High Resolution Programmable Divider

I have an application where I need to generate frequencies from 500,000hz to 640,000hz in one hz steps. I'm looking for the simplest way to accomplish this. I'd like to do it with a uP, but I think the resolution required is beyond the counter/timer circuits of any uP that I'm familiar with. I suspect it'll have to be a hardware logic solution. Of course, there may be a nifty part that I don't know about that will do something like this.

The frequency could obviously be higher within reason, as long as I get the required resolution over the same 20% range.

Buy one of these.
https://www.thinksrs.com/products/CG635.htm
and control it through the serial port.

If you want to build hardware, the keyword to be looking for is "DDS" Direct Digital Synthesis:

Analog Devices (www.analog.com) has a pile of chips:
**broken link removed**

And you should be able to google for some more related links

Since this is going into a piece of equipment, I doubt the boat anchor is the answer. I'll check the other links...

Actually, I want something that generates digital signals, I'm not looking for analog outputs. I just need to generate an extremely accurate digital frequency.

Most of the DDS chips do have a digital output. The thing is that the analog might be more useful - put it through a smoothing filter to get rid of some noise, then to a high speed comparator.

Keep in mind that the difference between 639.999KHz and 640KHz is 2.5 picoseconds. There is going to have to be some analog component to this solution unless you're willing to modify the specs.

Good point, but since I need the high accuracy, I'm not sure if that will do the trick. I may be embarking on the impossible mission. I'm trying to emulate a high quality pressure transducer that puts out 4300hz to 4800hz full scale range, so my idea was to divide that frequency down by 128 after generating it. The trick, as you say, is actually generating it.

If you filter the sine wave coming out of the DDS, it *is* accurate, just noisy. Once it is filtered, all the oddball harmonics disappear and you have your nice clean signal with (possibly) 2.5 psec resolution.

The proper question probably should've been to state that you need some sort of waveform in the 4.3-4.8KHz range (sine wave? triangle? square?), or just tell us the sensor type so we can look it up and get some more background info. Like whether is this part of some calibration procedure - and there may be some relatively simple ways of transforming the problem into something easier.

As it is, most of the DDS chips seem to have 32 bits of resolution, and should be able to hit your initial requirement (but I've never used one, so I can't say for certain). I've only used DDS techniques in software...

James

gunrunnerjohn said:

Good point, but since I need the high accuracy, I'm not sure if that will do the trick. I may be embarking on the impossible mission. I'm trying to emulate a high quality pressure transducer that puts out 4300hz to 4800hz full scale range, so my idea was to divide that frequency down by 128 after generating it. The trick, as you say, is actually generating it.

Click to expand...

Why not just generate the 4300-4800Hz directly?.

You also don't give any specs on how accurate and stable it needs to be?, presumably similar to the original sensor - but what is the spec on that?. There seems little point getting carried away striving for 'super accuracy' to emulate a sensor that is only 5% accurate or something?.

By the way, assuming 1Hz resolution (at 4000Hz), you're talking 500 steps, that's 0.2% resolution - which I would imagine probably exceeds the resolution and accuracy of the original sensor?.

1Hz resolution at 500KHz, would give a ludicrous 0.0016% resolution!.

Actually, the sensor is way more accurate than 0.2%, amazing as that seems. It represents altitude from -1000 through 57000 feet, and it's absolute accuracy over time is +/-20 feet, and the resolution is much less than a foot. You can take a unit with the sensor running, and move it from a table top to the floor, and measure the height of the table from the pressure change. A standard air data quality pressure transducer has <0.05% Full-Scale Including Temperature Effects Over -40 to +85° C. That's long term accuracy, it's much better short term, and obviously the resolution is very good.

I have spent more time with the DDS solution, and that seems to be the way to go. I can use a TCXO as the input clock, and then divide it down to the correct range. Analog Devices has a number of them that look like a good fit.

1Hz resolution at 500KHz, would give a ludicrous 0.0016% resolution!

Click to expand...

Remember, one person's ludicrous is another person's requirement. I agree it doesn't require 0.0016%, but we want the conversion to be more accurate than the input data accuracy for obvious reasons, since errors here add to the total system error budget.

BTW, I'm not processing the sensor, I'm trying to emulate the sensor output for some external equipment.

You really need a frequency synthesiser do this and you can buy single chip PLLs that'll do this.

https://en.wikipedia.org/wiki/Frequency_synthesiser
https://www.google.com/search?clien...+synthesizer&sourceid=opera&ie=utf-8&oe=utf-8

Analog Devices doesn't seem to think that's the case.

New integrated Complete-DDS products present an attractive alternative to analog PLLs for agile frequency synthesis applications. Direct digital synthesis (DDS) has long been recognized as a superior technology for generating highly accurate, and frequency-agile (rapidly changeable frequency over a wide range), low-distortion output waveforms. DDS architecture (Figure 1) employs a precision phase accumulator and digital signal-processing techniques to generate a digital sine wave representation which is referenced to a highly-stable reference clock. The digital sine-wave data is then applied to a high-speed D/A converter (DAC) to generate a corresponding analog sinewave output signal.

Click to expand...

Do you know of any single chip PLL's that would do the job as well as the **broken link removed**? FWIW, I am going to use the digital output, I don't actually need the output DAC, it comes along for the ride.

I don't think standard PLL's would have enough resolution. Some of the newer "Fractional N" based ones might be able to though.

Well, if size/budget allowed, I'd probably want to spend the $$$ on a piece of test equipment that I know would definitely meet the specs, then use that to verify some cost targeted version (if needed). Incidentally, the frequency generator I linked to is a DDS generator with a PLL wrapped around it .. and a couple other bits of stuff. They have a pretty detailed manual on their site (the paper version of the manual has complete schematics, which surprised me).

The Analog Devices link you provided started me down the road, and I thank you for that. So far, it seems to be my most promising option. It appears to have the resolution, and given an accurate and stable clock, should be sufficient for the task at hand.

Truthfully, I only need about 30,000 steps through the range, I was "guilding the lily" with the initial requirement. I like to see how good it can be done, then crank back to the bare requirements if necessary. With the DDS solution, it appears I can probably get 60,000 steps, which would be great.

I didn't know about the DDS option when I started, so I was thinking along the lines of a programmable divider. I quickly came to the realization that would be difficult or impossible.

As far as test equipment, we plan on renting some equipment to prove the concept once we have the idea firmly in place. I don't think I want to spend to money to own it.