and verify those improvements you can post your information about those and potential videos of those here and thus help others that may wish to undertake a project like this.

Just build your own analog version that you have faith in and post your progress..

[Auto-Mation-Assist]

I have a update on the two main circuit boards for version AGD25.1. I decided to go ahead and have the Transmit and Receive blank circuit boards manufactured. I checked the circuitry as carefully as I could need to get at least one detector built up of testing and making bench type measurements on its performance and work out and bugs found. I think these circuit boards should arrive here before 15 January 2025. The TX board is smaller than the RX board and these back to back board combined save about one inch of depth in the enclosure. This gives a lot of space for the added rear panel board and its ferrite inductor. I have not used PcbWay for my circuit boards but will order the new front and rear panel circuit boards from them some time this week.

Here is a the layout of the initial AGD25.1 transmit circuit board.




The below is the initial AGD25.1 receive board. It actually has the same width as the above TX board.



Now its just time to verify my parts stock and order whatever is not on hand. Most all surface mount resistors that were still thick film type will be thin film type. The PreAmp stage on the bottom left as space for three metal film through hole 1/4 watt resistors. There is a horizontal test point pin strip for all gating pulses. There are a number of OpAmps that I have not used before.

I expect some bugs but think they will be minimal.

[Auto-Mation-Assist]

Temperature changes can cause a greater damping shift than 1 ohm of damping resistor change. For lowest flicker noise, use metal film instead of wirewound. Be careful using pots in the damping resistor, prolonged exposure to high flybacks can burn the sliding contact.

I agree that trimmer pots in high voltage circuits can have problems as you describe, but In the AGD25.1 detector they only have a adjustment range of about 27 ohms of the approximately 580 ohms required for damping. The simulation circuit is below my CoilTek 9 inch mono elite coils.



The voltage across the trimmer has been minimized. The four 2.5K resistors are all wire wound none inductive with a ppm/DegreeC of +/-20 and inductance of less than 1nH at 1Mhz. They could be replaced easily with Metal Film ones, and I have both types on hand. However none inductive wire wound are know to have the least amount of noise. I have used them in many tubed audio circuits for years and see no reason not to use them in this application. And in thinking in about the RF world, they provide a good out of band termination for the coil well past what is required frequency wise for a metal detector. Termination being a RF term for a proper load, damping in detector speak. Out of band meaning at frequencies outside its normal operating range.

If I were not to use the 5K trimmer that I specify I would replace it with my active circuit that can do the same thing but uses feedback and is also easy to adjust. I described and posted a diagram of how to do that in the past. Such a circuit can also use a simulated inductor using a OpAmp as a feedback element to accomplish the same thing. I think I also used a simulated capacitor as feedback in some testing.

For the AGD25.1 I decided to use trimmer resistors for the making coil damping fine adjustments. This is the adjustment that effects the ability to detect very fine gold by preventing over load (flat topping) at the final gain stage caused by peaking of the decay voltage. I have shown this type of peaking in a number of prior posted pictures.

I decided to use: DigiKey Part Number 987-1177-ND for the two trimmers. One 'selected' for mono coil and one for 'selected' DD coil.