I agree the Eye pattern looks too good, but why? Comparing the attached figure from Analog Devices AN-922 to Ken's filter plot suggests the current filter allows too much bandwidth.
Software to calculate component values for the raised cosine filter is what really is needed.
John
Nyquist Filter Values
Forum rules
This forum is run by the BATC (British Amateur Television Club), it is service made freely available to all interested parties, please do not abuse this privilege.
Thank you
This forum is run by the BATC (British Amateur Television Club), it is service made freely available to all interested parties, please do not abuse this privilege.
Thank you
Re: Nyquist Filter Values
- Attachments
-
- Raised Cosine Filter.PNG (31.25 KiB) Viewed 3655 times
Re: Nyquist Filter Values
I will let Brian answer but I think he is working on a digital filter for Digilite.
I wouldn't like to try and produce a filter like that in the analogue domain.
I did a quick Google and for low symbol rates something like the LTC-1569-7
might be suitable. Max cut-off freq 300 KHz with alpha of 0.5 not ideal
but better than nothing.
Another idea would be to feed the digital I and Q into two separate shift
registers which drive a ROM whose outputs drive an R/2R DAC I suppose.
Of course the above could be done with one of the £12 Chinese FPGA boards
on eBay some Veroboard + resistors for the R/2R DAC.
- Charles
I wouldn't like to try and produce a filter like that in the analogue domain.
I did a quick Google and for low symbol rates something like the LTC-1569-7
might be suitable. Max cut-off freq 300 KHz with alpha of 0.5 not ideal
but better than nothing.
Another idea would be to feed the digital I and Q into two separate shift
registers which drive a ROM whose outputs drive an R/2R DAC I suppose.
Of course the above could be done with one of the £12 Chinese FPGA boards
on eBay some Veroboard + resistors for the R/2R DAC.
- Charles
Re: Nyquist Filter Values
Hi John,
It appears that there are two goals you are trying accomplish with one filter.
A classic Nyquist for a transmitter is associated with the sampling rate of the incoming IQ data being sent to the modulator and has the purpose of filtering out aliasing spurs/images. So the Nyquist could be set to cutoff at 150% of the main signal (assuming that the sampling rate is very much higher) and do a good job of preventing aliases.
On the other hand, you are also trying to achieve a "wave shaping" filter to attenuate the ringing you describe coming from the serializer switching. It is a difficult problem you describe...and a Nyquist filter may not accomplish both goals.
Charles G4GUO (who better understands the math involved than I do) is correct in suggesting that digital filtering via software is going to be more effective than a filter designed in the analogue domain.
I have not used eye pattern analysis of transmission quality. Have you ever run an eye pattern side-by-side with a constellation analyzer (for example: Tutioune or SatLink) on the same signal? Seems like that would be an interesting comparison.
73...de Ken W6HHC
It appears that there are two goals you are trying accomplish with one filter.
A classic Nyquist for a transmitter is associated with the sampling rate of the incoming IQ data being sent to the modulator and has the purpose of filtering out aliasing spurs/images. So the Nyquist could be set to cutoff at 150% of the main signal (assuming that the sampling rate is very much higher) and do a good job of preventing aliases.
On the other hand, you are also trying to achieve a "wave shaping" filter to attenuate the ringing you describe coming from the serializer switching. It is a difficult problem you describe...and a Nyquist filter may not accomplish both goals.
Charles G4GUO (who better understands the math involved than I do) is correct in suggesting that digital filtering via software is going to be more effective than a filter designed in the analogue domain.
I have not used eye pattern analysis of transmission quality. Have you ever run an eye pattern side-by-side with a constellation analyzer (for example: Tutioune or SatLink) on the same signal? Seems like that would be an interesting comparison.
73...de Ken W6HHC