Jeff, I believe the answer lies in the spur reduction algorithm. This line is the key:
======================================================================== = long sr_tuning_word = (tuning_word & ~(0x80007fffffff)) | 0x000080000000; // clear first bit, low 31 bits; set bit 31 ======================================================================== = Setting bit 31 for the spur reduced tuning word (sr_tuning_word) may be what is keeping it from being a perfect 50MHz on the hardware. Eric Wachsmann FlexRadio Systems > -----Original Message----- > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] > radio.biz] On Behalf Of Jeff Anderson > Sent: Wednesday, December 07, 2005 7:47 PM > To: flexradio@flex-radio.biz > Subject: Re: [Flexradio] Interesting behavior when connected to a dummy > load > > Great! Many thanks for the reply, Bob. Could I ask a few more questions > to > try to get a better understanding of what's going on? > > 1. Does the SDR software skip the 9854's FTW of 0x4000(+ more zeroes) > that > would be used if the DDS chip was set to 50 MHz? I'm trying to explain to > myself why I don't see the beautiful spur-free spectrum that I see when > the > DDS is tuned to *exactly* 50 MHz (receiver tuned to 50.011 plus pocket > change) with Spur Reduction OFF, compared to when Spur Reduction is ON > and > I tune through the point where the FTW ought to be 0x4000... . > > 2. If clock leaks are worse at submultiples of 200 MHz, how does this > manifest itself? Again, my panadapter looks fantastic when the divisor of > 200 MHz is exactly 4. It's only as I tune off this point that it quickly > degrades. In other words, at a divisor of exactly 4, clock leakage (per > the > SDR1K's panadapter display, at least) wouldn't seem to be an issue. > > 3. Regarding the 9958 - how does the 10-bit DAC of the 9958 affect spur > performance compared to the 9854's 12-bit DAC? Per my understanding, even > if one has chosen FTW's that "zero-out" the phase-remainder (so that there > is no phase "quantization error"), there is still an amplitude error > introduced by the phase-to-amplitude converter's sin/cos conversion of the > phase to an amplitude that's 10 or 12 bits wide (depending upon DAC > resolution), and even with a completely linear DAC and no clock leakage, > I'd > think this should result in spurs, too. Are these spurs significant? > (I'm > just trying to gauge the drawbacks of the 9958 - it's faster, but the DAC > resolution is less). > > Thanks for any insights you can provide! > > - Jeff, WA6AHL > > > > -----Original Message----- > From: Robert McGwier [mailto:[EMAIL PROTECTED] > Sent: Wednesday, December 07, 2005 1:17 PM > To: [EMAIL PROTECTED] > Cc: 'Jeff Anderson'; flexradio@flex-radio.biz > Subject: Re: [Flexradio] Interesting behavior when connected to a dummy > load > > > Gents: > > Please bring up the message I put out a couple of days ago. I ended it > with the AD9854 has leakage of clock and some modest DAC nonlinearities > that cause some spurs. > > Jeff has discovered the worst of them. 200 MHz (clock)/ 50 MHz > (frequency of interest) is 4. The clock leaks on all submulitples of > 200 MHz but the ones at 50 MHz are the worst. This is an internal fault > of the AD9854 and its worst feature. With a slightly more complex spur > reduction routine, that uses both sides of the 1/f hump at zero, we > could help considerably and never see these things. You would still see > the impact of reciprocal mixing but its impact would be greatly > diminished. > > As I said before, Analog Devices has learned a lot and we are learning > along with them. > > Bob > > > > _______________________________________________ > FlexRadio mailing list > FlexRadio@flex-radio.biz > http://mail.flex-radio.biz/mailman/listinfo/flexradio_flex-radio.biz
