[time-nuts] yet another GPSDO design, or so
lists at rtty.us
Tue Jun 29 21:35:54 EDT 2010
Ok, here's a BOM:
TI DAC8581 $1.85 each, two for 3.70
Linear LTC 2493 $2.95
TI LM4040C50 $0.36
Freescale MCF52254AF80 $4.38
Quad Op-amp $1.00
Misc resistors and caps $3.00
Other semi's $2.00
Total $17.39 at moderate volume prices. Depending on your shopping approach likely less than $20 how ever you do it.
That gives you a CPU that's massive overkill (USB, Ethernet, 512K flash, 64K ram ...), a 4 channel ADC, a not so great reference, and dual 16 bit DAC's with pretty good performance. Spend a weekend writing code. Lay out a board the next weekend. fab it up and let it self train. Once you are done, you have a ~24 bit DAC with a 1 ppm or so INL. The linearity of your EFC is nowhere near 1 ppm, so INL isn't the measure you need to worry a lot about.
On Jun 29, 2010, at 9:01 PM, jimlux wrote:
> Hal Murray wrote:
>>> or with a pair of current output DACs and a resistive divider/summer so you
>>> have a "high order" and "low order" voltage.
>> If it were that simple, the manufacturers would package it up into a single chip. :)
> And they do... hence delta sigma designs..
> Back in the good old days before big monolithic converters were available you cold buy a fast wide DAC that basically was a hybrid with 2 smaller DACs and a prom that was burned at the factory.
>> I think there are two areas of interest. One is the obvious one that steps on the high-order DAC won't cleanly map into a constant number of steps in the low-order DAC.
> Yep.. but if you're driving it from a CPU, memory is cheap...
>> The other is things like temperature shifts. You have to work out the specs for both paths and take the worst one.
> It certainly isn't easy..
> But, if you need something that isn't readily available off the shelf (for one reason or another.. maybe you've got several thousand 8 bit DACs in your garage that you're dying to use... along with a well regulated power supply to run them all <grin>)
> time-nuts mailing list -- time-nuts at febo.com
> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.
More information about the time-nuts