Target response tester

[Skippy]

No Spice, I have occasionally been curious about simulation, but never enough to bother trying it.

It seems practical problems like real-world noise limit how much of a range you need, so maybe trying to perfect the low-level input behaviour of this discrete amp is not productive?

[green]

No Spice, I have occasionally been curious about simulation, but never enough to bother trying it.

It seems practical problems like real-world noise limit how much of a range you need, so maybe trying to perfect the low-level input behaviour of this discrete amp is not productive?

Thanks again for your help. Building the log amp in spice helped to understand what I needed to do. When testing the log circuit the other day I would sometimes get more than 800mv log out. 2.5 volts in(+PS) should be 800mv log out. After more time than it should I finally determined the scope over ranging was causing the problem so I added a resistor and diode to clamp the output going to far negative. Wasn't thinking when trying the capacitor decay, the clamp was causing the problem. Including schematic of log amp, if I had done sooner someone would have seen the problem. The 1usec decay looks like it could be better, need to try higher current. The diode is disconnected from common for todays tests. I tried a amplifier(offset adjusted to less than 20uv)between the capacitor discharge and log amp yesterday and it worked after lifting the diode. Today I tried capacitor direct to log amp and didn't see an offset problem.

I tied pin 1 of LOG114 to V- instead of common, forgot to change schematic. Not used so probably doesn't matter.

[Ferric Toes]

I have no Spice either, so I too am missing information with the .zip files.

I am building the logamp described in the applications data for the Burr Brown VCA610. It will be interesting to see how this approach fares.

Eric.

[green]

I have no Spice either, so I too am missing information with the .zip files.

I am building the logamp described in the applications data for the Burr Brown VCA610. It will be interesting to see how this approach fares.

Eric.

LT spice is free and not hard to learn. Some of the members are a lot better with it than me. I see some of them doing things I have no idea how to do. Doing some of the easy things has helped me a lot. I'll add a picture of an analysis plus the zip file for reference. Would have been better if I had added my schematic for the log amp sooner. Someone could have caught my simple mistakes.

[Qiaozhi]

LT spice is free and not hard to learn. Some of the members are a lot better with it than me. I see some of them doing things I have no idea how to do. Doing some of the easy things has helped me a lot. I'll add a picture of an analysis plus the zip file for reference. Would have been better if I had added my schematic for the log amp sooner. Someone could have caught my simple mistakes.

Buy yourself a copy of this book -> https://www.amazon.co.uk/SPICE-Circu.../dp/0131587757
Highly recommended!

[bbsailor]

I just thought I would run a plot for the silver quarter and clad quarter and here is the result. The top trace is for the two quarters and there is little difference. The midde trace for comparison is a US copper penny. The third trace is a US Cent.
Ignore the rulers and figures as I did not cancel them, and the graph is 500uS full scale rather than 1000uS in the previous ones. The quarters have not yet reached zero in this plot.

[ATTACH]42249[/ATTACH]
Eric.

Eric,

If we step back and take a more macro view of pulse characteristics for both the specific targets as well as the ground characteristics being hunted, we (collectively as Geotech1 forum members) may be able to provide some meaningful design information to get PI detector builders be better able to optimize their building efforts that better match the following:

1. Primary search area ground characteristics
2. Coil size and inductance for a range of target sizes, target Time Constants and ground characteristics
3. Coil style: Mono, DD, figure eight, concentric, probe
4. TX pulse characteristics optimized for specific ground conditions and target types
5. RX window optimized for specific ground conditions and target types
6. Trade-offs and methods to analyse these trade-offs
7. Desired variables to look for in micro-processor controlled PI detectors and analog controlled PI detectors.
8. Established tips and techniques for approaching the construction of a PI metal detector and optimization of related variables

In reading your forum and this Geotech1 forum over the years, I believe the above is a synthesis of the interrelationship of key topics that would benefit from a technically supported context.

Thanks

Joseph J. Rogowski

[green]

I have no Spice either, so I too am missing information with the .zip files.

I am building the logamp described in the applications data for the Burr Brown VCA610. It will be interesting to see how this approach fares.

Eric.

Wondering how your log amp is working. I lowered the input resistor like you and Skippy suggested. Looks better. The mv/decade changed so I had to make the gain adjustable. Doing some testing, will post some data when I think I've got it best I can.

[Skippy]

I've had a read of the LOG114 datasheet, seen a few things of importance, and come up with a few suggestions.

They talk about 8 decades range --- this is for CURRENT, if you put in volts through a resistor to the virtual ground, it performs badly, as the virtual ground can be 4mV off, the datasheet says 3 decades range only, eg. 10 Volts -> 10mV.
So you're going to have to drive it hard on the input, but not above 3mA or it goes 'off'. I suggest, based on your input voltage max being about 1.6 Volts, you have this drive 1.6 mA into the IC. Hence choose the input resistor (R3) = 1k0.
For your 'zero-volt' output, I suggest you choose this to occur at input current of 10mV. Hence your reference input current needs to be (10mV / 1k0) = 10 microamps. (eg. 250K to +2.5V rail)
Your output scale factor needs increasing, I suggest 800 mV / decade, which means post-amp gain of 2.13333 . (for example 4K7/2K2; 10K/4K7; 12K/5K6; 16K/7K5 ) .... the last one is easiest, just add another 15K across your existing one to make 7K5.
This would give a maximum logamp output of about +1.7 Volts, I assume it can do this from 2.5V supplies? And minimum output would be equivalent to 25 microvolts.

If you need more range, you're going to have to consider a voltage-to-current converter, that can be done with a modest opamp circuit, a Howland source:
http://michaelgellis.tripod.com/howland.html
(I've never tried it, the V -> I converters I've previously used were ones with a transistor in the loop, but they were sinks, not sources)

In addition, if you split resistor R5 'in half', eg 24K & 24K, you could decouple the tapping point to ground with a 1u capacitor, similar to the reference input circuit, and reduce noise injection.

[green]

More data with higher log amp input current. Changes to log114c.png, R1=200k, R2=2k, R3=2k, R9=1k pot(adjust log out volts to 400mv/decade) and D1 not connected to common.

The capacitors were connected to the R3 input and common. A 1N4148 anode was connected to Tx command(+-2.5v step signal), cathode to R3(log amp input). When command signal goes to -2.5v, capacitor discharges into the R3(2k input resistor). NPO ceramic capacitors, don't know how close they are to rated value.

Need to get PI coil and amplifier to decay to zero volts faster to get 2 decades of straight decay with short TC targets.

[green]

I've had a read of the LOG114 datasheet, seen a few things of importance, and come up with a few suggestions.

They talk about 8 decades range --- this is for CURRENT, if you put in volts through a resistor to the virtual ground, it performs badly, as the virtual ground can be 4mV off, the datasheet says 3 decades range only, eg. 10 Volts -> 10mV.
So you're going to have to drive it hard on the input, but not above 3mA or it goes 'off'. I suggest, based on your input voltage max being about 1.6 Volts, you have this drive 1.6 mA into the IC. Hence input resistor = 1k0.
For your 'zero-volt' output, I suggest you choose this to occur at input current of 10mV. Hence your reference input current needs to be (10mV / 1k0) = 10 microamps.
Your output scale factor needs increasing, I suggest 800 mV / decade, which means post-amp gain of 2.13333 . (for example 4K7/2K2; 10K/4K7; 12K/5K6; 16K/7K5 )
This would give a maximum logamp output of about +1.7 Volts, I assume it can do this from 2.5V supplies? And minimum output would be equivalent to 25 microvolts.

If you need more range, you're going to have to consider a voltage-to-current converter, that can be done with a modest opamp circuit, a Howland source:
http://michaelgellis.tripod.com/howland.html
(I've never tried it, the V -> I converters I've previously used were ones with a transistor in the loop, but they were sinks, not sources)

I was posting reply #69 while you were posting #68.

The input can go as high as 2.5 volts. Tried 2k input resistor which gives 1ma at 2volts in. R6 in the schematic adjusts offset.

I'll think about the other suggestions, something else to do right now.

[Skippy]

While I was thinking about the LOG114 circuit, I came up with some alternative output gain figures. I thought they may be of use, so I'm posting them here:
Output scaling = 500 mV /decade. This is almost exactly 150 mV / octave, so is a good 'dual-scale' output.
Post-amp gain needed = 1.3333; plenty of R pairs here: 16k/12k; 20k/15k; 24k/18k; 36k/27k; 6k8/5k1.

Output scaling = 200mV /octave. (equals 664 mV / decade )
Post-amp gain needed = 1.77; R pairs include 39k/22k; 12k/6k8.

I'm using the term 'octave' to mean a doubling/halving ratio, as in filter specs like -6dB/octave, hope I got that right. There seems to be a lack of words for these 'ratios', I'm sure I've seen 'hexade' used for a 16-fold change, and maybe 'dodecade' is 20-fold?

[green]

While I was thinking about the LOG114 circuit, I came up with some alternative output gain figures. I thought they may be of use, so I'm posting them here:
Output scaling = 500 mV /decade. This is almost exactly 150 mV / octave, so is a good 'dual-scale' output.
Post-amp gain needed = 1.3333; plenty of R pairs here: 16k/12k; 20k/15k; 24k/18k; 36k/27k; 6k8/5k1.

Output scaling = 200mV /octave. (equals 664 mV / decade )
Post-amp gain needed = 1.77; R pairs include 39k/22k; 12k/6k8.

I'm using the term 'octave' to mean a doubling/halving ratio, as in filter specs like -6dB/octave, hope I got that right. There seems to be a lack of words for these 'ratios', I'm sure I've seen 'hexade' used for a 16-fold change, and maybe 'dodecade' is 20-fold?

Don't know if this makes sense, but it's the reasoning I used to set output scaling at 400mv/decade. Scope scale, 100mv, 200mv or 500mv/division. Wanted major divisions/decade.

With 400mv/decade. Scope 200mv/major, 2major(10 minor)div/decade, 4decades total. Scope 100mv/major, 4major div/decade, 2decades total. Scope 500mv/major, would need 1v/decade to get 2div/decade, scope doesn't have 250mv/major(400mv/decade looked like a better option).

Target time constant(straight line decay on linear X log Y chart)=time difference/2.3*(number of Y div/decade)/number of Y div crossed. TC=time difference/2.3 for 1 decade crossed, /4.6 for 2 decades crossed, /6.9 for 3 decades crossed. If the data was copied to Excel it wouldn't matter because Y scale could be .375v/major(10minor div/major) with .375v/decade

[Skippy]

I understand your reasoning, having awkward scope screen scaling is best avoided, though scaling such as 2.5 major divisions / decade isn't too troublesome (this would be 500 mV/decade logamp output, with 200 mV/div on the scope).
I was just thinking about the spare general purpose opamp in the LOG114, which could be used to provide an alternative output scaling, so you had more choices, to help cope with a wide variety of different targets.

Just curious, what's involved in getting data out of the scope and into a PC ? Presumably it's a USB connection. Does it just allow captured data transfer, or can you do more advanced stuff, like having the PC control the scopes behaviour, with the PC monitor showing what the scope screen does, etc?

[green]

I understand your reasoning, having awkward scope screen scaling is best avoided, though scaling such as 2.5 major divisions / decade isn't too troublesome (this would be 500 mV/decade logamp output, with 200 mV/div on the scope).
I was just thinking about the spare general purpose opamp in the LOG114, which could be used to provide an alternative output scaling, so you had more choices, to help cope with a wide variety of different targets.

Just curious, what's involved in getting data out of the scope and into a PC ? Presumably it's a USB connection. Does it just allow captured data transfer, or can you do more advanced stuff, like having the PC control the scopes behaviour, with the PC monitor showing what the scope screen does, etc?

I have a RIGOL DS1052E, lower cost scope but works for the application. USB connection to a memory stick. Store bitmap for scope pictures or CSV to copy to Excel. Don't think it can do advanced stuff.

More test data. Charted the quarter at two distances, one amplifier saturated for about 30usec and another with amplifier not saturated. Three different record times. To determine TC, record time should be 4 or 5 times target TC. With the detector I'm building the target sample + ground sample is taken in less than 150usec. It's hard to see what happens the first 100usec where the decay isn't straight line decay with 600usec full scale. Trying to determine the best scaling and if saturated or not saturated amplifier is best.

I get a glitch in the no target picture that shows up in some of the other pictures, not always at the same signal amplitude. Two of the foil pictures it peaks at about 1.5volts, with some of the quarter amplifier saturated pictures glitch is near full scale. Any ideas what is causing it or how best to determine what is causing it?

I'm starting to like the log amp vs the way I was charting the data.

Charted California ground with Excel to get log X scale.

I see the first picture shows 50usec line instead of 150usec line like it should be.

[green]

Have seen a reply in another thread about adding resistance in series with the coil so it flattens out before turn off being better. Constant current vs constant rate Tx. With the same average Tx current, constant rate gives a higher Rx signal than constant current. Did a test with a quarter with Tx width 1, 2 and 4 times the TC awhile back. Tx width from 2x the TC to 4x the TC made a larger increase with constant rate than constant current. Any thoughts why? Repeated test with 1oz copper coin. Similar results, more change with the copper coin than quarter with constant current.

Skippy mentioned octave in another reply. If I'm doing my math right, 400mv/decade with 10 minor div/decade three minor div would be an octave?