BATC Forum

I have been looking for Nyquist filter values for the lower symbol rates for some experiments I am doing.
John O'Loughlin kindly sent me an Excel spreadsheet he made for Digilite.

I have inputted the ones I needed and the results are below, of course I will need to select the nearest preferred values.
Posting here for info and any comments.

Colin G4KLB

Hello,
For My Digilite at SR 250, I have used : 4.7nF / 100µH / 15 nF / 122 µH / 4.7 nF
seems not to be the best, I obtain only a Max of 17 dB of MER when receiving with Tutioune1600_lowSR.

I can have 30 dB of MER using another modulator and Brian can obtain also similar value using his Digithin.

I will try new values.
Jean Pierre F6DZP

Thank you Jean Pierre,

I have been trying to find the best Values for the low symbol rates, as I believed we don't have the optimum values yet,
Talking to John he states 'The calculations on my spreadsheet simply scale the values from Dave’s 4Ms/s chart entry according to the frequency change'
I assume some rounding up/down would have been done on the 4Ms/S to use preferred values, so using these values to calculate say 250Ks/S would introduce a significant error.

We have been unable to find a calculator that gives component values.

I posted the chart to provoke some discussion, so apology's to John for dropping him in it! He had only sent it to me for info, and was not expecting it to be published!
I had no other source for the values to use, so was a reasonable starting point.

Hopefully someone reading this will have been down this route and be able to guide us.

Colin G4KLB

Hi Colin G4KLB,

G4GUO introduced the DATV-Express team to a software program called "ELSIE" (pronounced "L C") that is contained inside the ARRL HANDBOOK on a CD-ROM with an assortment of other software tools. Elsie is very useful and made by http://www.tonnesoftware.com (not freeware).

Here are two examples of the values used in the list of your original posting. I plugged in values for 4 MSymb/sec and 1 MSymb/sec to just show a little of what Elsie can accomplish.


Elsie tool using 4 MSymb/s values from original posting


Elsie tool using 1 MSymb/s values from original posting

I hope this info on the Elsie tool is useful to you....de Ken W6HHC

Hi Ken,

The DigiLite filter is terminated in 100 ohms, could you re-plot with that value?

A raised cosine filter is the norm for QPSK filtering, any chance you have found software for that type?

John

Hi John,

My main motivation was to lead Colin to a useful filter-design-and-analysis tool called "ELSIE" that he/others can also use. It turns out that www.tonnesoftware.com has an apparently free "Student version" to download that is limited to 7-poles. They also sell a key to their "Professional version".

Here are two re-plots using 100 ohm termination for the ~1 MSymb/s region...one using the default Bessel values selected by the Elsie tool and one with plugin values for 1MSymb that were listed in original posting.


Elsie tool using 1 MSymb/s values via Elsie default for Bessel design (100 ohm termination)


Elsie tool using 1 MSymb/s values listed in original posting (100 ohm termination)

Finally, I have attached a screen-capture of all the filter-family-types that Elsie supports (center of screen). Also notice that Elsie has ability to choose closest 5% component values for user if they wish (see red arrow).


Elsie Screen-shot showing all filter types that are supported.

73...de Ken

The Nyquist filter reduces the transmission bandwidth required. However the pulses from the serializer passing through the filter cause ringing on the output. The energy of one pulse can spill over into subsequent time slots causing ISI - Inter Symbol Interference. To reduce this effect the filter should be designed to cause the nulls in the waveforms from the previous pulses to occur at the peak of the current pulse. This makes it much more than just a low pass filter.

www.feng.pucrs.br/~decastro/pdf/ShapingFilterDesign.pdf

The Eye diagram should show if this result has been achieved. I haven’t had a chance to try this on a DigiLite yet. I think the Eye pattern could be displayed on a Digilite by syncing on the clock signal at PIC pin 25 and displaying the output of the filter.

http://users.ecs.soton.ac.uk/sqc/EL334/MODEM-L4.pdf
Example Eye diagrams on pages 11-13

Thanks to Ken and John for your info ...... keep it coming

I got a chance to make the Eye pattern on a 4Ms/s Digilite today. It looks surprisingly good. Now someone who runs low symbol rates needs to try it.

John

A bit too 'good'.