EMRFD Message Archive 298
Message Date From Subject 298 2007-01-04 04:14:14 neomag_magneo Bandpass filters with series resonators: a problem in the design Hello group!
Thanks for the advice concerning the H-mode mixer transformers!
Another problem still bothers me:
I have tried to reproduce the filter designs (triple tuned circuit,
series resonators) as given for the Lichen transceiver in EMRFD. The
MathCad sheet (or its equivalent in Excel spreadsheet) seems to fail
when the numbers for 80 m TTC are plugged in, the bandwidth cannot be
more than 500 kHz. When analyzing the very filter with GPLA or
equivalent, it seems to work as expected, anyhow.
Is there something wrong with the MathCad formulas, or have I missed
something. All my efforts with parallel resonators have been quite
succesful thus far (receicer preselectors and BPFs for transmitters),
but now I am really stuck !
73 de Heikki (OH2LZI)
300 2007-01-04 15:21:45 Wes Hayward Re: Bandpass filters with series resonators: a problem in the design Hello Heikki, et al,
Many thanks for the comments regarding the MathCad sheet on my web
site. There is no error, but there is a problem.
For most readers who have not been chasing this issue, this is not
directly about what is in EMRFD. Rather, it regards an extension on
my web site having to do with the design of higher order filters
using the EMRFD programs. That piece includes some MathCad sheets
that show the equations for the design of double and triple tuned
circuits that have series resonators. A problem shows up when you
insert the filter parameters for a filter with center frequency of
3.75 MHz using 15 uH inductors with a Q of 50 and a filter bandwidth
of 700 kHz. Dropping the bandwidth to 600 kHz produces a viable
I have updated the discussion on my web site. See
The problem comes about with wide bandwidth which sometimes produces
a negative value inside a square root symbol. This then generates a
complex capacitor value.
The programs include k and q values for Butterworth filters. If
Chebyshev k and q values are substituted, the complex valued
capacitors are avoided. Chebyshev k and q data are available in
Zverev's text, other texts, or are generated by a routine built into
XLAD, the crystal filter program included with EMRFD.
I analyzed the filters in Fig 6.135 of EMRFD and the values shown
there are OK.
Hope this helps.
301 2007-01-05 03:50:57 neomag_magneo Re: Bandpass filters with series resonators: a problem in the design Hello Wes and others,
First, my apologies for not referring to the W7ZOI-website where the
Mathcad stuff can be found.
Thanks for the clarifications, it seems that I just have encountered
one of the inevitable limitations in current design methods. I agree
with Wes that all the filters given Fig 6.135 of EMRFD should be OK, I
have analyzed (with AADE Filter Design, freeware, highly recommended
!!!) the 3.75 MHz version and also built one with results very similar
to the simulated frequency response. This very filter, however, (with
the component values given in the EMRFD table), cannot be an outcome
of the Mathcad-procedure (for the reasons explained by Wes), and this
is what made me doubt the whole method in the first place. When the
bandwidth is reduced, a viable filter can be designed. The only
problem lies in the Q-value (50) which is kinda low for toroids, at
least. Furthermore, the whole design is highly dependent on this
Q-value, which in my opinion, should be as high as possible.
I have a feeling that the true problem is the high inductance value
which could be reduced by a factor of two at least. The similar filter
found in the Triad tranceiver (or its predecessor CDG 2000) has 8 uH
inductors for the same bandwidth. Also switching into
Tchebyshev-response (with 0.1 dB ripple) does the job, even with 15 uH
inductors. I cannot quite understand the popularity of Butterworth,
when better equal ripple filters are available with similar effort,
the small ripple should be tolerable in most applications...
In summary, the design methods do require some learning effort, and in
that sense this short episode has been most educative for me, at
least. From that viewpoint, I was actually lucky to get stuck with the
single example not fully compliant with the design mathod given. A
good starting point for an average design might be a reasonable
inductor, following that rule I would not have encountered the
problems with fitting some 55 turns on a T50-coil form to make the
magic (!) 15 uH inductor, hihi.
Finally I tried a 4th order series resonator filter design as proposed
by Wes in the wed-article and got some nice results. After building
quite a few parallel resonator filters with almost zero top coupling
required, I will aplly this new (?) method for the receiver under
There is an active group and website in Scandinavia, the aim is an
ssb-tranceiver utilizing currently available smd-components, DDS etc.
The stuff can be found starting at:
The language is variable, some English, too ! The transceiver is also
available in kit form, as the trend seems to be these days.
Personally, I am no real fan of any construction kits, but only wanted
to inform the group of the activities on this side of the pond.
302 2007-01-05 12:38:18 Wes Hayward Re: Bandpass filters with series resonators: a problem in the design Hi Heikki (and the group),
One of the fun and exciting things about filters is that as we get
deeper into any part of it, we start to see more and more about the
complexities. As the risk of making things even more complex, I
expanded the triple tuned circuit part of my web page to include some
Chebyshev filter designs. That turned out to be most interesting.
By allowing a passband ripple of 1 dB, it was possible to extend the
bandwidth of the filter we have been considering to 1.34 MHz.
I put the details on my web page, and included some plots of S21 and
S11 for a filter with very high Q inductors. The only change of
significance when a low Q=50 inductor is used is increased passband
You are correct -- we do always want to go for Q values as high as we
can get. But this brings problems. The high Q elements are often
larger than we would want. They can become expensive. I picked a
Q of 50 for these 15 uH inductors because it is possible to purchase
inexpensive RF chokes with this L value that have this Q. When we
use them in a wide band filter, the loss is not an issue.
I'm not sure that it is valid to call the CDG-2000 a predecessor to
the Triad transceiver. G3SBI, W4ZCB, and W7AAX were working on the
the Triad as early as 1995, if not before. It has been kicking
around for a long time, much of it the result of the early H-Mode
Mixer work of Colin, G3SBI. I don't recall when the CDG-2000 made
it into print. I think that it generally followed from the front
end work of the Triad. W4ZCB may be able to shed light on this.
The CDG-2000 includes a lot of other details, especially in the DSP
area. The later variation is the STAR. While I have not kept up
on it, this design has become very popular. I believe it also uses
the same G3SBI front end topology.
Many thanks for the link to the Finish transceiver. That should be
an interesting one to investigate.
304 2007-01-06 06:50:52 w4zcb77 Re: Bandpass filters with series resonators: a problem in the design > By allowing a passband ripple of 1 dB, it was possible to extendthe
> bandwidth of the filter we have been considering to 1.34 MHz.and
> I put the details on my web page, and included some plots of S21
> S11 for a filter with very high Q inductors. The only change ofpassband
> significance when a low Q=50 inductor is used is increased
> loss.Wes, I always attributed Chebychev "ripple" with return loss. (ie,
0.044 dB specified ripple will yield a 20 dB S11) I haven't been
following the thread sufficiently closely to understand where you're
going with this, wouldn't another topology be prefered for the goals?
> I'm not sure that it is valid to call the CDG-2000 a predecessor
> the Triad transceiver. G3SBI, W4ZCB, and W7AAX were working onthe
> the Triad as early as 1995, if not before. It has been kickingmade
> around for a long time, much of it the result of the early H-Mode
> Mixer work of Colin, G3SBI. I don't recall when the CDG-2000
> it into print. I think that it generally followed from thefront
> end work of the Triad. W4ZCB may be able to shed light onthis.
> 73, WesWell, It all started with Makhinsons HDR in the fall of 1990
Communications Quarterly. (I miss that almost as much as I miss Skip
Tenney's "Ham Radio".) That started Colin G3SBI to thinking (He's a
great outside the box thinker!) and he shortly removed the signal
path from the LO path with his "H" mode mixer. Initially performed
with the SD-5000 from Siliconix, it blossomed with the advent of the
quad buss switches as in the FST 3125/6 series. (And is now
progressing very interestingly with the FSAV 332 series and his new
fundamental frequency squarer.) If we wait just a bit more
patiently, technology is going to solve all our problems, but my
personal definition of Hell is to have the perfect radio and no
interesting hams with which to have an intelligent QSO.
I still have a shop SD-5000 receiver, Ip3 measured >+50 dBm, and no,
I didn't measure that, Colin did with the Queens finest test
equipment. My capabilities poop out in the vicinity of +40 and even
that somewhat depends on which cheek I hold my tongue in.
Bill and I communicated with Colin as early as 1993, and I met him
first time on a visit to Manchester in 1996 when Alice and I
travelled to the UK celebrating our 50th anniversary. We trekked to
Darsbury Labs where he measured my first efforts.
The CDG 2000 was an effort spearheaded by George Fare, G3OGQ and a
superb homebrewer despite a career in a totally different discipline
than electronics, and Dave did a magnificent job on the software and
controller for the transceiver. A three handed effort, not bad for a
product that transcends most of todays commercial efforts with
staffs of hundreds, but then, they didn't have to bring out a new
radio every 30 minutes.. The radio was published in RadCom, the
RSGB's equivalent of QST in a 3 part series in October, Nov and Dec
of 2003. I know that Dave has done some work on a DSP package for
the radio, but am not aware of anything published on that addition.
The PicaStar, G3XJP's magnificent project spanned some 20 issues of
RadCom, starting with the first in June of 2003. By that time, 8 of
us in the Beta group had finished 8 beautiful DSP transceivers and
were enjoying a truly unique experience. A radio that kept being
reborn with every new issue of Peters software. I've had something
like 20-25 new radios in the last 4 years despite my trusty dusty
old Signal One serving me rather faithfully for over the previous
(Those 20 issues of RadCom would have had me turn in my life
membership in QST if it had only been printed on heavier paper and
in standard US sized magazines so it would fit my bookcase.)
The Picastar Yahoo group is now nearing 1000 members and the radio
transceiver is being built in many countries. The Ukraine has
several, a large group of students in Syria are building them as a
part of their head of department in the EE school of Engineerings
efforts. There are presently two operational in the US and only one
in Canada, but the list of European constructors is a very large
Peter is presently working diligently on a STAR II, second
generation DSP radio, using the original hardware as a vehicle,
there's a whole new paradigm around the corner.
> Regards to the group