Ok, so I want to build a frequency to voltage converter for a signal between 1 and 100 MHz, mostly sinusoidal with some small varying DC component. I’ve been looking and it sounds like I need a frequency discriminator of some sort, but I haven’t found a good, simple way to do it. Seems like there should be an IC or some such simple, cheap solution but I’m drawing a blank.
Any ideas?
Thanks!
100Mhz? are you serious or is that a typo?
what kind of frequency resolution do you need?
what kind of voltage swing can you handle?
Here is a circuit to think about. This uses fairly simple components.I attached a sketch of a charge pump. The inverter symbol was supposed to have a symbol to identify it as a schmidt trigger. That gate needs to be fast enough to pass your highest frequency. Amplify the input signal as appropriate. Use a small C1. Some charge passes into C2 with each cycle of the RF. R1 drains the charge pumped into C2.
This circuit will react to strong RF nearby.
yes, I was serious about 100 MHz. I guess I really only need a 25 MHz bandwidth although 50 MHz would be better.
A little about my application:
I’m building wavelength stabilized lasers, and one way to measure their stability is to shine both lasers on the same photodiode. This works the same as adding any two sources with different frequencies, you get a beat note with the sum and difference frequencies.
I’m working with HeNe lasers at 633 nm, and I believe that I’m stable to the range of 1-5 MHz for short time scales. So, I’ll take one laser at 632.81600000 nm, and one at 632.81600667 nm, combine them on a photodiode, and get a 5 MHz signal out. I want to track the relative stability of the two lasers, so I want to convert his beat frequency to a voltage. A 25 MHz bandwidth gives me enough range for a larger constant wavelength offset between the two - wider bandwidth just gives me more range.
@analogon - thanks for that, it is an interesting idea. I guess I was hoping for a more integrated solution but I’ll keep this in mind.
I’ve seen a reference to using a philips SA568 PLL, and taking the error voltage as the measurement. This part seems not to be available, and I’m currently looking for alternatives. Seems that digital PLLs are more common - but perhaps I could use a comparator (schmidt trigger?).
I don’t know the amplitude of your signal; you may be able to amplify it or offset it on to a dc level to bring it up to digital levels. If you can get up to digital levels you can run it thru a prescaler (divider) and get it down into the frequency range of a PLL with VCO like a 4046 / 4096. Then just use the error voltage that’s supposed to drive the VCO as your frequency error signal. It’s not a 1 chip solution. you have to provide a reference freq to the PLL to compare your freq. under test to and discrete parts for the loop filter etc.
I think that the differences in error voltages may be too small to observe as a voltage due to other circuit noise that usually finds it’s way into measurements and limited resolution of A/D converters. You may need to look at freq counters that have pc connectivity
A few things come to mind.
“You could use a PIC as a frequency counter. It should work up to about 50MHz.”
what kind of PIC are you talking about? All the PICS I’ve used require 4 clock cycles per instruction and the fastest PIC I’ve heard about is the 32FXXX family @ 80 Mhz(or 20 MIPS).
Well the PIC itself cannot run instructions at that rate, but the hardware can. And with the help of the prescaler, you have yourself a fairly decent frequency counter. This has been done many, many times. Take a look at the following app note from Microchip.
[Frequency Counter Using PIC16C5X
Long time ago I fiddled with this just for the heck of it and was pleasantly surprised. With an external prescaler you can of-course go even higher (but would then have to “switch” ranges externally).](http://ww1.microchip.com/downloads/en/AppNotes/00592d.pdf)
If you want accurate measurements, I would recommend:
a) Use a PLL if you need noise rejection; the PLL will essentially bandwidth-limit the frequency excursions you can detect.
b) measure frequency or period digitally.
You get more accuracy for less money in the time domain, and the data you’re trying to collect is time-domain. You can roll your own frequency counter (or period timer), or if the assignment is about the lasers and not the test equipment, the lab might have either a stand-alone frequency counter or a nice cheap digital oscilloscope that’s been calibrated recently.
"…Take a look at the following app note from Microchip.
Frequency Counter Using PIC16C5X…"
interesting; thanks for the link.
some of the 18F’s can probably come close to 100Mhz; and output result to a PC via USB for embedding into a automated test .
I had been thinking about measuring the period with the edge detectors then subtracting the code overhead time. While I think it’s more accurate, I don’t think you can exceed the MIPS. Haven’t looked into it though.
I think the PLL route is the way I want to go - I’m interested in the spectrum of the beat note between the two lasers.
The original paper I saw used a Philips SA568, but this chip seems to not be available. I’m looking at the 74HC4046. If there’s any suggestions for a better PLL I’m all ears, otherwise I’ll just get one and go play with it a bit and see if it works.
Thanks all!
If you are interested in the spectrum of the sideband signal from mixing the lasers, your best bet is to simply digitize the photodiode output with a digital oscilloscope and fourier transform that data. If you are interested in only lower frequency components modulated on that signal, then the PLL is an excellent filter, and you can digitize the PLL oscillator output with an oscilloscope. In either case, a calibrated oscilloscope is your best data capture tool (if you have access to one, of course).
Right - sample fast enough to get into the frequency domain. Then all sorts of analytical possibilities emerge that aren’t possible in the time domain.
I would use a PLL like this. Should give you a -5 to +20 Mhz bandwidth. In this system you are basically locking the laser to produce a 5 MHz beat note. The reference is the 5 MHz crystal oscillator.
If you really need a +/- 50 MHz bandwidth, you can do it too (with a different circuit)–you would need to upconvert the beat note out of the photodiode to 50 Mhz, and use a 50 MHz crystal oscillator. You would need to convert the 5 Mhz beat note to 50 MHz using an image reject mixer and another local oscillator (a second crystal oscillator) at 45 Mhz.
In any case, you want to make a system that locks up producing a 5 Mhz beat note, but not being able to lock up with a -5 Mhz beat note!
Vout would be ~proportional to the instantaneous deviation from the desired 5 MHz beat note as long as the loop stayed locked.
There are probably other phase frequency detector chips out there too. There is a 50/50 chance you would have to swap the “U” and “D” pins to get it to lock. And you need to choose the R and C values to make a control loop with some phase margin.