Most hobbyists use crystals as an external clock signal for a microcontroller. A less common use would be to make a bandpass filter (BPF) for an RF signal. [Dan Watson] explains his crystal ladder design on his blog and links to several sources for understanding the theory and creating your own crystal ladder band pass filter. If you want a set of these purple PCBs you can order them straight from the purple fab.
[Dan]’s schematicOne of the sources that [Dan] cites is [Larry Benko]’s personal site which is primarily dedicated to amateur radio projects. Which you can find much more in-depth information regarding the design of a xtal BPF. [Larry] goes into detail about the software he uses and some of the applications of crystal ladder filters.
BPF designed by [Larry]The process includes measuring individual xtals to determine which ones will work together for your target frequency. [Larry] also walks you through the software simulation process using LTSpice. If you aren’t familiar with Spice simulation you can get caught up by checking out the series of Spice articles by our very own [Al Williams].
18 thoughts on “Designing A Crystal Ladder Bandpass Filter”
Been there, done that.
You must buy 50 or 100 crystals, and measure every one of them just to select those 4.
No rocket science….grab some (very) old book and everything is explained.
Exactly, AT LEASST 100 pcs. And you will need a well calibrated vector network analyser like VNWA or NWT, they are also not the cheepest HAM tools. Looks like it is better to buy a quality assorted xtal set and forget about the measurements and chinese xtals. Or buy a calibrated xtal filter. Do not forget that xtals wears out, slowly but they will …
Are you sure? You might be, but I really don’t see that claim being backed by any numbers.
Considering you get 14.31818 MHz oscillators as a standard frequency¹ and you can get one with 25ppm frequency accuracy for < 1.50€² , I'd assume that ordering a set of four will give you four devices with a *very* close parameter set for 6€. Doesn't sound that bad. Still sounds cheaper than buying 100 cheap ones…
Yep, I’m sure. You gave nice numbers and PPMs here but this is only a theory. How many xtal filters have you built? Or how many xtals have you bought with 25ppm and have measured them? This is what I’m talking about! REAL 25ppm xtal sets are not cheap. There is a definition “matched crystal set” and forget about cheap <1.50€ xtals. You usually realize it when you measure, e.g. lot of wasted time and money …
It’s better than the G3UUR design in the ARRL handbook. Trask explains why.
I’ve been working on an automated unit built around an MSP430, but that project got stalled for a while by other tasks. Buying a filter is certainly a popular option, but not cheap. But if you’re willing to expend labor instead, you can get xtals for US$0.10 each.
I’m a bit baffled by the “xtals wear out”. I’ve certainly never seen that mentioned in any literature I’ve read. They do age, but that’s not wearing out.
You don’t have to measure all the xtals. You just need a set with the correct parameters. You could get lucky and find the first 4 you measure will work, but that’s not likely.
Without a VNA you will never have the required results, especially when you are building your rig all DIY. Calculate everything, but only when you measure the filter you will see what’s happening with your signal.
Yes, pencil. You can then erase it if you go too far.
Grinding raises the frequency, the pencil lowers it. You can’t go to far in either direction, the crystal will lose its ability to oscillate, I seem to recall because by hand it’s an uneven process. I did it at least once, to move a surplus crystal a tad so it would be on frequency after it was multiplied up.
And that was mostly in the days of ft243 and the like holders, which could be unscrewed to get to the crystal blank.
Hermetically sealed crystals are trouble enough opening, but I’m not even sure how they are connected electrically to the blank, and the blank is smaller and more fragile.
But this only gets you on frequency, if it could be done with current crystals, and you can measure their frequency accurately. It doesn’t address other criteria that might make one crystal better than another.
Yeah, the trouble is that modern designs like this one never seem to bother explaining how to do that. They just assume that the reader knows all the technical details.
For a long time “crystal filter” meant a single crystal. Very selective, but not great skirt selectivity. To make it wider, it was loaded down with resistors.
When SSB started becoming popular in the fifties, something better was needed, so the half-lattice filter became common. You needed crystals spaced a certain amount apart, and since one section usually wasn’t good enough, you needed two crystals per frequency. Initially these were down at the “classic” IF frequency, 455KHz. There were a lot of ft243 type crystals left over from WWII, and if you couldn’t find the matching pairs, you could open them up and grind them. Nobody had frequency counter, so everything was fairly vague.
Later, some used higher frequencies, in the shortwave range. But there wasn’t likely to be ground o precise frequency, since the move was to hc/6 type crystals. Though one guy from British Columbia was building direct conversion SSB transceivers with the crystal filter on the signal frequency, so they had to be exact; he unsealed the crystal cases and ground them, but still he had to start with crystals close to frequency.
In hobby circles, crystal ladder filters didn’t appear till the seventies, in European publications. But not much talk of crystal matching or individual specs. Back then, one was to make assumptions, and scale from the schematic.
Tat sort of thing didn’t appear in US hobby publications till the late seventies. I think the 1979 ARRL Handbook has half a page about crystal ladder filters.
I think it was the late eighties, maybe early nineties, when the whole routine was discussed, match the frequency, learn the specs, and then design from there. Obviously they’d be better filters, but some of it probably came from more calculators or computers, to do the math, and frequency counters being much more common. Also, as digital electronics became o much more common, crystals became a commodity, so instead of just being able to buy 100KHz and maybe a few others off the shelf, the variety became much wider. You could find crystal near here you’d want it, and they were cheap compared to custom ground crystals. Even when bought in quantity so you could find matches.
Has anyone considered simply thermally stressing the ones which are “nearly” right ie 20+ hours at 205C? this is a well known hacker method and if nothing else is also good for pre-baking crystals on critical projects to weed out ones likely to drift in the first 1K hours.
I’m just preparing to make one crystal ladder filter, which is supposed to pass 27 MHz, but greatly attenuate (it does not have to be perfect, 100x attenuation will be great) any frequency more than 10 kHz away. How many crystals would be required for this sort of selectivity? Based on my calculations, one would be enough. I don’t care much about the shape of the bandpass curve. Thanks.
Crystal filters are defined by bandwidth. The way you describe it, it sounds more like you need to knock out the other, which might best be done with notch filter at the offending frequency.
You do care about the shape f the bandpass curve since you are concerned bout the ignl 10KHz away.
Selectivity is “easy”, wider bandwidth is harder, and good skirt selectivity (so the sides of the filter are sharp rather than broadening out) requires cascading sections. The higher the frequency the harder it is to make a crystal filter, especially when the crystals may not be on their fundamental frequency up there.
Almost a hundred years ago Howard Armstrong solved some problems by inventing the superheterodyne receiver, where the incoming signal was converted to a fixed frequency, where it was easier to have gain and selectivity. So you wouldn’t have to have selectivity on the signal frequency, If the high selectivity is on the signal frequency, it has to be exactly on frequency, and if you need a different frequency, the filter has to be rebuilt.
Only exotic applications would put a narrow filter on the signal frequency.
That said, one simple method of crystal filters is if you have a multiple stages of amplification, you put a crystal across the emitter resistor (or cathode resistor if a tube). It will give most gain on the crystal frequency, and doing this a few times will improve the skirt. I’ve thought about doing this for a WWV receiver, since one if its frequencies is 10MHz, and crystals are easily available cheap for that frequency. And it’s an easy way to get selectivity with minimal effort.
Thanks for your reply. Believe it or not, it is a regenerative receiver with one stage of added RF amplification and 3 stages of LF amplification. It will receive low powered telemetry data (20 mW TX at the moment). I currently have issues with some CB interference. Voices sound like they are from the southern US, even though I’m in Ontario… This thing you mentioned about emitter resistor is very interesting, I will probably try that.
![BPF designed by [Larry]](https://hackaday.com/wp-content/uploads/2016/03/20m_xtalbp_s21close_7-23-2009.gif)
Been there, done that.
You must buy 50 or 100 crystals, and measure every one of them just to select those 4.
No rocket science….grab some (very) old book and everything is explained.
Exactly, AT LEASST 100 pcs. And you will need a well calibrated vector network analyser like VNWA or NWT, they are also not the cheepest HAM tools. Looks like it is better to buy a quality assorted xtal set and forget about the measurements and chinese xtals. Or buy a calibrated xtal filter. Do not forget that xtals wears out, slowly but they will …
Are you sure? You might be, but I really don’t see that claim being backed by any numbers.
Considering you get 14.31818 MHz oscillators as a standard frequency¹ and you can get one with 25ppm frequency accuracy for < 1.50€² , I'd assume that ordering a set of four will give you four devices with a *very* close parameter set for 6€. Doesn't sound that bad. Still sounds cheaper than buying 100 cheap ones…
¹ http://de.farnell.com/webapp/wcs/stores/servlet/Search?catalogId=15001&langId=-3&storeId=10161&categoryId=700000005192&sort=P_PRICE&showResults=true&aa=true&sf=502,708&pf=312004262
² http://de.farnell.com/txc/7w-14-31818mba-t/osc-14-31818mhz-3-3v-smd-7-0x5/dp/1842094
Yep, I’m sure. You gave nice numbers and PPMs here but this is only a theory. How many xtal filters have you built? Or how many xtals have you bought with 25ppm and have measured them? This is what I’m talking about! REAL 25ppm xtal sets are not cheap. There is a definition “matched crystal set” and forget about cheap <1.50€ xtals. You usually realize it when you measure, e.g. lot of wasted time and money …
A VNA is handy, but not required. It does require lots of measuring though. Chris Trask has a good design for an xtal test set:
http://home.earthlink.net/~christrask/Crystal%20Test%20Set.pdf
It’s better than the G3UUR design in the ARRL handbook. Trask explains why.
I’ve been working on an automated unit built around an MSP430, but that project got stalled for a while by other tasks. Buying a filter is certainly a popular option, but not cheap. But if you’re willing to expend labor instead, you can get xtals for US$0.10 each.
I’m a bit baffled by the “xtals wear out”. I’ve certainly never seen that mentioned in any literature I’ve read. They do age, but that’s not wearing out.
You don’t have to measure all the xtals. You just need a set with the correct parameters. You could get lucky and find the first 4 you measure will work, but that’s not likely.
Without a VNA you will never have the required results, especially when you are building your rig all DIY. Calculate everything, but only when you measure the filter you will see what’s happening with your signal.
About the “xtals wear out”, yes. The frequency will drift with the years, all the xtal filters need maintenance/tuning. Have a look here:
http://www.conwin.com/pdfs/aging_perf_crystals.pdf
Or draw on the crystal with a marker pen a bit at a time.
I thought a standard graphite pencil was use when working with crystal holder that made it easier to get to the quartz wafer.
Yes, pencil. You can then erase it if you go too far.
Grinding raises the frequency, the pencil lowers it. You can’t go to far in either direction, the crystal will lose its ability to oscillate, I seem to recall because by hand it’s an uneven process. I did it at least once, to move a surplus crystal a tad so it would be on frequency after it was multiplied up.
And that was mostly in the days of ft243 and the like holders, which could be unscrewed to get to the crystal blank.
Hermetically sealed crystals are trouble enough opening, but I’m not even sure how they are connected electrically to the blank, and the blank is smaller and more fragile.
But this only gets you on frequency, if it could be done with current crystals, and you can measure their frequency accurately. It doesn’t address other criteria that might make one crystal better than another.
Michael
Yeah, the trouble is that modern designs like this one never seem to bother explaining how to do that. They just assume that the reader knows all the technical details.
What “really old books”?
For a long time “crystal filter” meant a single crystal. Very selective, but not great skirt selectivity. To make it wider, it was loaded down with resistors.
When SSB started becoming popular in the fifties, something better was needed, so the half-lattice filter became common. You needed crystals spaced a certain amount apart, and since one section usually wasn’t good enough, you needed two crystals per frequency. Initially these were down at the “classic” IF frequency, 455KHz. There were a lot of ft243 type crystals left over from WWII, and if you couldn’t find the matching pairs, you could open them up and grind them. Nobody had frequency counter, so everything was fairly vague.
Later, some used higher frequencies, in the shortwave range. But there wasn’t likely to be ground o precise frequency, since the move was to hc/6 type crystals. Though one guy from British Columbia was building direct conversion SSB transceivers with the crystal filter on the signal frequency, so they had to be exact; he unsealed the crystal cases and ground them, but still he had to start with crystals close to frequency.
In hobby circles, crystal ladder filters didn’t appear till the seventies, in European publications. But not much talk of crystal matching or individual specs. Back then, one was to make assumptions, and scale from the schematic.
Tat sort of thing didn’t appear in US hobby publications till the late seventies. I think the 1979 ARRL Handbook has half a page about crystal ladder filters.
I think it was the late eighties, maybe early nineties, when the whole routine was discussed, match the frequency, learn the specs, and then design from there. Obviously they’d be better filters, but some of it probably came from more calculators or computers, to do the math, and frequency counters being much more common. Also, as digital electronics became o much more common, crystals became a commodity, so instead of just being able to buy 100KHz and maybe a few others off the shelf, the variety became much wider. You could find crystal near here you’d want it, and they were cheap compared to custom ground crystals. Even when bought in quantity so you could find matches.
Michael
Or in the old XTAL holders, you could take the wafer out and grind it down to the higher freq.
Sharp, but this can get even narrower.
Has anyone considered simply thermally stressing the ones which are “nearly” right ie 20+ hours at 205C? this is a well known hacker method and if nothing else is also good for pre-baking crystals on critical projects to weed out ones likely to drift in the first 1K hours.
I’m just preparing to make one crystal ladder filter, which is supposed to pass 27 MHz, but greatly attenuate (it does not have to be perfect, 100x attenuation will be great) any frequency more than 10 kHz away. How many crystals would be required for this sort of selectivity? Based on my calculations, one would be enough. I don’t care much about the shape of the bandpass curve. Thanks.
Crystal filters are defined by bandwidth. The way you describe it, it sounds more like you need to knock out the other, which might best be done with notch filter at the offending frequency.
You do care about the shape f the bandpass curve since you are concerned bout the ignl 10KHz away.
Selectivity is “easy”, wider bandwidth is harder, and good skirt selectivity (so the sides of the filter are sharp rather than broadening out) requires cascading sections. The higher the frequency the harder it is to make a crystal filter, especially when the crystals may not be on their fundamental frequency up there.
Almost a hundred years ago Howard Armstrong solved some problems by inventing the superheterodyne receiver, where the incoming signal was converted to a fixed frequency, where it was easier to have gain and selectivity. So you wouldn’t have to have selectivity on the signal frequency, If the high selectivity is on the signal frequency, it has to be exactly on frequency, and if you need a different frequency, the filter has to be rebuilt.
Only exotic applications would put a narrow filter on the signal frequency.
That said, one simple method of crystal filters is if you have a multiple stages of amplification, you put a crystal across the emitter resistor (or cathode resistor if a tube). It will give most gain on the crystal frequency, and doing this a few times will improve the skirt. I’ve thought about doing this for a WWV receiver, since one if its frequencies is 10MHz, and crystals are easily available cheap for that frequency. And it’s an easy way to get selectivity with minimal effort.
Michael
Thanks for your reply. Believe it or not, it is a regenerative receiver with one stage of added RF amplification and 3 stages of LF amplification. It will receive low powered telemetry data (20 mW TX at the moment). I currently have issues with some CB interference. Voices sound like they are from the southern US, even though I’m in Ontario… This thing you mentioned about emitter resistor is very interesting, I will probably try that.
Keep up the helpful work and producing in the crowd!