All,
?
Two weeks ago I reported about the relatively high noise floor level which I observed on my HiQSDR board. Based on the feedback I received, I ran several tests and experiments which eventually resulted in an improved noise floor.
- Change of the power supply did not help to resolve the high noise floor issue. The on board LDO¡¯s do their work as anticipated.
- Using a heat gun I noticed that the noise floor rose by 10 dB when I slightly heated up the ADS5500 (ADC) from the top side. When I subsequently applied the same heat to the bottom side of the PCB, the noise would drop very fast to the original (but a bit too high) level. Hence, I started to wonder whether the hidden ground pad of the ADS5500 was perhaps floating (i.e. not properly soldered)? So I removed the ADC, and resoldered it (that¡¯s the short version!). The result? NO DIFFERENCE whatsoever! The pad apparently was soldered properly after all.
- Next I inserted a 10 Ohm resistor in the Vcm feedback path between the ADC and driver (as per data sheet), but that did not noticeably improve the noise floor.
- Running out of options I wondered about the quality of the ADC clock source, but the type used on the board by DB1CC runs a very low jitter (< 50 fs), so could not be the cause, or could it? DB1CC uses a series resonance circuit (220 nH + 7.5 pF) between this 122.88 MHz oscillator and the ADC. On my oscilloscope the levels looked fine, but when I replaced the 7.5 pF cap with a tune cap I could fine tune and let the noise floor drop by 8 dB!! That accounted for much of the high noise floor that was present in my version. I just wonder what the use is of this series resonance circuit? The oscillator runs in a fundamental mode, i.e. there are no ¡°sub harmonics¡± present at its output. As an additional experiment I by-passed the series LC-circuit and gained another dB of improved noise floor. But that¡¯s not the end of the story.
I was curious about the dynamic range, so I applied a carrier and went all the way to -25 dBm (on 10.7 MHz) and found no obvious spurs popping up within the 300 kHz pass band. That¡¯s > 90 dB of dynamic range and the noise floor did not change during this test. Then I removed the by-pass and used the (tuned) series LC circuit again in the oscillator chain and noticed something peculiar: Applying the the same strong RF signal as in the previous test now raised the general noise floor by 10 dB! So by-passing the series LC filter provided much better performance than including it. I wonder if others have performed the same exercise?
?
Can anyone can enlighten me what the purpose is of this series LC-filter (perhaps improving the clock symmetry?).
?
73s,
?
Peter, PA3BIY
?
