My AGD analog detector work

[Auto-Mation-Assist]

After adding the last two circuit additions to my detector the next thing I wanted to do is to reduce the affect of temperature related receive waveform changes because these change the point at which the decay curve is sampled and likely present in most all PI detectors.

Since components used in electronic circuits have temperature related drift in their values which will affect all signals being processed. I think that in most PI detectors the temperature related drift has to do with parts at are just before the first Rx gain stage, namely the clamping diodes but any series resistors type and value must also be carefully chosen. Any temperature related changes here are not only amplified by the receive gain stages but will also end up looking like timing changes.


Here is a schematic of my receive board input.



I have measured the diodes that I use to clamp the input signal from 74 to 140 degrees Fahrenheit and find that the voltage drop across a single diode changes 44.2mV over that temperature range with only 1.28mA of current applied. The current that they actually need to handle is about 90 times that amount but varies with peak coil decay voltage which can cause significant warming of the diodes over time. Nearby parts that also warm up should not be placed to close to the clamping diodes.

The actual voltage across my test diode which was mounted on separate circuit board was 0.3984 V at 74 degrees F and 0.3542 V at 140 degrees F with only 1.28mA flowing through the diode. This temperature related voltage shift will end effectively changing the point at which the receive waveform is sampled.


Any PI detector that uses clamping diodes will suffer to some degree but my test have indicated that the diodes may contribute more that 85 percent of all temperature related drift and that it becomes more of a problem when a detector has timing parameters set to detect very small flakes of gold. In such a case parameters set on the workbench for a detector may not anywhere near correct for locations where temperatures are significantly higher. Transmit pulse length has control of the heating process and a detector using multiple pulses instead of one may cause more heating of clamping diodes.

The best way to at least help alleviate this is to have a temperature sensor in the detector and have its output adjust the timing parameters to compensate for this drift automatically, and the second best and always desired is to keep the clamping diodes from warming up to much to start with by choosing the best possible suitable location for these components. I did my testing using the location shown in the picture below below with heat sink compound for my actual in detector test.

My initial test clamping diode heat sink - The loose wire goes to the flatten circuit which I removed for the picture.



The AGD detector currently uses 2ea MMBD452LT1G dual diodes in parallel for the diode clamp and I used a type K thermocouple to make my all my temperature measurements. The range I was most interested in right now was for 74 to 85 degrees F which has a 7.5mV change 0.3984V cold and 0.3909V warm. Keeping the diodes cooler shows a very significant improvement in overall temperature related drift and this can be seen clearly by looking at signal at the output of the analog gate feeding the fast sample and hold circuit. Particularly at the leading and trailing edges of the signal was temperature increases after turn on.

This is a typical wave form picture when the diodes are cool. When the temperature of the diodes increase the spike at the top left side will drop down below the horizontal line by about the same amount as they are above the top with the heat sink. Without the heat sink the spike will drop down to about 1 horizontal line up from the bottom and also lower the overall signal level. That shows that we are sliding down the decay curve by perhaps 0.2us or so. That is a lot for the small temperature change during bench tests. The part of the wave going up towards the right is the action of the automatic offset control in the detector and ends straight down and then the second slower channel starts to sample.



I plan to keep the clamping diodes temperature as stable as possible by putting them in a U shape clamp with some 1mm thermal silicon pad material surrounding them on top and bottom. This will then press into the slot in the clamp and have a nice tight fit, and still be removable if required. The standoff in the picture will likely become part of the clamp. I will also plan to do some testing by adding a temperature sensor onto the clamp and then into the timing circuit itself. That would allow for a greater temperature compensation range.

My current test procedure is with my swinging small flake of gold attached to a golf ball shown in prior pictures at a 1 inch distance which still gives a very strong audible signal, and use machinist gauge blocks for ease of setting my distances.

[Auto-Mation-Assist]

Most all of my test are done with the CoilTek 9" Round Camo Mono Ellite coils and I think that these are really good coils but I feel that they have a problem with flexing of the flat surface of the top of the coils.

I have noticed that it takes very little pressure on the top of the coil at an area about 2 and a half inches in front of the cable entry point to give me a signal that is many many times stronger than my normal test sample gives on the golf bar swinging above the coil. Last summer I also noticed some modulation of weak signals that I now suspect many have been top cover resonance vibration after bumping into something. There could be other potential reasons for this but top cover flexing is something you can actually see and feel.

I think that the internal shield may be sprayed on the inside of the top cover and thus any flexing of the top cover will change the spacing to the coil windings since there is a gap between the two. I have two of the 9 inch coils and also a 14 inch which flexes quite a bit more. Also when I run my fingers across the flat part of the top cover I get a sense that it very thin and that it is floating all by itself.

I discovered this flexing issue while tracking down some hefty noise spikes that would occur at random but have not yet the source of those. It may be be possible to drill a hole in the top and run some kind of filler into the coil to fill the void inside but that would make potential disassembly nearly impossible. I think that these mono coils are really good but I'm not happy with they way the top covers where chosen to be built.

Has anyone very had one of these Elite mono coils apart and looked at the internals or have any suggestions?

[Auto-Mation-Assist]

The updated TX circuit boards arrived for the detectors that I built last year. These boards added the modification that I posted before which controls the coils damping resistive load during the time when the decay waveform gets low enough to make the clamping diodes stop conducting. Since these boards need to go into existing detectors there was no good way to increase the board dimensions to make all newly added components fit on top of the board and thus the larger metal film resistors that that circuit uses were placed on the bottom of the board.



And the bottom




With the added function the sampled waveform can now be corrected for the loss of partial coil loading when the clamping diodes become a gradual open circuit. In order to be able to display this and make the adjustment I added several test points to the sample and hold circuits on the RX board. This allows for looking at the receive signal only during the correct sample time. To get the proper wave form at the correct time both the point at which sampling starts has to be set and also the delay for the sample and hold gate has to be set correctly. This may be about 0.205us and is set by adjustment on the RX board. This compensates for signal delay caused additional signal routing before the signal reaches the sample and hold circuit. Adjustments values depend to TX pulse width and voltage levels driving the TX coil along with coil damping load. These control peak coil decay voltage and the length of time it takes to decay. The decay complete signal is master and other adjustments are made after this best time to trigger is set and is before well before the decay reaches its zero volts point. For me it is typically 2.805us after the TX pulse ends which activates the first RX gate 0.445us later or at the 3.25us point.

To make the two adjustments, for 'decay complete' and 'waveform correction' the TX board has two adjustment trimmers. The first is a time delay based control and the second is a voltage based control.

Back in a prior post I made note that clamping diodes are a bit temperature sensitive and It helps to keep them as cool as possible. I looking at my old detectors more carefully I decided that it was best to move them to under the board and clamp them between the board and its mounting tray with 3mm silicon pad material top and bottom. See picture below which show part of the receive board bottom.



As a alignment aid I added two test points for looking at the outputs of both sample and hold gates on the receive board.



In preparation of adding the circuitry for the flatten circuit board I added wires to the point were its output signal will go. These are the yellow and blue wires. The circuit board which will cover most of the receive circuits and also act as a shield so it will replace the existing metal shield. Power and TX pulse for the flatten board will come from the TX board.



I expect the flatten circuit boards to arrive in about a week. I have tested by just using the features on the TX board with a spacing of 1.3 inches with my normal golf ball test set up and it detects the small 0.015 gram gold flake just fine. The only circuit change was reducing the value of the resistor going to pin 3 of the AD8033AKSZ in the wave form correction circuit from 7.32K to 1K to reduce the chance of getting a bit of ringing while still having some common mode level protection. It is possible to lower this to zero ohms.

[pito]

diodes can account for more than 85 percent of all temperature-related drift; keep them as cool as possible. =
1. Some crystal oscillators are equipped with a thermostatic heating element that maintains a constant temperature.
2. Put the diodes into the thermos, it can be a piece of styrofoam ( coffee in a styrofoam cup cools slowly )

clamp diode heatsink = does not prevent temperature changes

[pito]

Post # 82
will get damaged by the 700 volt recoil voltage of the transmit coil = maybe connect 2 MOSFET in series?

Has anyone very had one of these Elite mono coils apart and looked at the internals or have any suggestions?

https://www.youtube.com/watch?v=UaHe...kxgNVAh3gCoAEC

[Auto-Mation-Assist]

Pito, I read your post while I was working on my own post and the things I'm working on are only refinements to optimize performance and stability. These refinements affect not only my own detectors but also other persons detectors since we all use more or less the same components. I share my work on this for others to ponder over.

When high currents flow through a tiny little diode it will heat up significantly and putting it a insulated enclosure will make things a lot worse.

All of my frequency counters use ovens for their frequency standards.

Also while I was designing and building high end audio class a Power amplifiers some 30 years ago I temperature stabilized those at 126 degrees F. So temperature related problems are not new to me.

MOSFETS ?? There is absolutely no need to consider changing the coil driver MOSFET unless TX pulse width is raised above about 45us. If that is done it will be desirable to handle the higher peak recoil voltage since it will be higher than 650 volts that the current MOSFETS are rated at, and thus 1000 volt MOSFETS will be needed so that the recoil voltage will not be clipped by the body diode in the MOSFET, which will cause it to heat up.

The only thing that will get damaged if the clamping diodes fail to a open condition is the immediate destruction of the first gain stage opamps.

[Auto-Mation-Assist]

About five days ago I ran some tests while heating up the clamping diodes that are show in a prior post and the chart I made based on my measurements.

This test was with the clamping diodes on top of the RX board and with clamps to the standoff with .025 thick alum top and bottom of the diodes and with the diodes surrounded with 1mm thick silicon pad material.
The diodes were heated to the temperature indicated or cooled. The heating effect is normally caused by the large peak currents from reducing the TX coil recoil voltage down to around 1.3 volts.
Recoil peak voltage can reach 650 volts in my design and will reach that level if TX pulse is increased to more than 45us.
Current flowing to these small diodes will cause them to warm up and when diodes warm up all their operating parameters become degraded. Thus when dealing with diodes in critical areas it pays to keep them as cool as reasonably possible by choosing the best possible locations for them and using small heat sink if desired.




Nano seconds with + occur later than the reference time of 25 degrees C.
Nano seconds with - occur before the reference time.
DTD = Decay Time Delay trimmer on the TX board and R25 is the resistor in parallel with it.


Degree C---Nano seconds --- Resistance of DTD and R25 combined correct timing (refer to TX board diagram).

10 ..............+082.258 ......3946
14 ..............+072.984...... 3926
18...............+072.581...... 3920
20...............+050.605...... 3895
22 ..............+038.105...... 3877
25...REF.....+000.000.......3822
30.............. -019.355........3790
35.............. -032.056........3770
40 ..............-050.605........3743
45 ..............-072.782........3720
50...............-094.960........3680
55...............-128.226........3631
60...............-200.101........2527

Total timing change from 10 to 55 degree C was 210.484ns with the small heat sink.

Total resistive change required to correct timing over same range was 315 ohms.

Normally the temperature of the clamping diodes reach 30.8 degree C or 87.3 degree F while on the work bench. This is about 15 degrees F higher than ambient temperature.

I think it is not practical to correct for the 60 degree level with analog means. The change from 55 to 60 degrees C is to large.
Without a heat sink or a poor location of the clamping diodes will cause the temperature related drift to be significantly higher.

[pito]

putting it a insulated enclosure will make things a lot worse. = maybe not, it is the chance that temperature will stabilize after 1 or 2 minutes, at least you will isolate them from outside temperature changing
use stronger diodes or metal case diode.

[Auto-Mation-Assist]

putting it a insulated enclosure will make things a lot worse. = maybe not, it is the chance that temperature will stabilize after 1 or 2 minutes, at least you will isolate them from outside temperature changing
use stronger diodes or metal case diode.

Absolutely not !!!!!!!

[Auto-Mation-Assist]

Good morning to everyone.

In working with my analog ADG detector and applying the modifications below have made significant improvements in detect sensitivity while also reducing background noise levels.

1. Added the waveform correction circuit on the TX board as in the schematic AGD23.4.B
2. Moved both clamping diodes below the RX circuit board to better control their temperature increase from the high current pulses of the TX coil kick back voltage.
3. Made a change in the automatic offset circuitry to better stabilize the reference voltage level used to drive the alert tone generator.

Not yet installed is the circuitry to add what I have referred to as the flatten-er which is designed to increase available dynamic range by cancelling out part of the coils kick back voltage that enters the first gain stage. I’m have not yet received the circuit boards for this addition and only wish to do further testing on this as it would installed in the detectors, I assembled about a year ago. I’m working on a new version of the detector that will have all these enhancements built in.

The enhancements listed in the three numbered items have made it possible to raise my swing ball test to two inches as shown in the picture. This is an increase of 1.25 inches over the last few months for detection of the test gold sample of 0.015 grams. I think that this is very good for a PI analog detector with no microprocessor and only CMOS logic controlled by analog user controls.

In order to detect such small items at distance the point in time directly after the TX pulse ends becomes very important. From that time reference the proper number of nano seconds needs to be added to set the sampling start point. The number of anno seconds to be added will vary with TX pulse width since it determines the height of TX coil kick back voltage which in part determines decay time, as does the coils damping resistive load. The shape of the coils decay waveform along with its duration controls everything. Improper adjustment of timing parameters can make it impossible to detect very small items at reasonable distances.

In the picture below shows the coil and ball with 2 inch spacing between the coil and ball holding the 0.015 gram of gold. I used 40us wide TX pulses for my tests which gives peak kick back voltages just below the 650 volt clipping point with my current power supply voltage of 13.78 volts and is close to the medium voltage of a LIFEPO4 lithium 12 volt battery provides during its discharge curve. I can clearly hear the ball swinging in the other room containing my test equipment. I think that it can go a bit more than two inches but need to find a better way to support my coil.



And the gold sample

[Auto-Mation-Assist]

On the top of the list of critical items in a metal detector is that routing of signal must occur at exactly the right times which is complicated be delays that components cause.

Any components placed in the signal path will cause a delay in moving the signal from one point in the circuit to another. Components will can also cause frequency based amplitude and phase changes. It is always important to how much a signal gets delayed from the input of an active stage to its output, and also how long it takes to open or close an analog gate.

Below is a picture of the pulses generated for the AGD detector based on the signal routing delays for the signal up to the inputs of the sample and hold circuits.

Channel 0 - Violet trace


The purple trace shows the negative going edge of the TX pulse which was 40.08us wide.

Channel 1 - Blue trace


Is the decay end signal generated by the TX board. It goes high a few microseconds after the TX pulse starts. It goes high when the TX coil decay is complete enough to allow passing of the receive signal for further processing. The point at which this pulse goes positive is adjusted by trimmer on the TX board and its proper setting is very important. This signal can be considered to be the master clock for the receive side. Nothing happens without this signal. This adjustment basically allows you to slide up or down the coils decay signal to a point that is above the zero crossing point and within the signal handling ability of the analog gain stages at selected normal TX pulse width.

Channel 2 - Green trace

The AGD detectors has its first analog gate after the first receive gain stage. It is there to limit the receive signal to pass more signal than required. Thus it only allows the signal that is withing the desired time period to pass and be routed to the next gain stages. The analog gate that this pulse controls normally closed the gate for 11.16us. I typically use minimum settings but the time it remains closed is user adjustable and depends on combined time of the pulses show for channel 3 and channel 4. Maximum combined being 66.9us. The first signal gate closes on a positive pulse. This pulse occurs 470ns after the rising edge of the blue trace.


Channel 3 - Yellow trace

I like to refer to this as the fast channel and is the one that is critical for finding small gold. This pulse controls the analog gate going to the fast channels sample and hold circuit. When the pulse is low it closes the analog gate and the signal is averaged for the duration of its time. Making this pulse narrow increases sensitivity. Making it to wide will lower sensitivity since a larger portion of the decay slope is averaged. For me the gate is normally closed for 1.96us but adjustable by front panel control to 12.57us

Channel 4 - Orange trace

The slow channel. I normally do not use this channel for anything but it can be mixed with the fast channel by whatever the ratio is desired by front panel control. The polarity can also be changed by front panel control so that the slow channel signal can be added or subtracted. It can be useful under some circumstances if the surroundings created detector nervousness. The slow channel gate closing time is adjustable from 8.8us to 53.92us. The slow sample and hold gate closes an a low pulse.

There is a 300ns gap between the end of the fast channel Yellow pulse and the start of the slow channel Orange pulse.

And I just got word that my flatten circuit boards are being shipped, nice.

[Auto-Mation-Assist]

The picture above is the receive waveform at the input of my fast sample and hold channel and is not what I expect to see there. There should not be a large negative going spike there.

I will show why this is happening, the cause, and my fix that may help other persons working or building their own detectors.

I have made a number of post relating to flattening the downward side of the TX coils decay slope to help with obtaining the most possible receive signal handling headroom. My prior work on this has been tested and can be made to work but it cannot fix the problem with the waveform pictured above. The reason for that is that this problem is not created by the coils decay waveform and is instead created by the clamping diode capacitance, stored energy, op amp input capacitance and also a small amount of circuit board capacitance. This problem shows up as the large negative going spike in the picture. My own simulations has show that this negative going spike has a level of -32mV out of my first gain stage and turns out to be actual reason I developed a way to flatten it. The spike is there is not preventing me from detecting small gold but it is something that should be fixed as a improvement to the existing circuit design.

If you study the partial diagram of part of the front end of the AGD detector you will find that if you just consider everything wired to the output of the three 20W 1W metal film resistors 'to be capacitors' then what we have is actually have a sample and hold circuit that gets set and reset by each signal pulse received from the TX coil decay signal. With each received pulse the parts are charged up until limiting starts and then they decay by whatever the circuits time constant turn out to be. This is not a good thing to have going on at the most sensitive part of a detector.

It is something we have to live with and work around the best we can. This negative spike is generated and amplified by the parallel op-amps to -32mV and it likely ends up being 3.2 volts or more later on, and is a lot for something that is mostly fixable. The spike as a time constant of 357.4 nano seconds from its leading edge to its peak and has a total period of 838 nano seconds between its zero volts points. Since this is fairly fast some of the pulse is filtered out a bit before entering the actual sample a hold circuit by its preceding low pass filter.


To fix this problem the arrangement of the clamping diodes need to be changed from a normal configuration to one in which currents going through the clamping diodes and be measured and also a time difference has to be established so that measurement differences relating to time can be made. This is what I will do using all analog methods to suit the analog detector.


Lets look at some before and after pictures first that also have a 1Mhz sine wave added just to verify that there is no loss of the actual desired signal.

The test waveform with the problem.



The wave form after I applied my fix to decrease the negative going spike to a minimum. There is just a very slight rise of the 1Mhz injected sine wave in the first two cycles. This is a great improvement over the waveform shown above.




My circuit diagram for my clamping diode and associated parts negative going spike fix that is caused diode capacitance, stored energy, op amp capacitance and a very small amount of capacitance in the PCB layout.. The changed parts are in the the rectangular box. It uses three diodes as shown in the diagram, two of which generate the source for the correcting signal.





By using this method I see no need to use the flatten circuit that developed and I tested before since I think that this addresses the source of the problem I was attempting to cure and had not yet properly identified. The waveform correction addition that address loosing part of the TX coils damping resistive load after the clamping diodes stop conducting is working well. The clamping diodes in the above circuit also provide desired signal isolation so that the actual desired receive signal remains unchanged when the circuit is used as can be seen by looking at the 1Mz since wave that was injected into the simulated TX coil as shown in the diagram above.

The blue trace between the two vertical lines shows the time were the fast channel sampling takes place. This is a old picture just for reference.

[Auto-Mation-Assist]

I was getting ready build this into the detector that I have using for testing and noticed that I had left out a Schottky diode D4 in my prior post which was not required for the simulation is in my actual circuit diagram. This diode keeps the peak voltage at the input to the op-amps to 599mv when the peak TX coil kick back voltage reaches 659 volts. The updated diagram is below and has some added comments relating to voltages and diodes used in the simulation and the ones actually in my detectors. The file name is the same as the old one.

[Auto-Mation-Assist]

In one of my recent posts I stated this about the negative going pulse that I wish modify, "that this problem is not created by the coils decay waveform". That assumption was not correct and the negative going part of the waveform is an actual requirement for detecting small gold. The negative going portion of the waveform is actually and indication that the coil is slightly under damped which is a requirement for looking for return signals that decay very rapidly. My intent has been to significantly reduce the depth of this part of the waveform while retaining its content and thus reduce the level of the signal to be processed. An other issue with doing this is that it affects signal averaging and would likely force width the width of the sample period to be reduced in order not to loose detect sensitivity.

I have tested the circuit in the above post and did get it working at a slight reduction in sensitivity. I did notice that the detector became a bit more nervous which may be due to injecting some added noise into the first stage of the detector. That could partly explain the slight loss of sensitivity. The parts values that simulation came up with were not workable and I ended up changing some values. I also added several diodes for isolation after resistor R8. I have a updated diagram that I can post if anyone is interested.

[Carl-NC]

I always try to slightly underdamp the coil, but not enough that the overshoot lobe rails out the preamp. I've tried to follow this thread but it's spread out over time, so I should go back and re-read everything. Interesting stuff.