EMRFD Message Archive 13578

Message Date From Subject
13578 2017-02-10 00:22:04 Hans Summers Re: Digest Number 2831

Hi Allison
 A vfo running at say 39-81mhz  and then divided by 2

 gets a good start for isolation and stability for the HF range.

 it helps if the tuning range is greater than 2:1.

 For the 3 to 30 range the number of filters are not too bad.

I tried 2:1 filter ranges first before arriving at the final filter configuration 60-40-30-20-15-10m. 

I think that it all depends what performance is desired. The original poster of the EMRFD thread seemed to want 1dB flatness and -40dBc spurii. So I took that as the "requirement" to aim for. A "sharp filter" often means a more exotic filter topology, which introduces passband ripple - which could perhaps damage the "1db flatness" requirement. *IF* it is a requirement. 

With Power of 2 frequency ranges is I think very difficult to achieve satisfactory performance. Squarewaves DO have content at even harmonic. Yes it is a lot lower than odd harmonics but it is still there. In the border case where you have the sig gen on just over 2MHz and the filter has a 4MHz cut-off, the 2nd harmonic is not significantly attenuated. 

Again it comes back to what performance is required. Everything is a trade-off as usual. With the 2:1 range the 2nd harmonic shows up on some frequencies, at -23dBc. So I decided 2:1 is not good enough... not for those performance targets anyway, if we want them to apply across a wide frequency span. 

Therefore I used the final filter lineup of 60, 40, 30, 20, 15, 10m. The ratios are a lot more forgiving. The results are I think, excellent over the range 3-30MHz: comfortably within the 1dB flatness, and the -40dB requirement (my flatness is +/- 0.4dBm and worst case spurii was -45dBc). Best of all, completely un-optimised, filters built just by "counting turns", and the whole thing lying unshielded on the bench. So there is lots of scope for improvement if people wanted. But for my target "mere mortal" good results can be achieved without special equipment or skill. See graphs:

73 Hans G0UPL
13580 2017-02-10 10:02:02 kb1gmx Re: Digest Number 2831
HI Hans,

I had to dig hard to see your text (view as source), no idea why.

Yes 2/3 octave filters would work better but more switching.  Plus I was talking about continuous coverage from say 1.8 to 30 as signal gen rather than ham bands only.

My approach to the SI570 problem was simpler.  It only gores down to about 10mhz so that had issues.
the other was harmonics.  Solution in my unit was run the NCO from 80 to 150mhz into a 9 pole low pass 
with a flat passband snf cutoff of 150mhz, attenuate it, and mix with 80mhz crystal osc taking the result 
through a 9pole low pass and attenuator this time with a 60mhz cutoff.  The resul is near DC to 60mhz
very flat and most spurs 35db down or more.  Why only 35db?  Mixers are non linear and make harmonics inside (DBM are notorious).  When at setup was paired with octave filters everything in the suboctave range was better than 40db down (some bands more than 50db) as it stripped the harmonic mixing.

I tend to assume most sources have some harmonic output as even -40db is fairly good.  Thing like IMD measurements need clean sources for high accuracy or you end up measuring the sources own 
contributions.  The other measurements like S11, return loss, and SWR are impacted by harmonic content.
Those however are less sensitive and if a tuned detector (or spec ana) is used then harmonic content is
less problematic.

In the end there are times when gear is handy and maybe required for critical work but for most uses 
as in does it work, or basic measurements, are far less critical.