Since this made it to HN an Hackaday, I'd like to address several quips from folks who I think just skimmed the beginning of the article.
1) Supply voltage: if you have a circuit that uses higher voltages, the picks mentioned at the beginning obviously don't work! I am recommending these ICs because 95%+ of contemporary designs are single-supply 3.3 V or 5 V circuits. If you need more range, there's plenty of modern chips that fit the bill and work better than LM741 & co. For example, OPA1615 is great.
2) Cost: the ICs mentioned at the beginning are inexpensive, but they're not the cheapest op-amps out there. They're just balanced picks with good specs. If you need the lowest possible price, you're still better served by a 21st century design. This may include redesigned, drop-in replacements for the vintage chips - e.g., LMV324A.
3) If it works, it works: sure. I'm not saying you have to switch if you have a drawer full of LM324 and it happens to work for you. But in practice, the internet is full of posts from people trying to make sense of crossover distortion, phase reversal, or signal voltage limitations of these old chips. They're just not very good op-amps, and novices should probably stay away.
Cool article! I'm actually looking at dusting off my op-amp skills, and this has made me think. I am thinking about trying my hand at a phono preamp, but now I have the problem that the input and output signals are symmetrical around a "ground" level, and thus an op-amp using a dual supply +/-5V or so would be simple (output voltage swing I aim for is +/-1V).
So now I have to add a voltage reference to my circuit. I can basically use a voltage divider, maybe a zehner diode, or probably most appropriately an actual voltage reference (TL 43something?) . Thoughts?
So tradeoff is: dual supply vs. additional chip. Basically it depends on what kind of power supply I can get without building my own (which is not too difficult, especially for low power applications, but adds cost and build time). Using a typical 5V wall wart will likely give pretty shitty results due to switching noise, I guess. I'll dig through my pile of stuff and see what I can find. (and I need to get to grips with SMD, even if it's those big packages, my eyesight isn't what it used to be)
In the paras below aren't the datasheet parameter back to front?
Such devices are known as rail-to-rail output (RRO). If in doubt, look at the “common-mode input range” or a similarly-sounding parameter in the datasheet.
In the same vein, quite a few devices accept input voltages that extend a bit below the lower rail; some can also go slightly above the upper supply voltage. This reduces the need for awkward biasing and is denoted as “output swing” in the datasheet.
I have recently tried to make sense and understand the construction of the Colpitts oscillator, and how it works. Between half a dozen different articles and example circuits (most using transistor, not op-amp, though one article mentioned using op-amp instead), there were quite a few different constructions in regards to where the two capacitors in series were connected.
The goal was to understand how to make a metal detector...
The all analog oscillating circuits have to fulfill the Barkhausen criteria, that says an amplifier circuit with a feedback loop may become generator only in case, when the output signal is in-phase (not shifted or shifted multiplications of 360 degrees) with the input one.
So, in other words, read all those caps as a feedback loop phase shifter.
Now imagine a sine wave being applied to a base of transistor.
Since the rise of a base potential decreases the collector potential, there is a 180 degree (or Π) shift and the criteria is not fulfilled.
In contrary, the rise of a base potential results in rise of an emitter potential, the phase shift criteria is fulfilled. And there is a feedback loop (the capacitors) attached.
I'm hoping the above explanation will help You reading the generators circuits.
The design challenge with metal detection is that a modern constructions are a Digital Signal Processing platforms that analyze the spectrum for a different wavelengths (and perhaps antennas as well) responses of objects made out of different materials. I.e. gold & silver are diamagnetic and they will respond in other way than ferrous materials (ferrous are attracted by magnets, while gold and silver opposite). Filtering and processing the signal to discriminate materials is a real challenge here.
Designing a quality and precise analog circuit is truly challenging. A few months ago, around September, I've been asked if I could make a controller for a fertilizer mixer. At a first glance - nothing fancy, just a pair of PID regulators for EC / pH each.
From the perspective of this project I can tell, that designing the right analog measuring circuits is a state of art. It is not only limited to selecting the right op-amps (most of the important factors You have nicely described above), but also many different aspects of the analog design.
As an example, the EC/pH probes are extremely high impedance devices, so using JFET op-amp is a must. But that's not all. I've learned the hard way, that EC/pH measuring circuits cannot be supplied with power from a same source. The electrodes of the probes not only have their high impedance [GigaOhms], but also own potential, and if they are put together into a conducting liquid, the currents start flowing in between probes rendering circuit readings unreliable.
Then, my second idea was to use a micro DC/DC inverters like TBA 1-0512E from Traco. It didn't work out either. The readings were unstable as before... despite measuring circuits were separated! Then, by using DSO I found a nice DC freewheeling diode ripple noise on the power line. By reading carefouly datasheets, I've learned Tracos have 100[mV p-p] ripple noise, where - i.e. the pH probe generates in between of -177..177[mV]. So yet another fail.
And I have mentioned, the circuits have to be separated. Then how to separate the analog circuits from the digital one? Either by converting Voltage-to-Frequency or by ADC converters that use isolated power lines, like AD7793.
Making long story short, after several months of hard work I am getting 0,7% of accuracy from the designs I have, and I believe I could do better in several aspects. Mainly that's the reason quality, industrial-grade analog products are pricey - and perhaps, in this very case I should go for the EC/pH transmitter device instead of reinventing the wheel.
Some could say that my words are sort of abstraction, because there are EC/pH meters on the market for a dozen of bucks to buy. Not gonna start my rant on them, let's say buying a six-pack would be a better use.
Thanks. As an electronics hobbyist beginner, I’m stuck for a while at the copy/paste stage if I want to get anything done. I do make an effort to understand the circuits and the data sheets of the components included in what I’m copying, but I’m a long way from being able to select my own components.
I look forward to your articles — many of them are slightly above my head, but you write clearly and I do manage eventually to comprehend them.
Getting the level right for people like me is something of an art (I teach in my own field, so have 1st hand experience).
If you have questions about specific designs, or are just scratching your head about one of the existing articles, don't hesitate to reach out (lcamtuf@coredump.cx). I can probably help, and it's a good way to improve the articles or discover new topics to write about =)
Since this made it to HN an Hackaday, I'd like to address several quips from folks who I think just skimmed the beginning of the article.
1) Supply voltage: if you have a circuit that uses higher voltages, the picks mentioned at the beginning obviously don't work! I am recommending these ICs because 95%+ of contemporary designs are single-supply 3.3 V or 5 V circuits. If you need more range, there's plenty of modern chips that fit the bill and work better than LM741 & co. For example, OPA1615 is great.
2) Cost: the ICs mentioned at the beginning are inexpensive, but they're not the cheapest op-amps out there. They're just balanced picks with good specs. If you need the lowest possible price, you're still better served by a 21st century design. This may include redesigned, drop-in replacements for the vintage chips - e.g., LMV324A.
3) If it works, it works: sure. I'm not saying you have to switch if you have a drawer full of LM324 and it happens to work for you. But in practice, the internet is full of posts from people trying to make sense of crossover distortion, phase reversal, or signal voltage limitations of these old chips. They're just not very good op-amps, and novices should probably stay away.
Cool article! I'm actually looking at dusting off my op-amp skills, and this has made me think. I am thinking about trying my hand at a phono preamp, but now I have the problem that the input and output signals are symmetrical around a "ground" level, and thus an op-amp using a dual supply +/-5V or so would be simple (output voltage swing I aim for is +/-1V).
So now I have to add a voltage reference to my circuit. I can basically use a voltage divider, maybe a zehner diode, or probably most appropriately an actual voltage reference (TL 43something?) . Thoughts?
So tradeoff is: dual supply vs. additional chip. Basically it depends on what kind of power supply I can get without building my own (which is not too difficult, especially for low power applications, but adds cost and build time). Using a typical 5V wall wart will likely give pretty shitty results due to switching noise, I guess. I'll dig through my pile of stuff and see what I can find. (and I need to get to grips with SMD, even if it's those big packages, my eyesight isn't what it used to be)
In the paras below aren't the datasheet parameter back to front?
Such devices are known as rail-to-rail output (RRO). If in doubt, look at the “common-mode input range” or a similarly-sounding parameter in the datasheet.
In the same vein, quite a few devices accept input voltages that extend a bit below the lower rail; some can also go slightly above the upper supply voltage. This reduces the need for awkward biasing and is denoted as “output swing” in the datasheet.
Yes, thanks!
"To get the RMS noise voltage, you can multiply the figure by the square of the bandwidth the circuit is passing through."
I think you meant square-root of the circuit bandwidth here.
Oops, yes!
I have recently tried to make sense and understand the construction of the Colpitts oscillator, and how it works. Between half a dozen different articles and example circuits (most using transistor, not op-amp, though one article mentioned using op-amp instead), there were quite a few different constructions in regards to where the two capacitors in series were connected.
The goal was to understand how to make a metal detector...
The all analog oscillating circuits have to fulfill the Barkhausen criteria, that says an amplifier circuit with a feedback loop may become generator only in case, when the output signal is in-phase (not shifted or shifted multiplications of 360 degrees) with the input one.
So, in other words, read all those caps as a feedback loop phase shifter.
Now imagine a sine wave being applied to a base of transistor.
Since the rise of a base potential decreases the collector potential, there is a 180 degree (or Π) shift and the criteria is not fulfilled.
In contrary, the rise of a base potential results in rise of an emitter potential, the phase shift criteria is fulfilled. And there is a feedback loop (the capacitors) attached.
I'm hoping the above explanation will help You reading the generators circuits.
The design challenge with metal detection is that a modern constructions are a Digital Signal Processing platforms that analyze the spectrum for a different wavelengths (and perhaps antennas as well) responses of objects made out of different materials. I.e. gold & silver are diamagnetic and they will respond in other way than ferrous materials (ferrous are attracted by magnets, while gold and silver opposite). Filtering and processing the signal to discriminate materials is a real challenge here.
Designing a quality and precise analog circuit is truly challenging. A few months ago, around September, I've been asked if I could make a controller for a fertilizer mixer. At a first glance - nothing fancy, just a pair of PID regulators for EC / pH each.
From the perspective of this project I can tell, that designing the right analog measuring circuits is a state of art. It is not only limited to selecting the right op-amps (most of the important factors You have nicely described above), but also many different aspects of the analog design.
As an example, the EC/pH probes are extremely high impedance devices, so using JFET op-amp is a must. But that's not all. I've learned the hard way, that EC/pH measuring circuits cannot be supplied with power from a same source. The electrodes of the probes not only have their high impedance [GigaOhms], but also own potential, and if they are put together into a conducting liquid, the currents start flowing in between probes rendering circuit readings unreliable.
Then, my second idea was to use a micro DC/DC inverters like TBA 1-0512E from Traco. It didn't work out either. The readings were unstable as before... despite measuring circuits were separated! Then, by using DSO I found a nice DC freewheeling diode ripple noise on the power line. By reading carefouly datasheets, I've learned Tracos have 100[mV p-p] ripple noise, where - i.e. the pH probe generates in between of -177..177[mV]. So yet another fail.
And I have mentioned, the circuits have to be separated. Then how to separate the analog circuits from the digital one? Either by converting Voltage-to-Frequency or by ADC converters that use isolated power lines, like AD7793.
Making long story short, after several months of hard work I am getting 0,7% of accuracy from the designs I have, and I believe I could do better in several aspects. Mainly that's the reason quality, industrial-grade analog products are pricey - and perhaps, in this very case I should go for the EC/pH transmitter device instead of reinventing the wheel.
Some could say that my words are sort of abstraction, because there are EC/pH meters on the market for a dozen of bucks to buy. Not gonna start my rant on them, let's say buying a six-pack would be a better use.
Thanks. As an electronics hobbyist beginner, I’m stuck for a while at the copy/paste stage if I want to get anything done. I do make an effort to understand the circuits and the data sheets of the components included in what I’m copying, but I’m a long way from being able to select my own components.
I look forward to your articles — many of them are slightly above my head, but you write clearly and I do manage eventually to comprehend them.
Getting the level right for people like me is something of an art (I teach in my own field, so have 1st hand experience).
So, please keep up the good work!
If you have questions about specific designs, or are just scratching your head about one of the existing articles, don't hesitate to reach out (lcamtuf@coredump.cx). I can probably help, and it's a good way to improve the articles or discover new topics to write about =)