My AGD analog detector work

[Mr.Jaick]

Hi Auto-Mation-Assist
How much performance improvement do you get for that 0.5 to 1 us of faster sampling on a 0.5-gram nugget? just curious.

[hunter63]

Hello. Would your friend share the PCB documentation in PDF format?

[Detectorist#1]

Hi AMA,
I admire your efforts to go very deep in the problems of every stage of PI MD. All we see that you have deep knowledge in hardware. It is good that you post time-diagrams with nanoseconds details. But when I try to use
pinpointer with 3us first delay, I found that in real world I need to increase this delay to 5us to avoid false signals from wet grass. Maybe more of your efforts to do extremely precise PI MD are not useful in real practice.
In this time, maybe Pito is right - post real movies for your MD.

[Detectorist#1]

Hi AMA,
Last year, I tested diffused capacitance of some fast diodes in attempt to use them as blocking diodes in the input of PI MD and for serial diodes for MOS FETs in end TX stage. Situation is similar - no time when
these diodes are under reverse voltage especially for blocking diodes. In this situation "reverse recovery time" in datasheets is not applicable. With my test circuit, I found that super fast schottky diodes needs
serious more time than simple low Si diodes as 1N4148 to switch-off. This is marked by Carl from long time. If you use schottky diodes, you first do a problem and after that you look for fix the problem.

[Detectorist#1]

Hi AMA,

When I tested BAT81S schottky diode ( 40V 30mA 1ns) , my tester pointed 2us time for discharge after 0.37A current with 1KoHm discharge resistor. You use 6KoHms discharge resistor in your schematic!
Yes, 670 volts by 6KoHms give appr. 100mA current, but the problem stay in this circuit.

[Auto-Mation-Assist]

I have not had a chance to read everyone's latest comments carefully yet since I have been very busy laying out the circuit boards for my AGD24.1 version. The AGD24.1 version will have two back to back circuit boards. I have completed design of those boards and submitted them for manufacturing to my board house and expect them to get to me before the end of March. In the past I ran out of room on the main circuit board and ended up adding a separate board to handle the TX pulse and this has worked very well especially after adding the wave shaping circuit to it.

The 24.1 version will retain all its analog circuitry but will allow for more flexibility and space for replacing the CMOS logic and VCO to become MCU controlled later on.

One of my goals was to move the CMOS logic from its -10V power source to the +10 volt source so that can get rid of the level shifting circuits presently used. These use zener diodes which cause considerable spiking on the leading and falling edges due to their high capacitance. The other thing reduce front end noise as much as possible by reducing the resistor values in the first gain stage which includes the clamping diodes and the series resistors that feed them, and is also part of the TX coil load. The TX coil damping oad is a bit more complex because of the added wave shaping circuit that starts to be active after the clamping diodes stop conducting.

I have disassembled two of my detectors to get them ready to use the new circuit boards. Since the boards will be plugged together back to back I need to be able to get to the connectors on the lower board that connect to the front panel controls. This required machining the bottom of the PCB mounting trays so that they have large openings on the bottom side. I did fill up some old holes with quick set JB Weld.



AGD24.1 info:

The lower board, AGD24.1B will have the audio amp, demodulated signal handling, VCO and CMOS Logic. If a MCU is added it will be on this boards replacement. The lower board will take its power from the upper board.

The upper board, AGD24.1.A will have all the TX, RX, Sample and Hold, Demodulated signal Low Pass Filters, Plus and Minus 10 volt power supplies, and the isolated 12V to 12V DC/DC converter for the +10 regulated supply. It also outputs -12 battery power to the lower along with the +10/-10V regulated power.

The upper board has inputs for DC/DC converter on/off, TX pulse, and RX gating pulse, RX pulse 0, RX pulse 1, RX pulse 2. It outputs, Coil has decayed signal, Chan 0 and Chan 1 demodulated signals, Regulated power good. Its basically a complete analog PI TX/RX system that only needs control signal inputs.

The lower board can be replaced with anything that can control the upper board if desired and thus allows for adding user desired features.

I was looking at the current timing parameters that I use right now.

Loop rate 266us
TX pulse width 40us,
Decay end detection 2.62us after TX pulse ends,
Close first RX gate 480ns later, (this gate stays closed for the total time of the two channels below).
Close the fast channel gate 170ns after that, (keep it closed for 1.96us this is my minimum).
Close the slow channel gate 80ns after all that, (keep it closed for 8.8us this is my minimum).

I have never needed to use the the mixing feature of the fast and slow channel. Mostly just used during testing to make sure it functions properly.

[Auto-Mation-Assist]

There is always a not of interest in what diodes to use for the front end protection clamping diodes. I have tried a good number of different types and all worked to protect the front end, even the ones that are so small that they become a challenge to put on a circuit board. Fortunately they only need to handle fairly high current values (based on their physical size) for a fraction of a microsecond. The diodes that I use have proven to work very well.

The data sheet for the diodes that I have been using is attached.

[Auto-Mation-Assist]

In one of my earlier posts I described a way to use a separately generated decay wave and subtract its output using the differential inputs of the first gain stage. I did develop a circuit broad to test this and it was able to do the job. I think that further testing is required. I did find that with the waveform corrector circuit that I now use in the TX board that it does not to have part of the decay waveform eliminated by the use for the circuit board in the picture below. This board was also designed to be a shield for the RX section in the ADG23 V detectors to reduce picking up RF signals for nearby transmitters at my location. One of the connectors for the TX board to plug into is on the left.

I should also mention that having the RX and TX functions operate on separate circuit boards has not presented any problems.

[Auto-Mation-Assist]

A while back there was a comment that the AGD detector provided a nice eye pattern. The AGD detector has a built in test mode that allow control of the analog gates without having a TX coil present. This allows for testing the response of the entire receive circuitry using test equipment. Below is a picture of an eye pattern using a 200Khz sine wave. It shows the analog switching after the first analog gate and the following fast channel gate. The slow channel is not shown in this picture.

[Auto-Mation-Assist]

One of the most important things to check is the frequency response of the first analog gain stage and how it responds to fast rising signals. If there is significant overshoot on fast rising signals then a real problem exists since it will equal a increase in gain which can lead to serious signal clipping and thus cause part of the receive wave form to become useless. The result will be terrible sensitively to small gold. Remember that at the times these peaks occur the voltage level of the decay wave form is the highest, you can easily run out of head room and drive you gain stages in to severe clipping.

Here is a picture of one of my own detectors first gain stage without proper compensation. The test was using a 200Khz square wave at the RX input connector.



The peaks may not look like a lot but in fact they represent a real problem that must be corrected. After correcting this the wave form will look like the below picture.



This picture is my actual first gain stage response to a 200 Khz square wave.

If to much compensation is used the areas were the peaks were will be reduced to a level lower that the straight line part and lower the RX bandwidth.

The severity of the peaks sown in the first picture can vary greatly depending on circuit design, parts used and circuit board layout. No peaks with a lot of roll off indicates insufficient front end bandwidth. This could be because the opamps are not fast enough for this use or there is a lot of high frequency signal loss before the RX signal reaches the amps inputs.

There is no way to test for these kind of things without proper test equipment, small details can make a lot of difference in performance.

[moodz]

One of the most important things to check is the frequency response of the first analog gain stage and how it responds to fast rising signals. If there is significant overshoot on fast rising signals then a real problem exists since it will equal a increase in gain which can lead to serious signal clipping and thus cause part of the receive wave form to become useless. The result will be terrible sensitively to small gold. Remember that at the times these peaks occur the voltage level of the decay wave form is the highest, you can easily run out of head room and drive you gain stages in to severe clipping.

Here is a picture of one of my own detectors first gain stage without proper compensation. The test was using a 200Khz square wave at the RX input connector.



The peaks may not look like a lot but in fact they represent a real problem that must be corrected. After correcting this the wave form will look like the below picture.



This picture is my actual first gain stage response to a 200 Khz square wave.

If to much compensation is used the areas were the peaks were will be reduced to a level lower that the straight line part and lower the RX bandwidth.

The severity of the peaks sown in the first picture can vary greatly depending on circuit design, parts used and circuit board layout. No peaks with a lot of roll off indicates insufficient front end bandwidth. This could be because the opamps are not fast enough for this use or there is a lot of high frequency signal loss before the RX signal reaches the amps inputs.

There is no way to test for these kind of things without proper test equipment, small details can make a lot of difference in performance.

Good and precise work ! Are you able to indicate the frequency response of your preamp.

Below is the typical response of the pream I use for PI detectors @ around 35 microsecond TX on time.

[Auto-Mation-Assist]

MOODZ, I will post bode plots for my old and new versions later this week. Many thanks for posting your bode plot.


On an other subject not related to the above. I have been getting ready to figure out proper values for the TX coil damping resistors in the 24.1 version of the AGD detector. Since there will be a increase of the load presented to the TX coil by the new preamp all resistor values that relate to coil damping will need to be changed with the exception of the TX-Coil-Decay-Control-Circuit of which I posted the design for in some older post. I referred to it before as a waveform corrector.

This post will concentrate on the now named "TX-Coil-Decay-Control-Circuit" because it it a very important addition and usable in many PI detectors with only minor changes in the detectors a long as they have the required power sources to power the circuit itself. The issue that this circuit is designed to solve is that when the TX coils decay voltage decays to a point lower than the voltage for the RX clamping diodes to conduct the load presented to the coil becomes basically a open circuit. This causes a large voltage spike during the first 0.5us of the signal being routed to the receive preamp as shown below and for the most part should be considered to be a very large error.



The picture indicates that this spike is 32.955mV and when amplified by the detectors RX signal gain amplifiers it can become amplified enough to cause signal clipping following gain stages. Again the reason for this spike is that there is no longer a damping or as we call it in the RF field a lack of termination of proper impedance. Trying to provide a termination with a simple resistor to solve this is not very practical because it will need a very large decrease in damping resistance to settle down this spike which will cause decay time to increase by many microseconds. Fortunately this problem can be fixed without increasing decay time, but by decreasing decal time and also at the same time increasing recovered signal level by up to 30% as compared to just damping the coil more with a simple resistor.

Here is a picture that shows that the problem spike is no longer there and that a much better damping control is being used. At the 32.955mv it is more difficult to see the actual crossover times to a near zero volt condition. For that I will use a 2uv level later on.




Lets look at the 2uv samples now

For the the fix using a simple resistor below you can see that the curve decays to zero with no spike being produced, but the decay time is extremely long at 8us and not very use full.



Now lets look at the same 2uv one for the fix which has an active part to automatically adjust the damping load on the TX coil very slightly during the last part of decay and see how it compares to the above. You can see that there is a gain of about 3us less time required to decay for the fix that used active loading to control the final part of the decay curve. The other nice thing is that now you have a fully adjustable damping load by just using a trim pot. The only the original damping resistor values need to be modified to allow the fix ti be installed ince it will replace parts of the normal damping resistor load. Which is what I'm working on now for the 24.1 version of the AGD. This fix is working well in the AGD23.4 version with its updated TX board which has this fix circuitry installed.




Here is a picture and some info on this fix. I hope you try implementing this. It is bound to improve your detector if it has RX clamping methods simulator to AGD detector, and if not it can likely be adapted to work as designed in most detectors. The proper choice of active device is important.

[Auto-Mation-Assist]

MOODZ,

Here are the bode plots you requested along with some additional information that you may find usefull.

23.3 version



23.3 version -1db point



24.1 version



24.1 version -1db point




These are for the first stage which uses parallel op amps for noise reduction.

There are two versions, both are wide band both with a gain of just over 28dB.
They are followed by a separate variable wide band gain stage adjustable for 6 to 18 dB gain with 1db points at 1.3Mhz.
After that is a low pass filter with a gain of 6db and just restricts the same bandwidth slightly. Its main job is driving the sample and hold circuits.

The 23.3 pre-amp has limited bandwidth which to me was a bit slow. Its -1db point was 551.438K with -3dB being at 999.801K - Gains verified with Fluke 8920A
The 24.1 pre-amp has increased speed to twice the 23.2 and (not yet verified) 3 db reduction in noise level. Its -1dB point should be 1.082Mhz and -3db at 1.995 Mhz

Both versions of the detector utilize dc offset control back to the pre-amp to maintain constant VCO control voltage based on surroundings or ground conditions and operates in frequency range below 10 mHz (milli herz).
ALL circuits in the AGD are DC coupled with the audio amp being the exception.

Some of my testing to verify overall sensitivity and sample and hold functions is done with sine or square wave signals between 20 and 0.5Hz.

[Auto-Mation-Assist]

I got word that the lower circuit board for AGD24.1.B has been manufactured and is on the way to me. I hope to have the board populated and tested in the next seven days.

I was able the get the shields that will cover the critical parts of the receiver circuitry covered cut out. These will help a lot to keep RFI out of the front end while working on the work bench when the detector is not in its case, but will also keep wires away from this critical area once the end covers are closed. There is no way of knowing were internal wires from the rear panel end up once closed.

Internal shields for the TX-RX boards. The lower cutout is the connector for the RX coil. The upper one to the left is for the variable gain amp control. and the hole right on top of that is for access to the pre-amp output test point. One per detector.

[Sean_Goddard]

Well, well. Look at that, a PROPERLY designed detector. Rx in screened box on a properly laid out board of its own. No surprise that this machine will be super sensitive and VERY quiet in its final version.