Introduction

[Carl-NC]

HH1 and its direct spin-off HH2 were both projects that were pretty much fully designed before I posted them. HH3 is different. I have been toying around with several approaches, and never completely pursued any of them.

What I would like to do with HH3 is develop it on the forum, with (hopefully) plenty of participation from other forum denizens. We can take it one piece at a time and incorporate the best of previous designs.

I think the best approach is to put as much into software as possible so that the design is very flexible in its processing ability. I'd also like to have some sort of LCD interface. And, of course, USB.

- Carl

[Ray-NM]

Sounds like a great idea Carl.
I thought about using a basic stamp or a propeller chip from Parallax.
From my testing it looked like the basic stamp was not quite fast enough
Havent tried a Propeller chip yet But it looked quite versitile but used a differnet programing language.
Anyway just my 2 cents.
Ray-NM

[gef12]

Ground balance HHIII

Hi Carl
One thing I and some others would like to see on the HH III would be a ground balance .. but I guess you may have thought of that already as previous HH have not had this .. although I have seen some attemts at circuit design into this area.
cheers
gef

[Carl-NC]

Yes, HH3 should have a means for GB. I'm still stuck trying to get HH2 finished so 3 will be further down the road. Too many projects...

[mts]

Do you see HH3 being able to remove the need for the differential integrator? It would be great if the microcontroller could directly sample the output of the preamp and then perform all integration and processing internally in the digital realm.

I'm currently working on such a detector based in part on the HHII and GoldPic3 designs. So far so good but my depth is not as good as I think it should be. I'm getting about 7" on a quarter using brute force techniques. So now I need to go back and refine my design. But I think there is a lot of potential here and the circuit is significantly less complex.

As you've said before the 10bit ADC in the PIC is actually not that bad and seems to give decent results. I'm only using 8bit mode and I'm still getting what appears to me to be a workable system (albiet, with less than optimal depth).

I'm using the PIC16F690 which comes with the PICKit2 programmer and demo board. It's a 20 pin DIP so it has plenty of I/O lines and ADC inputs for things like threshold, delay, and other inputs as well as output lines for LCD's and such. There are certainly much better microcontrollers out there but this one is cheap and seems to get the job done. Plus it is through hole which makes it easy to breadboard the solution first. Eventually I will move toward one of the dsPIC versions for sampling multiple times in an attempt to provide some level of discrimination. But for now the goal is to produce a decent PI detector with minimal pin count and cost.

[Carl-NC]

I'm not sure that a micro ADC has enough resolution for directly sampling the preamp. But I think that is the right approach. I am doing some exploratory work with direct sampling using an 18-bit ADC, and I'd like to have more than that. I'm sampling at 2us intervals all the way through the on & off times, and getting some very nice results. This is an official White's project so I can't share details, but I will try to form a hobby project based on similar concepts.

- Carl

[mts]

Well I can understand the desire to sample multiple times during a given interval. That would be ideal but would require a pretty high sampling rate. Your old HH2 design was really only sampling twice during each interval and I expect that the samples were far enough apart that they could be implemented using a fairly decent PIC. I don't remember the exact numbers but I think the PIC I'm using takes about 15usec to finish a sample. There is a minimum setup time as well so you might be talking about 30usec between samples. That's not very fast but is probably close to what your analog system was doing using two sample and hold circuits before the differential integrator.

As for the number of bits I can defnitely see where more is better. Getting 12 or 16 bit samples is usually going to be better than 8 or 10. Unfortunately, I'm finding that 10 bits isn't giving me any better depth than 8 bits due to noise. So that's going to be the biggest issue that anyone is going to need to contend with. I'm not sure what my problem is yet. Perhaps the sample and hold circuitry in the PIC is not up to the challenge and the signal is changing too quickly so there is a cap on precision. I don't know. Using a faster external ADC might solve the problem. I'll keep investigating it. I'll also look into using a higher end PIC. Unfortunately, the dsPIC flavors tend to only come in surface mount versions which makes breadboarding much harder.

I'm glad to hear that you are already heading in this direction. I really see this as being the way to go. The circuits are significantly simpler and you can do all kinds of things with digital filters that are very hard to do with analog circuits.

I'll be watching closely as I'm very interested in hearing what your results are.

Thanks!

[solomon]

a junior member

Hi,I am a junior member of your site i see your project i am so interested in oder to give a comment send to a sample of the design,block diagram& schematic diagram of the matchless metal detector.

[uvgunkhuu]

high performance simple circuit pi metal detector

hi dear membershipsg
hammerhead pi 1 problems
i needed
that working video link
please help?

[Qiaozhi]

hi dear membershipsg
hammerhead pi 1 problems
i needed
that working video link
please help?

If your problem is with Hammerhead I, then why have you posted your question in the introduction thread to Hammerhead III?

[GT Blocker]

Well, it has been awhile now, hasn't it?

Hey folks,

It seems to have become pretty quiet here on the subject of the HH3 and PIC uControllers. There are many new PICs every year with ever better capabilities. So I have decided to reach out to the PIC MicroChip forum to see if I can get a bit of insight. I'll see how that works out and let you know.

Questions that I still have are mostly in the ground balance arena. How many integrators are needed for reliable ground balance? Can a PIC handle the integration AND timings, or do we still need to go with an analog/digital solution to handle both? Do we need dual PICs to handle both signal processing and LCD display information? Do we need a bit of flash memory (probably)? Do we need a stand alone A : D converter to achieve the needed resolution and speeds?

It just seems to me that we need to push ahead. Lets start pounding this out to a board so we can get some code finished.

Just my thoughts on this...
GTB

[Dean Sarelius]

Hi All,
In seeking to move this thread along and hopefully promote some more discussion around the HM3 design I would like to make a comparison of the merits and pitfalls of a pure analog + switches differential integrator compared to a high resolution 16,18,24bit high speed ADC approach.
Firstly the analog approach as I see it provides the benefit of a low cost solution for performing the differential sampling of S1 and S2 as these are typically spaced within 15-20us the sampling of these values can be performed reliably and easily given the analog switches + opamp approach.
Performing this task with an ADC would relegate this to a high MSPS ADC and most high resolution high speed ADC's are very costly. I wonder if the cost of such an ADC would be considered worthwhile, I certainly think it would if the benefits were great enough.!

In order to process the 16bit ADC output as a minimum will probably require a 16bit MCU, no probelm since there are a plethora of good 16bit MCU's available from PIC and ATMEL and these are now relatively low in cost and coding in C or assembler in 16bit is very similar to programming in 8bit just a larger word.

So assuming the following, S1=Target ID, S2=GB, S3=EF1 and S4=EF2 and the relationships are typically FET OFF<25us>S1<10us>S2>, S1<50uS>S3 and S2<50us>S4 then I would like to consider how the ADC can add value over and above the normal analog switch options.

As you know within the first few microseconds of the return pulse the voltage change per microsecond is still very high while after say 20-25us most of the pulse has started to settle so only small variations of the voltage occur over a given time. The analog switch plus op amp approach is not very capable of sampling the voltage variations within the steep falling edge of the return pulse since the sampling mechanism itself requires a minimum amount of time to sample and since the voltage is decaying at such a high rate the average sampled value is highly dependent on the sample and hold input impedence charactersitics of the circuit.

However if an ADC were fast enough it may be possible to sample this edge accurately and it is my guess from my own experimentation so far that the sampled value on this part of the slope could be beneficial for target identification as opposed to the sampling of the pulse after or during the settling phase.

Not withstanding the rapid number of ADC samples over the pulse which provides numerous ways of manipulating this data in software I wonder if there are any other advantages anyone can add to this..?

[Silver Dollar]

Well for VLF 24 bits is good and a fast processor can handle the generation and collection of data as well
as output a nice audio tone. This is my starting point for that;

http://www.geotech1.com/forums/showt...dec-Bare-PCB-s

But a PI is a different animal all together. In studying past projects I see a push towards more bits
and faster sampling but you never have enough horsepower to do the job. Then a shift in thinking
comes and they switch to a comparitor type setup where you watch to see where the decay falls
below a certain value as this will tell you what type of metal is under the coil.

You have to find the "pivot point" and then can tell which side a particular target is on. This type
of PI detector will work nicely with some sort of Micro or the analog input with the micro decoding
the target and providing audio. A full waveform sampling unit will be difficult to get the performance
from (I think) as the speed of the sampling is a bit excessive.

[Dean Sarelius]

I'm not sure that a micro ADC has enough resolution for directly sampling the preamp. But I think that is the right approach. I am doing some exploratory work with direct sampling using an 18-bit ADC, and I'd like to have more than that. I'm sampling at 2us intervals all the way through the on & off times, and getting some very nice results. This is an official White's project so I can't share details, but I will try to form a hobby project based on similar concepts.

- Carl

Hi Carl,
About selecting a suitable ADC, it would appear to me that the Delta Pulse 3 circuit Pasha has developed is using the built in 10bit 100ksps A-D converter inside the PIC18F4520.
Pasha is even displaying the sampled back emf pulse on the 128x64 display and according to his videos he is able to discriminate and cancel out iron and aluminium leaving just the signal for copper.
The Microchip data on the ADC (p22 shows that it is capable to perform 10bit AD conversion within 2.4us which is plenty fast enough as far as I can see..!

So my question is do you really think that it is necessary to go higher than 10bits given the good performance which Pasha is already demonstrating..?

By the way the reason I am asking this is because I would also like to directly sample the pulse after the pre-amp but would prefer to see how far the performance can be made with the 10bit ADC before investing in higher speed higher accuracy ADC.

DP3 Metal Discrimination:
http://www.youtube.com/watch?v=RSJ1H_bG-sA
DP3 schematic
http://www.geotech1.com/forums/showt...&highlight=DP3
PIC Data sheet
http://www.microchip.com/wwwproducts...#documentation

Cheers,
Dean

[Qiaozhi]

About selecting a suitable ADC, it would appear to me that the Delta Pulse 3 circuit Pasha has developed is using the built in 10bit 100ksps A-D converter inside the PIC18F4520.

With the 10-bit ADC of the PIC18F4520, the resolution of the sampled signal is 5/1024 = 4.88mV. If the preamp has a gain of 60dB (1000x) then the smallest change in the target signal that can be detected will be 4.88m/1000 = 4.88uV. So, if around 5uV will provide enough sensitivity for the application, then a 10-bit ADC will suffice. Of course you could use oversampling and decimation to increase the ADC's resolution, but this would cause the detector to react slower, and is probably not be a practical solution.