Modernizing the BFO

[Yourownfree]

Interesting concept. It amazing to me how many different designs there are. That's why I love electronics so much. There is always a way to get to the same result and there is not just one way to do things. If there was, sooner or later someone will come up with something else. I love how people think and the difference of concepts that they come up with. I'm fascinated by the designs. I could spend hours going over schematics. I just do, don't know why. Maybe when you get older half the time is just spacing out. Thank you for sharing this diagram .

[Yourownfree]

uodate: I had to train myself on a program called Easy EDA. Easy my foot. I started watching a two hour video on how to build a simple USB power board and it took me four hours just to get into it for about 10 minutes. So I have many hours on this program and getting better at it. I probably have 12 hours in now. The good news is, I almost have the circuit drawn. I'm in the musical section now. Designing how the output will react and how you will hear it. I'm designing around 700-900 Hz filtering since that seems to be the best frequencies people can hear a change in tone better. Anyhow I hope to have this done in a day or so. After I think I'm done I have to go over each connection to make sure I didn't forget anything. I like to add little bells and whistles just because it adds to the things that can be tested by a user. A few things to adjust so its not just plug and play. I'll explain later. For now its back to work on this crazy detector. I've got 6 hours in today alone on it.

[Yourownfree]

I have the schematic done to share. This is my first design of my upgrades. Sorry it took so long. I had a learning curve trying to learn Easy EDA. schematic builder. It was a challenge for. It took days to get it right.

I can explain how it works in the next post as it is fairly complicated and has to be adjusted properly to work as intended. I had to(not) add some additional treats to it. Keep in mind there could be mistakes. I have never put this together or tested it in anyway. It is a diagram of how I designed it. It is ready for me to put together but I need to double check my work for errors. I'm just saying it may or may not function correctly. Once I build it I will fix any bug issues. I spent a lot of time on this schematic to get it to where I believe it will work. I designed it for 10KHz operating frequency at the coil, 800 Hz filtering in the audio and more. This unit is a BFO unit but without a reference oscillator to beat with. I removed the fixed reference oscillator found in BFO detectors. It simply doesn't use a reference oscillator. I did keep the old style search coil oscillator but changed it to a Hartley Osc. because they are more stable but it has to have a tap in the search coil so you have three wires to the coil instead of the normal two you are used of. More later.
ET_NBFO.pdf (the detector schematic) enjoy.

[Carl-NC]

Technically this is no longer a BFO, it's a relative frequency measurement. Here's another similar option:


Two items with your approach. First, the Hartley is generally less stable than a Colpitts, but probably not enough to worry about. Second, a strong target might create a 1% frequency shift while a weak target might create a 0.01% frequency shift. This is approximately true regardless of the absolute frequency. How much delta-voltage will this produce on the output of the LM2917? Is it enough to hear? This is a big reason why BFOs tended to run at a high frequency and use a mixer to create the audio.

[Yourownfree]

As for the delta voltage it looks like I can improve on that. After calculations it looks like it may be buried in the noise floor but not by much. The .01% So I propose to add a preamp instead of the emitter follower. That should give it better Signal to noise ratio which should lower the noise level and allow it to be detected. but even so, that's asking a lot. Like the wave of a hand. Is it true that humans can really only detect. Ok I need some more improvement. as it is now the delta is calculated on say a medium object buried at 61mv. so a 100 Hz change moves it about 15-20Hz. after looking into the info you gave me here I can improve on this diagram. The improvement should bring up the levels some. Should make it scream. Ill change the diagram with new delta in mind. I understand a person can only hear about a .5% change in frequency. The detector I can improve on knowing these things. Your feedback is greatly helpful. All the above is just theory as I have not tested it in the real world. Once I get the improvements then Ill build it.

[Yourownfree]

I've been looking over the circuit and figuring a few things. I can change the gain to achieve 91mv for 100Hz change. But I need to make a few modifications also to make this even a better detector. I will add a preamp after the osc with drive control on the input. need to scope the output for a defined square wave. may have to tinker with the input cap or whatever it takes to get that result. The output will need to be cut down before entering into the sensitive 2917 and adjusted by probing pin 3 on the 2917. I need to change the Osc. frequency to 5 KHz add a 9volt regulator for the lm2917 so the 470 ohm resistor doesn't have to work hard and give give the 2917 more headroom to scream as it is too close to the edge at moment. It doesn't like anything about 7 volts output. idle voltage was about 6.4 volts, so with a target screaming it would flatline. With new frequency the idle is about 5.14 I think it was which is perfect. I am going to toss the 555 timer chip for my vco frequency to hear but still can use it as a pulse frequency coming from speaker as a choice. I will look at that a bit more closely. but I will add a 4046 chip to give a better response of change with less frequency change. according to calculations at 5KHz 1% will be up there about 500- 1000 Hz. Yeah when this baby goes off you'll know it. May need to change your clothing. You will also have a better chance to hear depth and size and hear the difference. I will also be focusing on internal noise to get that down and .01% if there is not too much internal noise should be about 15-20Hz of change. Its slight but its something. All of this will be adjustable sensitivity and all. I like the more adjustments which helps me do in the field testing to see what works and what doesn't. I probably wont do it but I could ad an identical coil as the search head side by side and hook it up (oh great my brain quit on me, loss for words) to the circuit backward as the main coil so it is opposing it. That way when the sun bakes down on it it changes inductance evenly and helps keep the OSC stable. For now I just want to see how stable it is without it. I am working on a different design then this one that should solve drift anyway. I will work on a new circuit with the new ideas, preamp and cd4046 added and a few parts value changes to increase sensitivity. I want this machine to give a good battle with a store bought machine of the same type. I actually want to improve it to make it better than anything its type. That is the ultimate goal. Later on with some other of my designs we can go with some displays and processing goodies. For now I need to stay focused on this one until it defeats me.

[Skippy]

Some years ago, I was myself wondering what BFO improvements may be possible, using newer techniques, etc. That's what drew me to this thread.

My ideas included using the digital mixer, featured in the ETI-561 Wait design. I was intending to have a low search-coil freq, about 15kHz.
I decided the search oscillator was the one that needed to be designed for temperature stability, and the variable reference oscillator needed to be quartz-stable, but adjustable.

Designing a temp-stable oscillator is something Radio Amateurs have plenty of experience at - years ago, there was no other method. For guidance, you need Ham Radio design books. Obvious examples include the ARRL Handbook, a weighty volume, but available in digital format nowadays. WorldRadioHistory have copies, along with other radio design books.
WRH:
HOME PAGE WorldRadioHistory Radio Music Electronics Publications

BOOKSHELF ARRL HANDBOOKS: Ham Radio technical reference books

The way to achieve a stable, quartz-locked adjustable reference is known as "Huff and Puff" stabilisation. I have a book featuring it ... no doubt it can be found online, I will try and find some links. It 'rounds off' a VCO oscillator to the nearest 'spot' frequency, and locks it using PLL techniques. So, for example, you can choose the spot freq step size to be 25 Hz. If the VCO is running at 85005 Hz, the reference will lock it to 85000 Hz. If you tweak the VCO up to 85019 Hz, the locked output will jump to 85025 Hz. So, provided the drift of the VCO is low, the output will be quartz-stable, in 25 Hz steps.

An online link to Huff & Puff:
Huff & Puff reference library

I hope this info is useful to you.

[Yourownfree]

Well as it turns out, I spent way too much time again on this, this morning here in my time zone. I tossed around ideas the I hit the holy grail. I have to mention though before this idea a minutes before I was working on a concept that apparently has not been investigated so at some point I will do some experiments. Back to the holy grail. I will redo the circuit yet again but this time It will be different. You will say why hasn't anybody done this? Its too simple. As you know by reading I wanted to eliminate the fixed oscillator so we only have one to work with. I would like to write an article on this whole journey to get here. Ok her it is what you were thinking about too I suppose. Here Is how I propose to do it. I will use an oscillator, a Collpits, only because I can square its wave easier and at higher frequencies its more stable. But wait(record skipped) as for an oscillator drifting, get it close so it doesn't drift too much then forget about the drift. It wont matter if it drifts, who cares. why? because you are now going to run the oscillator into a Schmidt trigger to square the osc signal so it looks good. May have to have control over the drive signal which I will rewrite the schematic to include everything a person needs to build this. The next step after you make the wave look good you run it to a cd4046 Pll. Lets look this over for a minute. It will try and mimic the search head oscillator and not send a signal ack to it although you can later if need. but as of right now you don't. If the osc starts to drift so what, the 4046 will track it I didn't say it would change the frequency because we don't need to. All we need is a steady signal. If you use this detector on mineralized soil it will automatically change to compensate for the soil. Theoretically but may need some external control to help. Now you are thinking so what does the signal detection of a target? you design a sample and hold circuit. It keeps track of the steady signal even if it changes slightly due to drift of the osc it doesn't matter. The sample hold circuit will follow the voltage change. Now the secret sauce. You loosen the way it stays on the straight and narrow to maintain a steady signal following the osc. only enough to keep it locked. When a target(coin) is located it will set the unit off because just for a fraction of a second you lose lock on the pll. With that loss you use the error signal and process that into whatever you want. can be another 4046 used as a VCO to get a special sound or you can even use a 555 timer chip in astable mode. The signal is only going to be an on-off signal. yes or no I hit a target or take the additional 4046 and attach the vco output of the first and follow the signal change there. that way you can set the additional 4046 to output any tone you want a a base tone such as the 700-900 Hz premo signal everybody likes but use the first 4046 output to change the vco voltage on the second so that the baseline frequency, the sound you hear will change. of course you can add bells and whistles to this, but this is it folks. At least I believe so now. The holy grail of bfo metal detecting. The details and oscillator frequencies are determined by what you want them to be and you have to design around that. You have to calculate the part values for the 4046 etc to make this all work, but I'm sure it will work. If not it was a great try. Imagine not having to keep tweaking on a bfo to use it.
to the user its a no drift bfo. Ernie

[Yourownfree]

I guess as suggested this is no longer a BFO. I've taken the BFO, took away the reference oscillator that it beats against. It gone. So no BFO anymore. It just uses the search head oscillator, turns into digital, then processes it. Still has the drift problem but you can adjust the threshold at where it goes off when it sees something. The same adjustment can be used as a ground balance . The I posted will be a much better unit than a BFO. You don't have to listen to the beat frequency anymore as the sound is activated when it detects a change in inductance. Because of eddy currents again doing its thing, pushing back at the coil to change its value. The cleaner the signal is from the osc the better. Right now the way it is its sitting in the mud at .01% of 10 KHz and most likely not detectable. but if you clean up the signal, square it up, amplify it and then adjust the output so when it goes into the lm2917 its not getting fouled up. But still the detection was sitting at a point where 10% could be seen but would only give 15-20Hz difference in sound change. Thanks to Carl for pointing that out to me. I never gave it any thought. Then I did think about what he said. Now I know several new things that I can incorporate in the upcoming designs. By the way if you do swap out the 555 for a 4046 and readjust the gain of the2917 you will be able to see the .01% shift at 15-25Hz change and the 10% will go up substantially . I suggest you wait though for the next design I have figured out. I should say two diagrams away. I have some parts I ordered that finally came in. LM2917's,4046's and a few others for me to experiment with. The next unit I did kind of talked about where I am going to do a makeover. It will retain the Oscillator but it will now be a Colpitts Oscillator but get rid of the classic oscillator parts, transistor to be honest needs to go. It produces heat and is not very stable. I've even seen how you can take two legs of a transistor and use the transistor as a temperature sensor. it changes that much. I don't want the oscillator to change too much. Ok so it will be a Colpitts Oscillator because it has a cleaner signal on the output than the Hartley and we want a clean nice signal to start with. It will be an op amp type of Colpitts osc. I have not determined the part number yet but you can be sure it will be a good one not a 741 if you know what I mean. I have to go over the schematic to design from one end to the other so what I say at this moment is the basis on how I see it so far. The next step after the Colpitts is a trim pot. I like lots of trim pots so I can adjust various levels to be what they are supposed to be to keep the signal clean as seen on an oscilloscope. I want to keep the levels I think a Schmitt trigger 74hc14 because it is high quality in what it does. Now just a note, we want to keep the signal as a nice square wave with a 0-5 volt output which will be pefect for going into a 4046 PLL. The Pll in this circuit will follow the Colpitts osc. staying in time with it, even if it drifts slightly. Oh and another thing we need to do is to lower the frequency to 5 KHz instead of 10 KHz. because of certain values in what is needed in the diagram this will be the golden frequency. It will in the end give us the extra headroom we need to process the signal. That was one of the problems I ran up against in the first diagram. with that said we will have enough headroom, the voltage difference(Delta) that is used to process the signal to make this thing scream. Now I will use the LM2917 for the processing as it gives us more options in the end to play with. Now the secret chip is also after the Schmitt trigger, a LM567 tone decoder chip. Now this is going to be a fun chip to use. What it does it lock onto a signal you have designed it for and hah a window in which the lock lives. You can adjust it to 14%, which is how wide it can be tp detect. Imagine using this control to control threshold in a way, sensitivity adjustment or ?. you connect the 567 to the 4046. How it comes into play is it monitors the 4046 and tracks the frequency and with the 567window adjustment you can set it lock onto the 4046 signal. You can even add an LED so you can see that it is locked. When a target is found something is going to change just for an instant when it does it will come unlocked just for an instant and go into error in which it changes state on one of its outputs. This is what you use to activate your processing circuit and it comes alive. A go no go I found treasure sound. You have to calculate and change part values to allow it to not be so tight and quick because it is quick. Quicker than an Arduino I'm told. You have to do some re calculating on the 4046 as well to loosen it up a little so a target can sway the PLL of coarse just enough to allow for the target to come through. I think this could be a fun metal detector. By adding a second 4046 for a VCO(voltage controlled oscillator) this will gets its signal from either the first 4046 or the 2917 which when the output processing signal voltage changes just a little it will take a sound to the next level where you might think an emergency vehicle is coming. You will have the option of "go-no-go" sound or a continuous sound if you want to make you think its a BFO or both. You could run the volume lower on the BFO type of sound and have the "go-no go" mix in and give you two indications. You could have a switch for that or use a cd 4066 for switching. You should have auto ground-tracking due to the first 4046 following the Colpitts. You wont notice any drift. Another thing about the 567 decoder in the circuit is you can readjust the pot on it to lock onto the mineralized frequency of the 4046 after it auto adjusts its ground balance the 567 if you adjust it now it will be looking at the ground balanced signal and it will go active when a target is found. Now you may not know this but there are two signals going on in the circuit at once. One is frequency and the other is voltage. This will get you thinking. Think about discrimination for a moment. If iron gets in the target it is a magnetized kind of metal (Ferrous ) and it will change the inductance of the coil by adding more inductance. The result is now the search coil has more inductance and it causes the oscillator to lower in frequency. when it happens the voltage drops on the output of the 2917. The lm567 will more than likely come unlocked and activate at the same time. Now if it is gold (non ferrous) it will behave differently and have eddy currents involved which will push back on the coil in a way the inductance of the coil changes again. This time the frequency goes up because you have less inductance on the search coil. This causes the voltage on the 2917 to go up which is good. The 567 will lose lock on the 4046 because the frequency shifted. Doesn't matter which way, it's just looking for a shift. Now here is how you discriminate. Take the two signals and work with them. Gold= more voltage and 567 went into error which we use to say it activated for our purpose. Now iron on the other hand= lower voltage and 567 went into error activating our logic. We now can use logic circuit to tell us was it iron or was it gold? Trash or treasure? A different sound can mean treasure or an LED or visual display which we can add. I think with this circuitry we are on a roll. I am designing to get an enormous delta response to work with and the frequency drift timing can be adjusted too with an external knob control. Because all length of wires are important we have to design for voltage control and not a variable resistor (pot) attached to several inches of wire. You have to eliminate that idea. Here is just something to think about and figure out ways you can make this work for you, controls you would like to have etc. I took a quick look at some of the high end models and they mix them call them a hybrid. Like maybe a PI model mixed with a BFO. I think for really the simplicity of this circuit it can do a lot. I don't know what different make and models do as I don't keep up on it. But I did take a glance at a couple of them. They made it look like there unit did more because it had more things to twist on, including it had a volume control. You can look for yourselves and tell me what theirs can do that this one can't. I may be able to incorporate into this circuit theory .I do not know what all people like as functions but I'm willing to see if I can add those. After getting done with BFO modernizing Id like to take a look at the PI detector and see what I can do with that type to improve it. I have an idea how I can incorporate that with this model. just an idea though and for another time.

[Yourownfree]

I think I may have got a little carried away. I say that modernizing a BFO just really moves us forward into another category of metal detectors of some sort. We are now detecting both the frequency domain and the voltage domain. I did take a basic BFO and modified it but today I realized I can take this thing to a totally different level if I keep going. Adding a display, being able to set and do things from the display, GPS monitoring and data logging and on and on. I just need to make sure the above unit works and stop there for now and then I suppose start a new thread about a different kind of Detector. Or maybe explore the PI detector. I think realistically just removing the fixed oscillator and adding the circuitry do the same job but using voltage instead of audio mixing was probably enough to say we modernized the BFO. I took it too far and ended to the point its not called a BFO anymore. I don't know what to call this thing, "Frankenstein " ?. A frequency/Voltage Phase shift metal detector. "F/V Shift". I don't know. But if Carl wants me to continue I can build this unit here or cross over to another thread. or ? I can revise the first diagram to work better and be easy enough to build. Then leave all this new stuff for a different story If there is interest. For myself I would like to build the last info I gave as I have faith in that project. I can then use it in the real world to see how it reacts. I'm just curious. If it works out then I can glorify it more with all the bells and whistles. But if it doesn't work worth beans there no sense in continuing and Id rather put my efforts in a PI machine or VLF. Its all just for fun anyway. Ernie

[Yourownfree]

Skippy I wanted to say thank you for sharing your information. I agree with what you say. I have not had a chance to look it and I will the info you shared. I do understand the concept. I am sure that other metal detector companies have seen it and done it. I myself want to do that but say using a 1KHz crystal Osc as you say it will stay locked and in my case where I want to use a lm567 Its the key. you can move it up or down by whatever your crystal osc is outputting even through a divider if you want to lock every 25 Hz I wonder...... You sharing your info gave me an idea. I will research my idea based on your info so I guess I cant take all the credit. It was the 25 Hz that gave it away. If the idea works I'll have to call the idea circuit the "Skippy circuit".

[Skippy]

What I had in mind ( based on some of the many links on the Huff+Puff page ) was using a watch crystal ( 32768 Hz = 32 x 1024 ), and an oscillator/divider IC from the CD4000 series. This would then give me options for the VCO 'step size' of 16 Hz, 32 Hz, 64 Hz as I saw fit, 32 Hz being the likely choice. The Elektor article from May 1980 seemed a practical Huff+Puff circuit, I've just modified the reference oscillator a bit.

Elektor May 1980 ( see pages 18 on ):
Elektor Magazine
---------------
Another useful Ham Radio book that will have info on oscillator design:
RSGB Radio Communication Handbook:
Radio-Communication-Handbook-RSGB-5th-1977.pdf