EMRFD Message Archive 5855
Message Date From Subject 5855 2011-02-18 19:56:14 Richard Stasiak Crystal filter - pull frequency? I have been experimenting with a 2 pole crystal filter that I built from details in emfrd. It is about 2.5 kHz low of the desired 10 MHz centre frequency. I have tried a some random l and c paralled with the crystals. This changed the filter shape but not the centre frequency. Is there another way to shift this filter?
5856 2011-02-18 20:09:51 kb1gmx Re: Crystal filter - pull frequency? You need C in series with the crystals to "tune" them.
Looks at other examples in EMRFD for that.
5859 2011-02-19 13:01:30 ae5ew Re: Crystal filter - pull frequency? A quality trimmer or air variable would probably be best if you want decent stability.
5860 2011-02-20 01:03:34 Ashhar Farhan Re: Crystal filter - pull frequency? it doesn't help much, i have tried it. it only makes the impedance go
5861 2011-02-20 06:40:20 Leon Heller Re: Crystal filter - pull frequency? 5862 2011-02-20 06:56:39 WeiweiZheng Re: Crystal filter - pull frequency? YES, you need C in series with the crystals to "tune" them.
And the program named 'finetune' delivered with EMRFD is helpful¡£
5863 2011-02-20 08:40:05 kb1gmx Re: Crystal filter - pull frequency? The series C interacts with the shunt C so simply adding series C
will break a working filter. You have to start from a different calculation for each mesh (shunt c, series Xtal, shunt c) as the motional capacitance is effectively different. This is discussed
in EMRFD. I have made a few 8 pole ladder filters this way with
very good bandwidth and shape factors.
This approach will not fix(correct) crystals that are just off frequency. Also a bunch of random cheap crystals will not be
exactly on. For example I got a dozen 8mhz that were all
consistently 1.8 to 2.2khz high and another dozen that were
1.5-1.8khz low of the same brand but different lots. A third
batch were nearly exactly on frequency. All were within the
spec for accuracy.
I found they made very good half lattice type filters using
selected crystals from those three batches. Half Lattice filters require crystals with the pole and anti-pole frequencies to line
up so you use crystals of two or more frequencies separated by
about .5-2.5khz (depending on the desired bandwidth). They are
more involved but often have better shape factors.
So yes the simple ladder filters using 2 to 4 closely matched for frequency crystals is a simple affair but higher order or more complex filters are much more involved.
5864 2011-02-20 13:26:36 R Wall 100 Crystals for AU$5.00.... Hi all,
Found the following interesting link:
Maybe a homebrew PicaStar 11.98135MHz BPF, or whatever. 100 crystals for AU$5.00.
Roderick Wall, vk3yc.
[Non-text portions of this message have been removed]
5865 2011-02-20 16:59:21 Kerry Re: 100 Crystals for AU$5.00.... 5872 2011-02-22 19:51:07 Richard Stasiak Re: Crystal filter - pull frequency? Hi Glen
I have spent the last few nights determining the crystals motional characteristics using the
MyVna software and my N2PK vna. It's been fun as there are two methods to get the motional data, reflection and transmission techniques, and the results from each method are not correlating. I am planning to try and make sense of this tonight.
I also just discovered XLAD yesterday and will get familiar with it.
This is turning into a bigger project than I anticipated, however, somebody said, it's not the ending that's important, but the journey.
Thanks to and all of the others for your input on this subject.
BTW, I may be just down the road from you, qth here is St. Catharines.
On 2011-02-22, at 2:06 PM, Glen Leinweber <email@example.com> wrote:
> Not parallel capacitors - that will degrade
> the noise floor of the filter. If anything, use
> parallel inductors. This is awkward, since
> you need a LOT of inductance to resonate
> out the crystal holder capacitance.
> If your filter is low, then adding series capacitors
> to the crystal will shift it up. Series capacitors
> will also change the filter shape, and change
> the terminating impedance.
> The EMRFD software (XLAD) allows filter
> tuning with series capacitors. In any case,
> the two-crystal filter is dead easy to deal
> with, since both input mesh and output mesh
> are identical.
> I take it that your 10 MHz crystal have the same
> resonant frequency, so identical meshes
> will require identical series capacitors to
> bring center frequency up. XLAD should help
> you find the new terminating Z for your filter.
5874 2011-02-24 08:38:25 davidpnewkirk Re: Crystal filter - pull frequency? 5875 2011-02-25 09:32:55 g3xaq Re: Crystal filter - pull frequency? Hello Rick,
I hope you will find a few words of encouragement from a strictly hobbyist filter builder of some use...
In firstname.lastname@example.org, Richard Stasiak
> I have spent the last few nights determining the crystals motionalWell, if you have an N2PK then filter design becomes "easy" :-) My pal Steve/G3VMW has one and we've been using it to design and evaluate quite sophisticated filters built from a bag of 100 crytals we got off eBay for £1. Once you have used the VNA to measure all the motional parameters of a bunch crystals, and decided if their Q is high enough, you can use the latest version of Wes' excellent ladder filter design suite which you can download from here
> characteristics using the MyVna software and my N2PK vna.
to obtain capacitor values for your desired response shape. As always, the first step after downloading and unpacking the zip is to read the manual. It holds your hand through just about everything you need to know. We also found Bill Carver's paper on high performance filter design useful. It is included on the EMRFD CDrom.
xlad08 is the main design tool. There is a simulation and graphical display utility called gpla08, but G3VMW and I have found the freebie LTspice to have greater utility. This is because we are designing CW filters and want a narrow filter that does not ring excessively, unlike some of the offerings from the major Japanese manufacturers. As well as the usual amplitude response verus frequency, LTspice can simulate the time domain response of a filter when presented with a pulse of RF. A note on one of the forums showed us how to use the parasitic resistance, inductance and (parallel) capacitance in the standard LTspice capacitor model to create a "capacitor" that modelled a crystal's Rm/Cm/Lm/C0 with allegedly a significant improvement in simulation speed over separate components.
We use a 35wpm "dot" at our filter's centre frequency and then look at the output over a 50ms period. The simulations take ages to run because about 1000 time steps are needed per cycle of RF. However, the results are worth the wait. Narrow filters with Cohn, Butterworth and Chebyschev shapes look predictably horrible, with pings lasting many milliseconds after the main event. And the more crystals in the filter the worse it looks.
Avid readers of EMRFD and the rest of the literature will be unsurprised that the 8 pole filter design we selected is a "Transitional Gaussian to 12dB" shape. Our filter looks nicely selective for CW use, at least in LTspice: 2.7dB insertion loss, 360Hz at -3dB, 490Hz at -6dB, 1160Hz at -60dB, and 1950Hz at -100dB. Its time domain response is exceptional, with just a couple of pings at -37dB lasting for about 5ms after the trailing edge of the main pulse. And all this with a bunch of crystals that cost 8 pence total (about 6 US cents)!
Designing these esoteric filter shapes requires access to Zverev's "Handbook of Filter Synthesis", in which he laboriously tabulates the coupling coefficients and end-section q's for a zillion different filters. The data we want covers just a few pages of this book, so a technical library and a photocopier are probably the cheapest way to go. The VNA gives the crystal Q, which gives filter q0 via the 3dB bandwidth and centre frequency, and Zverev gives q1, qn and all the k's for this q0 for various filter shapes and number of crystals. You just type the numbers from Zverev into xlad08 and out pop all the capacitor and terminating resistor values. It's as simple as that!
And blow me, when you model the filter in LTspice it comes out pretty much bang on the money. By "pretty much" I mean within 0.5dB on the insertion loss and 5% on the 3dB bandwidth.
Meshtune08 addresses your problem of "my filter is too low in frequency", although moving it up 2.5KHz might be asking too much. The end of the manual in the zip file discusses this.
Steve and I have the opposite problem. Our 8 pole filter design is destined for use at the front of an IF chain. We also need a tail end filter to band limit the noise from the IF amplifiers. Something wider than the main filter that doesn't compromise the time domain response is what we want. Early simulations with simple 3 or 4 pole Butterworth filters around 1KHz or 2KHz wide showed them to be centred higher in frequency than the main filter. Their LF passband edge was close to the centre of the 8 pole's passband, which is about the worst possible place for introducing ringing. Moving the Butterworth tail end filter down in frequency to bracket the main filter would need about 10uH in series with each crystal (althogh some of the inductance could be absorbed into the adjacent series capacitor where there was one in the prototype design). Big inductors like this with low self capacitance are not trivial to make, so we looked at the Cohn Minloss design. This has no series capacitors in the inner meshes so is (I think) the shape that will yield the lowest centre frequency for a given bandwidth. Using the same crystals as in the main filter, a 4 pole Cohn with an 825Hz bandwidth has its -3dB points 20-30dB down the skirts of the 8 pole filter. Adding the tail end filter pulls in the -100dB bandwidth from 1950Hz to 1480Hz. It does not degrade the excellent time domain response because the 4 pole filter's wider bandwidth makes it respond "quickly", yet the main filter limits the rise and fall time of signals presented to it, and also restricts them to the "middle" of the passband where ringing is less severe.
Although we have built one prototype 6 pole Butterworth filter to validate the method, these 8 pole and 4 pole filters only exist inside our computers today. PCB layout is completed and SMD (Q>1000) capacitors are on order. Within two weeks we shall see if we are riding a horse or an ass.
5876 2011-02-25 09:55:03 Tim Re: 100 Crystals for AU$5.00.... I often end up finding that major-brand cheap microprocessor crystals from the major distributors (Mouser, Digikey, etc.) have average Q's in the 65000 to 150000 range.
In each batch there are a couple of "oddballs" that have markedly lower Q, and/or which do not seem to fall into the same center frequency distributi
5877 2011-02-25 11:00:49 w4zcb Re: Crystal filter - pull frequency? I wish, but actually, it goes the other way. the 8 pence comes to
something like 13 cents US.
Avid readers of EMRFD and the rest of the literature will be
unsurprised that the 8 pole filter design we selected is a
"Transitional Gaussian to 12dB" shape. Our filter looks nicely
selective for CW use, at least in LTspice: 2.7dB insertion loss, 360Hz
at -3dB, 490Hz at -6dB, 1160Hz at -60dB, and 1950Hz at -100dB. Its
time domain response is exceptional, with just a couple of pings
at -37dB lasting for about 5ms after the trailing edge of the main
pulse. And all this with a bunch of crystals that cost 8 pence total
(about 6 US cents)!
5878 2011-02-28 19:59:21 Rick ve3mm Re: Crystal filter - pull frequency? Hello Alan
Thanks for the words of encouragement. I have been spending some free time
learning the software tools that you have suggested. It will take me some
time (if ever) to get to designing and constructing an eight pole filter.
A question, regarding using the N2PK vna to obtain the crystal
characteristics. I measured both crystals using the MyVNA program using both
transmission and reflection techniques. In transmission mode I used a
Clifton Labs crystal filter test fixture (12.5 ohms). I directly connected
the crystal the DUT port of my reflection bridge for the reflection test.
The results of the tests were quite different for each technique. It
appears from examination, that the motional parameters derived for the
reflection technique are the most accurate.
I wonder what measurement technique you and Steve used and if you had
5879 2011-03-01 07:00:17 Stephen Wilson Re: Crystal filter - pull frequency? 5914 2011-03-13 17:43:50 William Carver Re: Crystal filter - pull frequency? HINT: While it may not move a filter as far as you want, you can move
the filter center frequency up by choosing a higher impedance level for
the filter (not your input/output impedance, but the impedance of the
Some xtals in distributor catalogs, 4MHz, 5 MHz, 8 MHz, etc, have a
choice of series resonant or parallel resonant calibration. The only
difference in the two is the parallel resonant crystals are series
resonant lower in frequency than the marked frequency. The difference is
usually several KHz, depends on the crystal frequency. But it's
something to keep in mind.
5915 2011-03-13 17:49:33 William Carver Re: Crystal filter - pull frequency? another HINT: A crystal is a high Q device. It takes
milliseconds....sometimes quite a few milliseconds....for amplitude to
settle down after a frequency change. You can't sweep quickly through a
network with an undamped crystal and get correct numbers for amolitude
I was getting totally bogus numbers and Paul Kiciak, N2PK, finally hit
on this issue. After I slowed the VNA down, allowing lots of time for
the crystal to settle down on the new test frequency , the numbers
suddenly started looking correct.
5916 2011-03-13 20:44:40 Richard Stasiak Re: Crystal filter - pull frequency? Awesome! Thanks for the hint. I'll try slowing down the scan of the vna, Bill.
On 2011-03-13, at 8:49 PM, William Carver <email@example.com> wrote:
> another HINT: A crystal is a high Q device. It takes
> milliseconds....sometimes quite a few milliseconds....for amplitude to
> settle down after a frequency change. You can't sweep quickly through a
> network with an undamped crystal and get correct numbers for amolitude
> and phase.
> I was getting totally bogus numbers and Paul Kiciak, N2PK, finally hit
> on this issue. After I slowed the VNA down, allowing lots of time for
> the crystal to settle down on the new test frequency , the numbers
> suddenly started looking correct.
[Non-text portions of this message have been removed]
5917 2011-03-14 06:10:08 Tim Re: Crystal filter - pull frequency?