wb4jfi at knology.net
wb4jfi at knology.net
Sat Jan 5 18:02:09 CST 2013
Hello there Karl. I did not see the LP filter design that you mentioned, but I believe the type of filter is an elliptical, or Cauer filter. You can use the free filter design and analysis program from aade.com to design them. I’ve used them for several years with DDS designs. The AD9850 is pretty old technology these days, but the price is pretty good. It’s output DAC’s resolution is not great, and it’s a power hog. I was playing a lot with the AD9850, AD9851, AD9832, AD9833, AD9834, AD9835 for several years, driving them with PIC chips. The newer AD995x and AD9912 are more iinteresting, as they have better DAC resolution, higher clock rates, and lower power requirements. But, they have a harder package to use.
Typically, you can design a Cauer filter to not only provide a traditional low-pass function, but you can often tune the response for maximum loss at the DDS clock frequency (sample rate). That helps to reduce the garbage seen from these older DDS devices.
One thing to keep in mind is that programming these devices requires 32-bit and sometimes 64-bit math, and sometimes even floating point. The Arduino CPUs (except possibly the new Due) does not have a good math library. It only has “float” types, no doubles (which causes poor accuracy). Also, I’ve had problems (failures) in trying to use long-long (64-bit) ints, even though that is supposedly implemented. Using 32-bit long ints is problematic in doing the frequency-tuning-words (FTW). I’m having a similar problem programming the Si570 with an Arduino, I had to give up on precise calculations, and am using a “close-enough” approximation.
From: Karl W4KRL
Sent: Thursday, January 03, 2013 1:54 PM
To: tacos at amrad.org
Subject: idealCircuit simulator
I ordered an AD9850 DDS Arduino shield on eBay for less than $6. It should be here from Thailand at the end of the month.
A quick reading of the application notes suggests that the output should go through a 70MHz low pass filter for use in the 0 to 40MHz range. There are several nice on-line filter designers:
But after looking at the low-res circuit jpg on eBay it turns out that it has an on-board filter. The values are vaguely similar to what the on-line designers calculate for a 7-pole Butterworth Pi but each inductor is shunted with a small capacitor. I thought it would be interesting to figure out what the filter was supposed to do and looked unsuccessfully for an on-line circuit simulator. There is a fair one sponsored by Digi-Key at http://www.partsim.com/simulator but I found it so awkward to use that I gave up after drawing the circuit. It did not produce any useful results.
Then I came across idealCircuit at http://ic.sidelinesoft.com/
The download is only 1.4MB and is free. There are probably a lot of limitations but it looks good enough for quick general use. One obvious drawback is the lack of part IDs. Maybe there is a work around in the manual.
I drew the filter and got a voltage and phase graph in just a few minutes with zero frustration. The circuit and results are attached. It seems to be a 70MHz low pass with some ripple and peaks near cutoff. Comments on the design and performance of the filter are welcome. Why did they shunt the inductors with caps?
Happy New Year
Tacos mailing list
Tacos at amrad.org
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Tacos