My AGD analog detector work

[Auto-Mation-Assist]

I never seem to get the name of the AGD26.1 signal processor circuit board correct. But here is a picture of the circuit board I will use for testing and continuation of coding work for the 128 pin STM32 MPU. I made this board the same size as the front end board for now and may move the lower mounting holes upward and trim it down to smaller size. Close to 4.1 inch wide and 4.75 in height.



The programming JTAG connector towards the upper left will be accessible through a access hole in the side of the enclosure. Once it this board is assembled and programmed the front end board gain will be adjusted so that everything plays together nicely.

Below is my current circuit diagram for this circuit board.

[ATTACH]n449377[/ATTACH]

[Auto-Mation-Assist]

I really like using individual controls for controlling my detectors and this version of my AGD26 detector is intended to fit inside my existing enclosures. They provide good RF shielding and weight just about the right amount to balance a completed detector with coil attached. The front panel control panel is getting a new layout since there are more switches now that will interface with the MPU. I designed a new front panel circuit board for the 26.1 version as shown below.


All the switches and controls mount to the back side of this PCB, it them mounts as an assembly to the 1/8 inch aluminum front panel. It interconnects to the signal processor board via wires that feed from the back side of the circuit broad and exit the board through the holes marked FP1, FP2 and FP3. Solder pads are provided to solder the wires into, FP1, a dual row 12 pin, FP2 and 6 pin vertical, and a 3 pin vertical that still as a small old FP1 label but is FP3 just for switching between internal speaker and headphone. This is simple circuit board and receives 3.3V dc from the signal processor board for the digital switch interfaces, and a 2.048 ADC reference voltage for handling the analog seven 2.5K linear controls. Their output voltages is routed to one of the STM32G474 MPU 12 bit ADC's, over sampled, averaged and the scaled as required. scaled.

The controls are intended to provide some adjustment range to, TX pulse width, channel 1 gate width, and channel gate width. For channel 2 gating there is a second control that allows for moving is gate period left and right. This is called CH2-Slide. The AGD26.1 always uses the the TX coil for its channel 1 receive signal. If a coil such as a DD coil is connected it can be operated like a MONO coil. Channel 2 can be switched to use the separate build in receive pre-amp to utilize the coils separate winding. Thus while in DD mode it can use both coils during the receive period. The DD-PRE switch handles that function.

The main signal processing is done by two 16bit ADC's, One for each channel. I'm sitll using the sample and hold circuit as in my prior detectors since the output of these also control the front ends board DC offsets. The sample and hold circuit adds a second receive gate which has a fast raw signal output and a filtered slower output, the sample and hold value for the gated period. I have added a switch to allow selection of which signal gets processed the the two 16 bit ADC's. Switching to that function is by front panel switch, SIG-SRC.

The MPU witll be programmed to have a default configuration which will be able to be modified via user control by setting the MAN-DISC switch on the front panel. Doing so makes the RATIO control active to allow for variable mixing of the two channel. Inverting channel 2 will be possible by using the CH2-INVERT switch on the front panel.

Most of these things are to get a bit more flexibility, time will tell if they will be beneficial or not. My diagram for the front panel does not show parts values but are 0.1uf caps, 1k resistors, 2.5K linear pots and bat handled toggle switches.



Here is the file in PDF format.

AGD26.1-Front-Panel.pdf

[ivconic]

Congratulations!
I have never seen a man who is so systematic, orderly and detailed.
But here I am looking... the date you started the topic is 03-03-2023.. and today is 06-12-2026.
Over 3 years on, we still haven't seen that detector in action... I wonder; do you also buy components on Aliexpress!??

[Auto-Mation-Assist]

ivconic: I have purchased from Aliexpress in the past but don't recall what I purchased. My usual parts source is Digikey, Mouser being second. For all other things that they do not carry it is internet search until I find something suitable.

Back to the AGD26.1 work

Back in 2023 I got interested in building a PI based metal detector without knowing anything about them. Since then I have built a lot of them and used each one to make improvements a bit at a time. I still use my AGD24.1A version that has damping resistor changes, and first receive stage and channel amps changes. The development work did not end there it has continued with focus on pulse generation, coil damping and very first receive gain stage. Every thing else depends on how well a detector will work. The nice thing is that the pulsar, damping and first gain stage can be simulated very accurately. Simulations can give great results, but poor circuit board layout and design can completely wreck those results. All this takes time, and even with great care problems can be expected to occur that require a bit of rework.

Time to some of the mechanical work that needs to be completed:

I got my drawings completed for modifications for two of my older enclosures,and their front and rear panels. I also drew out the shield that will go over the part of the TX section or the front end board. This will also be used as a heat sink to help control the heat generated by the three 2.5K 3W resistors.

Some times when thinking about the retirements for a PI metal detector it may be beneficial to think of it as being two delay lines which sync to a common trigger point, the end of the TX pulse. This starts the decay of not only the TX coil but also the targets signal. The delay line concept is what decayed those two two out until they reach the zero signal point. Drawing a chart helps to clarify a bit between TX coil decay and targets signal decay slope.

In the small chart below I set the known requirement to detect 0.015 gram of gold in my AGD24.2 detectors to have a decay time of about 3.9us after the TX pulse ends. It also shows the approximate 2.5us decay point of the AGD26.1. The targets return signal decay slope is still much higher at the 2.5us point than the 3.9us point. Just a bit further that 4us the targets return voltage becomes zero and no longer delectable. One additional thing that needs to be considered if speed. If the gain stages are two slow the signal to be processed may not be able to get properly time aligned.

There are various ways at looking at the signal relationships in a PI detector, I like to use something like the below chart:

[ivconic]

".. Simulations can give great results, but poor circuit board layout and design can completely wreck those results..."
Ok, clearly, I see that you have set yourself high goals and ambitions, sampling below 5uS is already something that exceeds ordinary DIY possibilities.
Although you wrote a lot of text and gave a lot of schematic parts and the appearance of your pcbs is outstanding...
still I didn't have the patience to read everything so I apologize for asking something that you may have already answered in your posts;
Besides the ambitions of sampling below 5uS, does your detector have GEB and DISC?
All the effort you put in would be pretty pointless without even a GEB option.
A video demonstration does not have to prove anything, but it can still tell a lot to an experienced eye. And that is what is missing from this thread.
When I initially asked you to make a video; I didn't mean to watch something pointless, but to get a basic sense of how that detector works.
If you followed closely other topics on the forum; you would have seen my sporadic resistance to most of the videos on Youtube,
which prove nothing and have the ambition to prove something.
But especially in this case, a good video is what we need here. Not to criticize and mock - but to get a basic sense of how it works.
Your high commitment to your project, as well as the many articles you have written so far, are worthy of attention.
But you have to understand that we live in a modern age when the story is incomplete if we don't have a video experience.

[Auto-Mation-Assist]

GEB is basically and change in the detectors zero reference level. To speed up this adjustment individuals generally move the detectors up and down to regenerate a larger difference voltage which aids in resetting the zero reference level a bit faster. Generally this is done automatically through design, it usually just takes a bit longer. DISC is a bit more complicated, and can use early or late signal mixing. Detectors that have built in digital processors can use more complex methods if they have built in math processing abilities. My AGD analog detectors have always included my version of GEB as shown in the circuit diagrams. It is done automatically.

The AGD detectors are meant to be used to find very small gold and have undergone changes that improved performance a bit at a time. In considering those changes I would have to say that more attention has been paid to TX coil decay time, which also requires greater dynamic range in the receive gain stages. Most all my postings have been related to the technical aspects. The signal processing board with its MPU will further improve the detectors performance. The combination of the high performance front end and the future digital processing addition will offer great flexibility via firmware code changes.

My post are of a technical nature and not a story. The technical content may be beneficial or not, but it is is open to comment and it is hoped that others experiment with the concepts and circuitry that I use or describe, and perhaps offer suggestions on how I could improve those further. I do not have a lot of spare time to monitor the forum so it is best to contact me via message.

According to UPS the two main circuit boards for the AGD26.1 should arrive by 9PM tomorrow. While waiting for those to arrive I updated my parts stock and worked on the changes required to my enclosure. Here are some pictures of that work.



Back side of front and rear panels showing anti rotation pockets for the switches and pots.



I was just going to put a shield plate over the TX section but changed that to have a bit of heat sink ability. To use this the three three watt coil delay load resistors would need to stand off the board 0.125 inch and the a 3mm thigh silicon pad placed between the resistors and heat sink. In the picture below I show it attached to a early engineering development circuit board. The slots in the heat sink will be at a slight upward angle when the detector to be is sweeping the ground. If I make more of these heat sinks I would rotate the slots 90 degrees.

[Auto-Mation-Assist]

The AGD26.1 Front end and Signal processor board have been assembled. I have not tested the Signal processor board yet but did find a issue with the front end board method that I used to control current sharing of the two parallel THS3092DDA current feedback op amps , which then resulted in making changes for that function in the main first gain stage for the TX coil damping pre-amp and the DD coil RX pre-amp.

The method I used for current sharing compared B against A and then making then equal. The problem here was that equal could be 300 or more Milli-amps of current flowing across the two 5.1 ohm resistors tying the op-amps outputs together. This still allows for the correct amount for gain but will generate a lot of op-amp heating, and may cause the negative 10 volt precision voltage regular to turn on its current limiter and reduce its output voltage. The negative voltage regulator is the last one to turn on in the power supply turn on sequence and thus can be used to help detect some circuit problems.

This issue with not having proper current sharing was caused by lacking a correct zero volt reference point for the integrator. It much better to do this with two separate integrators that use common ground as a reference. My correction of this problem is shown in the two below pictures and have been tested to function as desired when balance current being down in to the micro ma range. I expect that the value of the 10K resistors may be increased later and that the resistors marked with NI (Not Installed) may be added to control hunting if that shows up during final project testing once the firm ware source code for the project is complete.

I have the integrator as the dual channel AD8034ARZ since its package is easier to deal with, but the dual channel OPA2828IDGNR can by used instead of the original single channel OPA828IDGNR. Project parts numbering is may be different since the circuit board layout was changed and renumbering is not complete.


The Damping-RX pre-amp current balance change U21 circuitry.



The DD receive pre-amp change, U17 circuitry.




I'm going to take a break from this project for several months for some summer outdoor activities.

[Auto-Mation-Assist]

I have been testing the AGD26.1 two main circuit boards up to the inputs of the ADC converters after programming the STMG474 with enough code to get the analog gates opening and closing. I did find some self induced problems that I had to over come winch make my engineering circuit boards a bit messy but functional. The front end board has a temporary version of the modification I described in my prior post since the foot prints of the placement part could not use the foot print on the circuit board. The signal processor board schematic capture grabbed a old version of a 5 volt regulator instead of the new version. Don't know if that was caused by my change from Windows to Linux Mint on all my computers.

I programmed the the processor board for a TX pulse of 36us and a repeat period of 160us. The first analog gate pulse with was set to start at 2.5us after the TX pulse end and with a width of 1us. The second analog gate was also 1us but delayed from this first gate by 110 nano seconds. After adjusting the coil damping trim pot the circuit board pair was able to detect and show good signal level change at the signal that will be routed to the AD converters. I used a 0.155 flat alum foil for this test (3.937mm square) at a bit over four inches. This is not not good enough to detect my normal gold test sample but it is a good start for the analog section up to the test point that I used to view the waveform.

I would love to show a nice clean boards but all I have right now are my engineering ones. I plan to work on the firm ware when my time allows which when completed will make evaluation of performance a lot easier.

Here are some pictures of the work in progress.



The above shows how the two circuit boards fit together.



The signal processor board above had a old 5V regulator footprint that caused a short. A temporary ground wire is around the coil connector.


The below front end board is messy but functional and allows for overall testing an performance evaluation.



I will post the front panel control board in my next post.

[Auto-Mation-Assist]

I assembled the front panel control board. It will allow for further testing as it is needed to interface with the STM32G474 MPU.



The wires will run into three groups and terminate in plugs that plug into sockets on the processor board.

The below shows this assembly mounted to the back of the front control panel which has room for the power connector and power switch.



The last picture show the front of the front panel without engraving and knobs. The shafts will be cut so they are shorter.



The pots that affect timing will normally be disabled unless a change in setting is desired. The switch on the bottom right is momentary and must be toggled down for timing changes to take affect. If it is desireable to listen to tones while adjusting the switch can be held down. Hold it down for X number of seconds and timing will reset to default values. That is the plan.