Target response tester

[Skippy]

If you're wanting proper discrete log amp circuits for your simulation, here's a couple of Nat Semi app notes that cover the basics. Issues like temperature stability won't matter, as your simulation won't vary with temperature.

http://www.ti.com/lit/an/snoa575b/snoa575b.pdf

http://www.ti.com/general/docs/lit/getliterature.tsp?literatureNumber=snoa641b&fileType=pdf

[green]

Trying to simulate a log decay in spice with ability to add noise and or offset to the original signal. Couldn't find a log amplifier in spice. One possibility is make one. Another option? Think the simulation decay trace is correct but the Y scale is log instead of linear. Simulation has +-5mv of noise with zero offset. Change V2 to change noise and offset. Thinking decay trace isn't going to be any better than what I'm doing now. Is the simulation decay trace correct or am I missing something? The zip file opens in my computer but not here. Running security scan never stops. Anyone know why?

Tried to open a file from another thread, same problem. Guessing maybe the last computer update is causing the problem. I'm computer stupid, any suggestions where to look or what to try? Another question, how to do multiple passes on one screen in spice. Example: TRT_5, V2 offset -5mv, 0mv and +5mv on one screen. TRT_5, V2 used 5mv noise, zero offset in spice example. About the noise level I get at amplifier out, don't know if 5mv is good or bad. Including some scope pictures of amplifier out, scope on peak not sample. Wondering if they are what I should expect or should I be able to lower the noise level. Still want to try a log amplifier, just trying to learn something and what to expect.

[green]

Trying to simulate a log decay in spice with ability to add noise and or offset to the original signal. Couldn't find a log amplifier in spice. One possibility is make one. Another option? Think the simulation decay trace is correct but the Y scale is log instead of linear. Simulation has +-5mv of noise with zero offset. Change V2 to change noise and offset. Thinking decay trace isn't going to be any better than what I'm doing now. Is the simulation decay trace correct or am I missing something? The zip file opens in my computer but not here. Running security scan never stops. Anyone know why?

[Ferric Toes]

Here is a plot showing the small difference in the decay of a silver quarter (green trace) and a clad quarter (yellow trace). I had to to this in persistence mode, hence the fuzzyness of the trace which accumulates a lot of random noise. TX width was fixed at 1000uS.

Eric.

[green]

Hi Green, A clad quarter has a decay that is almost the same as the silver one, so my results will be valid for both. The TX width for the quarter was not optimum for the test as it should have been about 600uS and was in fact 400uS. Forgot to readjust it. Also you say 'Very good trace after 5usec delay'. The target trace starts after 10uS and I just wonder if there was confusion as the TB10uS is under the second vertical line which in reality is 20uS.

I have on order a PicoScope 2206B which should give improved results with their latest P6 software. It also has log X and Y axes as well as linear, but I may not be able to use them without some reprogramming as they only become active in the FFT mode. The Pico engineer I spoke to seemed quite keen to suggest it to the applications people.

Eric.

The target trace starts after 10uS and I just wonder if there was confusion as the TB10uS is under the second vertical line which in reality is 20uS.

Thanks, you are correct, I did get confused. I was using 10usec/div but wasn't thinking when I replied 5usec, saw the 10usec under the second line just like you replied.

[Ferric Toes]

Hi Eric, like your pictures. Very good trace after 5usec delay.

Including some pictures trying to repeat your test. Used a clad quarter since I don't have a silver quarter. My scope screen pictures aren't big enough to get TC numbers. Copied recording into Excel and charted, single recording not target-no target recording like I normally do. Our nickel traces crossed 20minor divisions at about 8usec. My quarter crossed 20 minor divisions at about 40usec instead of 30usec where yours crossed, changed target distance trying to get 30usec at 20 minor divisions for the 125 and 500usec Tx . Excel chart about the same size as your scope screen pictures. About the same readability.

Tried to get TC from your pictures. Nickel: time between 10 and 20 minor divisions 6.25usec/.7=9usec or time between 1 and 20 minor divisions 30usec/3=10usec. Subject to readability.
Quarter:time between 10 and 20 minor divisions 63usec/.7=90usec or time between 1 and 20 minor divisions 470usec/3=157usec. Subject to readability.
Looks like with a good zero and enough readability at 1 division the calculated TC isn't much different at 125 or 500usec Tx with my Excel chart. Maybe a 10% difference using time between 10 and 20 minor divisions. Wondering why the TC difference is higher between the two methods with your quarter. Maybe I'm reading it wrong. What was your Tx width?

Hi Green, A clad quarter has a decay that is almost the same as the silver one, so my results will be valid for both. The TX width for the quarter was not optimum for the test as it should have been about 600uS and was in fact 400uS. Forgot to readjust it. Also you say 'Very good trace after 5usec delay'. The target trace starts after 10uS and I just wonder if there was confusion as the TB10uS is under the second vertical line which in reality is 20uS.

I have on order a PicoScope 2206B which should give improved results with their latest P6 software. It also has log X and Y axes as well as linear, but I may not be able to use them without some reprogramming as they only become active in the FFT mode. The Pico engineer I spoke to seemed quite keen to suggest it to the applications people.

Eric.

[Skippy]

Quote:"I was wondering why the TC difference is higher between the two methods with your quarter?"
I think the likely answer is in part related to where you choose your 'Zero' on the vertical scale. Your estimate of a drop from '20' to '1' might actually be '19.7' to '0.7' , which is a drop to (0.7/19.7) = 0.035 rather than (1/20) = 0.05. Hence your division factor would change from ln(0.05) = 3.00, to ln(0.035) = 3.35, which is 12% different. I admit it's not explaining-away all the discrepancies. If you measure decay from '30' to '15', you get TC = (85 usec / 0.693) = 123 usec. Similarly from '10' to '5' gives (85 usec / 0.693) = 123 usec. both more realistic.
If you drew multiple versions of the graph with a log vertical scale, and used different 'zero' values, you'd find one with a decent straight-line, I'm sure.
It's also likely that readings before t = 50 usec may be skewed by the initial more rapid fall-off that your (Green) charts show. The problem is that on a linear vertical scale, deviations from the ideal constant TC don't show up, it's all curve.

You will be somewhat surprised to know that Eric's "5c/nickel" chart actually matches what VLF machines make of that coin. Info on the web will tell you both George Payne (Teknetics etc) and Troy Calloway (Troy Detectors) say it has a f_c about 17KHz. I've measured it (at 13KHz) as f_c = 16KHz; TC = 9.95 usec. This time, the target coin and my machine's freq are well-matched, so the results should be good. And... interestingly, the skin-effect doesn't have much influence, either. Skin depth is very dependant on material conductivity, and affects silver/copper the worst, and bad conductors like Cu-Ni much less. So my VLF will actually see all the 5c coin.

Here's a wiki page on Skin effect, it mentions the difference between Cu and Aluminium being significant, and they're both quite good conductors.

https://en.wikipedia.org/wiki/Skin_effect

Cu-Ni has a conductivity about 5% IACS, so it's skin depth should be Square-root (20) = 4.5 times that of copper / silver / the US 25c 'quarter' coin.

[green]

I've resurrected my digital set-up with a eight year old BitScope module and laptop to get better pictures. These I have inverted so as to get the decay the more conventional way up. Much better quality is achieved this way as the picture is saved to computer memory and no camera is involved.

1st picture is without target in coil. 2nd is with a Nickel and 3rd is the silver quarter. Time base 10uS for the first two and 50uS for the quarter.

Eric.

[ATTACH]41900[/ATTACH] [ATTACH]41901[/ATTACH] [ATTACH]41902[/ATTACH]

Hi Eric, like your pictures. Very good trace after 5usec delay.

Including some pictures trying to repeat your test. Used a clad quarter since I don't have a silver quarter. My scope screen pictures aren't big enough to get TC numbers. Copied recording into Excel and charted, single recording not target-no target recording like I normally do. Our nickel traces crossed 20minor divisions at about 8usec. My quarter crossed 20 minor divisions at about 40usec instead of 30usec where yours crossed, changed target distance trying to get 30usec at 20 minor divisions for the 125 and 500usec Tx . Excel chart about the same size as your scope screen pictures. About the same readability.

Tried to get TC from your pictures. Nickel: time between 10 and 20 minor divisions 6.25usec/.7=9usec or time between 1 and 20 minor divisions 30usec/3=10usec. Subject to readability.
Quarter:time between 10 and 20 minor divisions 63usec/.7=90usec or time between 1 and 20 minor divisions 470usec/3=157usec. Subject to readability.
Looks like with a good zero and enough readability at 1 division the calculated TC isn't much different at 125 or 500usec Tx with my Excel chart. Maybe a 10% difference using time between 10 and 20 minor divisions. Wondering why the TC difference is higher between the two methods with your quarter. Maybe I'm reading it wrong. What was your Tx width?

[Ferric Toes]

I've resurrected my digital set-up with a eight year old BitScope module and laptop to get better pictures. These I have inverted so as to get the decay the more conventional way up. Much better quality is achieved this way as the picture is saved to computer memory and no camera is involved.

1st picture is without target in coil. 2nd is with a Nickel and 3rd is the silver quarter. Time base 10uS for the first two and 50uS for the quarter.

Eric.

[green]

Think maybe 3 decades is all I'll need do to amplifier noise out.

Statement I made reply#2. Looking at what amplifier noise out does to the log out signal, 2 decades of usable signal is about all I expect to get. About what I'm getting now but the signal should be a smother with a little less noise. Be interesting to see which way is best. I had another X9 amplifier(x450x9x9) that increased the range to about 3 decades. Would work best with a 4CH scope and the amplifier went bad so I haven't repaired it.

[Skippy]

Capcitively-coupled? That sounds like the other type of 'log strip' amplifier, that are used in radio gear, to cope with widely varying signal strength, as an alternative to AGC. (they used to be horribly expensive, too, as you had to use multiple devices in cascade to get a decent compression range)

The photodiode log-amps seem DC coupled, but they may need some design tinkering to make them 'general purpose' , eg. the 0.5V offset on the input. I haven't read the datasheets / app notes for them, if you use dual supplies, you can probably get round that offset, it's really intended for single-supply operation.

[green]

"Not a problem now. The MAX4207 is $4.05......"
I'm aware of this... you've forgotten my post in the other thread already?

Home-brewing one from 2 or 3 opamps, a matched-transistor pair isn't that hard. They always tended to use really expensive super-matched transistor pairs, LM394, MAT-01, but with the availabilty of dirt-cheap dual-transistors in 6-pin SOT-23 or similar package, that 15 dollar part becomes a 15 cent part if you're not after super-precision. That's what was used in the design I worked on. The tricky part was the special resistor with about +0.3%/ degree C (proportional to Kelvin) tempco , which we wanted in SMD, and they seemed to be more common in leaded packages, including radial , that stood up high. I would say that if you're doing your experiments at home in a stable domestic temperature, you could miss out that part. Except you said you work in a garage which can be cold or hot depending on the season, which makes things more complex.
These are the things I'm referring to:

http://www.precisionresistor.com/AT-and-ST-Series-1K-3500-PPM-T.C.R.-Compensators/

It looks like audio guys use them in synths and effects units:

https://www.thonk.co.uk/shop/tempco-resistor-akaneohm/

https://www.thonk.co.uk/shop/tempco-resistor-krl-bantry-3500ppm-x2/

I ought to have some samples among my junk, but I dunno..

I could dig out some opamp/transistor logamp circuits from my library of stuff if you're interested.

Well you could build your own log amp. I'm not familiar with the latest IC's, but they are intended for photodiode sensing in the optical comms business. They're cheap (5 dollars?) and their limitations don't seem to be a problem. Examples from AD:

http://www.analog.com/en/products/am...mplifiers.html

Getting one in a package bigger than a pinhead could be a problem... though there's always the Evaluation Boards, thankfully.

That was the reason for looking at other log amps. Did miss or forgot the pinhead remark. The ones I looked at from AD seemed to be capacitor coupled. Think I'll try to select one of the log IC's instead of trying to make a log amplifier. Looked at the 810 Eric suggested, probably don't know enough about potential problems to make the correct selection.

The MAX4207 is the lowest cost and it looks like it might work so might be what I'll try first.

[Skippy]

"Not a problem now. The MAX4207 is $4.05......"
I'm aware of this... you've forgotten my post in the other thread already?

Home-brewing one from 2 or 3 opamps, a matched-transistor pair isn't that hard. They always tended to use really expensive super-matched transistor pairs, LM394, MAT-01, but with the availabilty of dirt-cheap dual-transistors in 6-pin SOT-23 or similar package, that 15 dollar part becomes a 15 cent part if you're not after super-precision. That's what was used in the design I worked on. The tricky part was the special resistor with about +0.3%/ degree C (proportional to Kelvin) tempco , which we wanted in SMD, and they seemed to be more common in leaded packages, including radial , that stood up high. I would say that if you're doing your experiments at home in a stable domestic temperature, you could miss out that part. Except you said you work in a garage which can be cold or hot depending on the season, which makes things more complex.
These are the things I'm referring to:

http://www.precisionresistor.com/AT-and-ST-Series-1K-3500-PPM-T.C.R.-Compensators/

It looks like audio guys use them in synths and effects units:

https://www.thonk.co.uk/shop/tempco-resistor-akaneohm/

https://www.thonk.co.uk/shop/tempco-resistor-krl-bantry-3500ppm-x2/

I ought to have some samples among my junk, but I dunno..

I could dig out some opamp/transistor logamp circuits from my library of stuff if you're interested.

[Ferric Toes]

Thinking about updating my target response tester to include a log amplifier and other modifications to compare with the decay curves I've been doing. Tx rate (1pps), Tx shape(constant current .5A peak or constant rate 1A peak), Tx width(adjustable 50usec to 50msec). Preamp(gain 450),post amp(gain 9). log amplifier(after preamp).

Discussion: constant current peak 1/2 of constant rate peak for same average Tx current, the lower the constant current peak the shorter the ramp time to peak, preamp and post amplifier the same as I've been using, need to modify constant rate circuit to allow adjusting Tx width while keeping 1A peak, Tx rate 1pps for low average Tx current,

Questions: any reason to make Tx rate adjustable? any changes or additions?

Comments or thoughts appreciated, would be easier to include changes now than modify latter.

Have a look at the Texas Instruments VCA610 and VCA810, both of which are wideband voltage controlled amplifiers. The 610 can be used to provide a 250kHz bandwidth while the 810 stretches out to 2.5MHz. The data sheets available online give all the details in the applications information. I have a few VCA610's from some years back and I did a quick test then and it looked promising, so I will have another go. Mine are DIP 8 which are easy to use.

Eric.

[green]

Unfortunately, the last time I used a log-amp (for a photodiode amp) it was a discrete design, in part due to some specific technical requirements, and partly due to the available IC's all being expensive, needing too many trimming pots, and usually in a big DIL package

.

Not a problem now. The MAX4207 is $4.05 for one in a 4x4mm 16pin package. Not sure I'll be able to solder to it.