Hi Eric and hope all is well with you. I was reading about some sampling circuits used in ADC's. Some of them use Coherent sampling and some Window sampling. I was wondering what the sampling circuits you use in your detectors are called? Thank you Eric and God Bless. John-Wirechief.
Find's Treasure Forums
Welcome to Find's Treasure Forums, Guests!
You are viewing this forums as a guest which limits you to read only status.
Only registered members may post stories, questions, classifieds, reply to other posts, contact other members using built in messaging and use many other features found on these forums.
Why not register and join us today? It's free! (We don't share your email addresses with anyone.) We keep email addresses of our users to protect them and others from bad people posting things they shouldn't.
Click here to register!
Need Support Help?
Cannot log in?, click here to have new password emailed to you
Hi Eric! Sampling circuits
- Thread starter Wirechief
- Start date
Hello Reg, Coherent is the sampling of a periodic signal with an integer number of its cycles which fit into a predefined sampling window Fin/Fsample=Nwindow/Nrecord. This is what Maxim says anyway. Now you might be coherent do you think Reg? It just caused me to be curious if the sampling circuits in PI detectors have a name or designation. Now if I could grasp FFT's! CU later Reg. Wirechief.
Hi Wirechief,
Eric's machines use analog circuits to process the RX samples. Samples are selected with synchronous (with the pulses) demodulators, and despite the fancy name, could be as simple as analogue switches. Usually only timing devices utilise digital circuitry. All mathematical operations (multiplication, integration, subtraction etc.,) performed on the sampled data is likewise done using analogue means.
The integrator is essentially the main filter (low pass), and it performs the corollary of convolution in the digital model, which is the real time equivalent of FFT in the frequency domain.
Up to this time it has been vastly more economical - design solution weight, size, cost, complexity, power consumption, R&D costs, etc., - to use the analogue solution. However, in just the last couple of years we have seen the price and processing power of the digital solution begin to approach the boundary whereby digital application could soon revolutionise metal detector technology.
Take for instance the microprocessor, about a year ago the industry standard was about 250 MHz, over the last couple of months I've been working on a consumer product that uses a 1 GHz processor. This sort of power coupled with specifically applied high speed A to D's, has the potential to raise the faintest of targets way out of the noise, and shout them out loud to us, and when atomic molecular resonance is utilised, 3D images of the target to boot.
This will depend upon the advances being made in UXO and geophysical sensing flowing over into the hobby market.
I'm sorry John, I didn't mean to give you a lecture, I must be half asleep still. I'm tempted now to just hit the back button and not post.
Oh well...here goes
HH
Kev.
Eric's machines use analog circuits to process the RX samples. Samples are selected with synchronous (with the pulses) demodulators, and despite the fancy name, could be as simple as analogue switches. Usually only timing devices utilise digital circuitry. All mathematical operations (multiplication, integration, subtraction etc.,) performed on the sampled data is likewise done using analogue means.
The integrator is essentially the main filter (low pass), and it performs the corollary of convolution in the digital model, which is the real time equivalent of FFT in the frequency domain.
Up to this time it has been vastly more economical - design solution weight, size, cost, complexity, power consumption, R&D costs, etc., - to use the analogue solution. However, in just the last couple of years we have seen the price and processing power of the digital solution begin to approach the boundary whereby digital application could soon revolutionise metal detector technology.
Take for instance the microprocessor, about a year ago the industry standard was about 250 MHz, over the last couple of months I've been working on a consumer product that uses a 1 GHz processor. This sort of power coupled with specifically applied high speed A to D's, has the potential to raise the faintest of targets way out of the noise, and shout them out loud to us, and when atomic molecular resonance is utilised, 3D images of the target to boot.
This will depend upon the advances being made in UXO and geophysical sensing flowing over into the hobby market.
I'm sorry John, I didn't mean to give you a lecture, I must be half asleep still. I'm tempted now to just hit the back button and not post.
Oh well...here goes
HH
Kev.
No thats ok because thats what I wanted to do is get someone to comment and that is how guys like me learn more, so I thank you Kev for the input. The low pass analog filters in Erics designs are maybe Butterworth, Bessel or others do you think Kev? Yes I would think that we are getting closer to ADC's that can give a much more accurate reproduction of signals and maybe A/D filters will meet Eric's stringent requirements soon. So you can lecture me anytime Kev because I'm all ears. These metal detectors have caused me to want to go further into the black box. Do you know of a good book that I might get on FFT's to read more about the subject? Not that I will understand but it don't hurt to try haha! Ok brother just keep this good stuff coming because I am willing to learn from others like you Kev. God Bless ya and CU later. John the Wirechief.
Carl-NC
Active member
Coherent sampling is where you sample a periodic signal an exact integer number of times per period. So, ferinstance, if you have a 2kHz sine wave, sampling at 50kHz will result in exactly 25 samples per period. If the sine wave is pure, then every 25 samples should contain exactly the same data. Sampling this sine wave at 51kHz produces a non-integer number of samples per signal period, which is incoherent.
When you use a Fourier transform to convert the sampled time-domain data into the frequency-domain, you want to have an integer number of points per signal period, otherwise the frequency information gets smeared out. And, with FFTs, you want to have a power-of-2 samples -- such as 32, 64, or 128 -- to take advantage of mathematical shortcuts. If you are forced to use incoherent sampling, then "windowing" can be applied to the time-domain data. This basically emphasizes the middle part of the data, and squashes the ends where the sampling discontinuities occur. Hanning, Hamming, and Blackman-Harris are popular windowing algorithms.
Metal detectors, both PI and VLF, use a special form of coherent sampling called synchronous sampling. Which is sampling exactly once per period. In PI, this sample occurs during the pulse decay, to see if there is a variation in the decay rate. In VLF, two (or more) synchronous samplers are used to compare the phases of the TX and RX signal to see if there is a phase shift.
There are variations in both PI & VLF but, to my knowledge, all are based on synchronous sampling.
- Carl
When you use a Fourier transform to convert the sampled time-domain data into the frequency-domain, you want to have an integer number of points per signal period, otherwise the frequency information gets smeared out. And, with FFTs, you want to have a power-of-2 samples -- such as 32, 64, or 128 -- to take advantage of mathematical shortcuts. If you are forced to use incoherent sampling, then "windowing" can be applied to the time-domain data. This basically emphasizes the middle part of the data, and squashes the ends where the sampling discontinuities occur. Hanning, Hamming, and Blackman-Harris are popular windowing algorithms.
Metal detectors, both PI and VLF, use a special form of coherent sampling called synchronous sampling. Which is sampling exactly once per period. In PI, this sample occurs during the pulse decay, to see if there is a variation in the decay rate. In VLF, two (or more) synchronous samplers are used to compare the phases of the TX and RX signal to see if there is a phase shift.
There are variations in both PI & VLF but, to my knowledge, all are based on synchronous sampling.
- Carl
Theres the man! Thanks Carl for the in depth description and that is exactly what I wanted to know! I think your at Maxim now Carl and I have been going to their website and getting some good info on filters and sampling. If you can recommend any textbooks on these subjects I would be interested. I knew there had to be a name for this type sampling used in metal detectors. You also have a great website Carl and I'm just thankful for guys like you and Eric, Reg, Kev and others that are willing to take the time to explain to the uninformed like me. I hope that I can at least store up enough of this good stuff to bless others down the road someday. God Bless and take care Carl. 73 de John the Wirechief.
Carl-NC
Active member
No. Sampling is done in the time domain, whether coherent or not. If you have incoherent samples and wish to run the data through a Fourier transform in order to see the frequency domain, then you would use a windowing algorithm on the raw data, before the transform. If the data is coherent, then windowing is not necessary.
- Carl
- Carl
So is the raw data the non-integer numbers Carl? But to see it in the frequency domain the coherent data must go through the Fast Fourier Transform? Thank you for correcting me Carl. What do you mean by smeared out? Ok Carl this is very interesting and I will talk to you again. Wirechief.
Carl-NC
Active member
It's not the raw data that's integer or non-integer... it's the number of samples taken of the signal. The sampled data values can be integer or non-integer, that doesn't matter.
Ferinstance, you take a signal -- say, a sinusoidal voltage -- and apply it to an A/D converter. During a single period of the signal, you sample it a certain number of times*. If the number of samples taken are an exact integer multiple of the signal frequency, it's coherent data.
Once you sample a signal -- coherently or non-coherently -- you have discrete time-domain values. You can then convert that time-domain data to frequency-domain data via a Fourier transform. Non-coherent data will need to be windowed first. The Fourier transform can be a standard algorithm, or a "fast" algorithm, known as an FFT. The FFT requires 2^N data points (normally), the standard FT can work with any number of data points.
If you have some background in signals and circuits, then a good book on data conversion and sampling is Analog Device's "Analog-Digital Conversion Handbook". They also have some seminar material available. Maxim probably has some lit as well, but I've only been there 3 months and don't know what's available.
- Carl
* This is called oversampling. There are ways to also undersample signals, but to try to explain that will severely muddy the waters.
Ferinstance, you take a signal -- say, a sinusoidal voltage -- and apply it to an A/D converter. During a single period of the signal, you sample it a certain number of times*. If the number of samples taken are an exact integer multiple of the signal frequency, it's coherent data.
Once you sample a signal -- coherently or non-coherently -- you have discrete time-domain values. You can then convert that time-domain data to frequency-domain data via a Fourier transform. Non-coherent data will need to be windowed first. The Fourier transform can be a standard algorithm, or a "fast" algorithm, known as an FFT. The FFT requires 2^N data points (normally), the standard FT can work with any number of data points.
If you have some background in signals and circuits, then a good book on data conversion and sampling is Analog Device's "Analog-Digital Conversion Handbook". They also have some seminar material available. Maxim probably has some lit as well, but I've only been there 3 months and don't know what's available.
- Carl
* This is called oversampling. There are ways to also undersample signals, but to try to explain that will severely muddy the waters.
Ok Carl I see now and that clears up the raw data question. I will look for the book you mentioned also for this is really making me more curious. If you undersample don't you have a problem called aliasing? Yes I have a background in electronics but I am just a technician and have worked with instruments used in the radiation field in a practical sense. I'm good at working with my hands and all my experience is practical repair. I have been in 2way radios, stereos, pagers, and other types of repair. For the last 15 years I have worked for a company that designs and builds instruments like Gieger counters, Scalers, Pulsers, Gate Monitors etc.. I have been in production, customer repair, assembly, and QA. Actually I would like to go to school to try and get a EE but it is a little late at 49 years old haha! Thanks again Carl and you have been very patient with me and I appreciate that more than you know. I am sure to come up with more questions, just to warn you haha! CU later Carl and God Bless. John the Wirechief.
Carl-NC
Active member
Aliasing always occurs when sampling, whether over- or under-. Aliasing is where all signal frequencies higher than half the sampling frequency (called the "Nyquist" frequency) fold back into the lower frequencies. For oversampling, a low-pass filter is often used prior to the sampler to prevent these higher frequencies from corrupting the lower ones. Undersampling can be used intentionally to mix down higher frequencies, in which case a band-pass filter is used to select the frequency band you want to mix down.
- Carl
- Carl
Thank you again for the lesson in Aliasing Carl. When the signal frequencies foldback on the lower ones is that sort of like being overdriven, or saturated? Also do you have one of Eric's detectors? I just picked up a Beachscan from the U.K. off of Ebay. It has the Thresh and Reject controls. Is metal detecting your main hobby Carl? Ok I will watch for your next lesson to correct me haha! CU later Carl. John the Wirechief. 73 de KA5QYR
Hey Kev! Thanks for the link and actually I have already looked at this fellows book and am tempted to order it. Actually I have quite a few textbooks and technical publications that has turned into a pretty good library. I don't mind investing in the printed word although after awhile it does stack up. I know the internet is here but I might not always have access to it and so I love having books around so that I can just pick them up and start reading. Have you taken a look at that book Kev and if you have do you think it is worth the price? Many thanks to you Kev for letting me know about this item. I wonder if Prospector Al is ever going to come back? I don't see him posting anything. I would like to visit with him because he has knowledge of the physics domain involved with these detectors. Well you take care Kev and God Bless. John the Wirechief.
Hi John,
I know what you mean about hard copies, it seems much more absorbable, and you can flip back and forth through the pages without needing to think about it, which often on the puter can brake your train of thought.
If you print it out, then it needs to be bound or bits end up all over the place, and there's something comforting about a nice library of your own.
As for it's value, I don't have enough experience with DSP to make that call, but I did find it understandable, where many books get too bogged down in theory, and mathematical proof. Maybe Carl could let us know if it's worth it?
I'm a late bloomer in the electronics field, and have a lot to learn, and I'm 2 years younger than you, and no you're not too old, don't let any young bucks put you off. You can't build nice furniture without the correct tools and knowledge of how to use them, just because you don't have all the tools nor the knowledge doesn't make you a lousy cabinet maker, no you could be a Chippendale who knows? We have one thing on our side, patience, I'm sure it comes with age. I see young ones fall over themselves to get stuck in, but then get bored, before the results begin to show
I admire the spirit you've shown, don't give in, keep collecting those tools and knowledge, and build your own unique detector.
Kev.
I know what you mean about hard copies, it seems much more absorbable, and you can flip back and forth through the pages without needing to think about it, which often on the puter can brake your train of thought.
If you print it out, then it needs to be bound or bits end up all over the place, and there's something comforting about a nice library of your own.
As for it's value, I don't have enough experience with DSP to make that call, but I did find it understandable, where many books get too bogged down in theory, and mathematical proof. Maybe Carl could let us know if it's worth it?
I'm a late bloomer in the electronics field, and have a lot to learn, and I'm 2 years younger than you, and no you're not too old, don't let any young bucks put you off. You can't build nice furniture without the correct tools and knowledge of how to use them, just because you don't have all the tools nor the knowledge doesn't make you a lousy cabinet maker, no you could be a Chippendale who knows? We have one thing on our side, patience, I'm sure it comes with age. I see young ones fall over themselves to get stuck in, but then get bored, before the results begin to show
I admire the spirit you've shown, don't give in, keep collecting those tools and knowledge, and build your own unique detector.
Kev.
Alright Kev I thank you for the kind words and encouragement to go forward. Promise me you will do the same. Well I took 2 and a half years of vocational electronics in high school and built lots of kits and went to Radio Shack everyday when I lived in Tucson. So I have played with tubes or valves, transistors and IC's and girls of course. But you are right patience does seem to come with age. I'm hoping that this is what this will lead to by asking all these questions and maybe I can come up with a new detector design. I agree with Al the next big leap in detectors is going to be in the physics domain. I print out just about everything posted here on the forum. I think Eric's detectors are really neat and different. I hope to try his Goldquest someday. Actually I just picked up a Minelab dealership and If I could I would just as soon carry Eric's stuff. Well I need to shut up so I will CU later Kev and I hope that we both can keep learning because I really have forgotten a lot on this end and when you don't use it you lose it! John the Wirechief. Oh yes maybe Eric or Carl knows something about that book.
Carl-NC
Active member
> When the signal frequencies foldback on the lower ones is that sort of like being overdriven, or saturated?
Folding doesn't occur because of overdriving, but a folded signal can cause signal integrity problems, including saturation. It's really too difficult to explain without some good supporting graphics, so it's best to leave it to published literature.
The ADI book can be had cheaply on Bookfinder (ex-library). I see ADI has published a new version, wish I'd grabbed a copy before resigning. And, I really need to get up-to-speed on the Maxim lit, so I can point folks there!
> Also do you have one of Eric's detectors?
Yup, Beachscan MkII, most sensitive PI I own. Wonderful machine.
- Carl
Folding doesn't occur because of overdriving, but a folded signal can cause signal integrity problems, including saturation. It's really too difficult to explain without some good supporting graphics, so it's best to leave it to published literature.
The ADI book can be had cheaply on Bookfinder (ex-library). I see ADI has published a new version, wish I'd grabbed a copy before resigning. And, I really need to get up-to-speed on the Maxim lit, so I can point folks there!
> Also do you have one of Eric's detectors?
Yup, Beachscan MkII, most sensitive PI I own. Wonderful machine.
- Carl