Вставка
- Опубліковано 23 чер 2024
- Episode 617
I show my design for a radio IF bandpass filter made with four crystals. Some theory, some schematics, some design, some measurements. Here is the article I showed:
www.arrl.org/files/file/QEX_N...
Be a Patron:
/ imsaiguy - Наука та технологія
I would recommend using the Dishal SW from Steder and Hardcastle, the two authors of the paper shown in the video. The most straightforward filter to build is the G3UUR’s QER, which is also described in the user manual of that SW. First, learn how to derive crystal parameters Lm, Cm, and ESR from the measurements of crystals using either a VNA (like nanoVNA) or some other set up (see the G3UUR’s 3-dB method). Also, measure their package capacitance Cp. As others mentioned, start with a large sample of crystals to pick those which have serial resonance frequencies close to each other (within 1% of the desirable filter bandwidth). The SW will tell you the center frequency, the input/output impedance, and the value of the shunt capacitors, all baded on values of your crystals and the desired bandwidth and number of crystals. Then, you would have to design an impedance matching circuit to whatever your circuit expects (normally 50 ohm). Hopefully after a few trials, you would have a filter with reasonable ripple and side skirts.
My mother used to work for Rakon in NZ making crystal filters by hand back in the day..... sanding the crystals, upon discussing with her just now it seems that they should in fact be "centered" upon the middle point, the fact that yours aren't probably suggests inferior crystal filter quality! right from the horses mouth hope this was some interesting insight for you it definitely was for me.
There’s one thing you didn’t mention on the video. The filter impedance is not 50r, and you can get rid of the ripple by inserting an impedance transformer on the input and output... and yes, making this filters is a real fun... thanks for the info!!
Hi IMSAI Guy, basically crystals are specified with our without a load capacitance value. When specified with a load capacitance value, the load capacitance value is the external capacitance in the circuit that is required to cause the crystal to oscillate at your desired frequency of oscillation when used in a parallel resonate oscillator.
Therefore you can pick up 2 different crystals stamped 10 MHz but each will resonate at a different frequency depending on how much external load capacitance was specified for their use when originally manufactured. One may have been specified as a 10 MHz crystal with a load of 20 pf and the other may have been specified as a 10 MHz crystal with a load of 30 pf. The one specified for a 30 pf load will oscillate at a different frequency than the one specified for a 20 pf load when placed in the same parallel circuit oscillator unless you use the correct load capacitance that matches what the crystal was designed to operate with. The load capacitance specification is based on the crystal being used in a parallel resonate oscillator but many oscillators operate at the series resonate point (you can still use these crystals in a series resonate oscillator but will need to add some series capacitance to obtain resonance at 10 MHz). An example of a parallel resonate oscillator would be one built around a digital inverter and here is a link to a paper that shows a schematic of one: ecsxtal.com/store/pdf/Oscillation-Circut-Design-Considerations.pdf
P.S. the QEX article you mentioned is one I often reference. I do in fact measure my crystals using the G3UUR method mentioned in that article and I also use the DISHAL program mentioned in that article to design my crystal filters.
Feb 3, 2021 update: Today I took a 10 MHz 30 pf cut crystal and using a VNA in a very crude series test set up its series resonate frequency was below 10 MHz (9.996960 MHz), but when I installed a 30 pf capacitor in series with the crystal its series resonate frequency was 9.999980 MHz and when using a series capacitor of 10 pf its series resonate frequency was 10.003860. I also tested a 10 MHz 18 pf crystal and it measured 9.997280 MHz but when I installed a 22pf series capacitor in series with the crystal its series resonate frequency measured 9.999500 MHz (I did not have a 18 pf cap which would have put the resonate frequency closer to 10 MHz). I normally use my crystals in a series resonate configuration and therefore have to add a series capacitor to raise their frequency to what I desire.
ITS THE JOY OF IT, so don't ever stop , ,
08:06 They specify the one of the peaks. Depending on whether the crystal is intended for series (first peak) or parallel resonance (second peak). This measurement is done at a precisely measured load capacitance. So if it says 10 MHz and the crystal is intended for series resonance at 22pf load capacitance, that's where the first peak will be under that load condition. So yes, you can use the marking outside of the can, IF you meet the specified load and operating conditions. Now that said, they're never perfect and differ a bit, even if they're from the same lot.
Most Xtal are designed to be used in a specific circuit. If you order a crystal to be ground specifically for you, you will have to specify the circuit and the capacitance of the circuit. This is why you have some "strange" values.
Hey, this is really great, thank you!
Another piece of the puzzle. Thankyou.
I have several hundred 8.3886 mhz crystals in HC48 cases.
You didn't try adjusting the input / output termination. All measurements were made into 50 ohms (assuming the instruments are so terminated). Ripple and slope are dependant on termination, so either L networks or matching transformers are needed. You can use variable resistors in series with the filter to see how the termination effects this.
Damn, Is There Any radio Wave Video You Don`t Have ? ;D
WONDERFUL VIDEO, Loved It And Learned A LOT ~!
interestingly, I measured my first crystal filter on our VNA at work. It's no slouch: $100k and 16 test ports! Apparently you have to pay extra if you want to measure extremely narrow bandwidth devices, however. I couldn't make a usable measurement at all. I sort of want to try again as it was a few years ago and maybe I made a dumb mistake. That said, the nanoVNA did an extraordinary job of helping me with the final tune up of my crystal filter.
Great stuff, thanks!
Glad you liked it!
I'm not an expert in the fields of crystals but I believe that they spec the crystals such that as they age they with stay within the PPM number specified. Age of a crystal makes the output resonate slower over time.
df/f0 = -K(T-T0)^2 (always negative).
Therefore the initial set point when the new will be on one end of the PPM. Measure the same crystal in 5 years and it will be in the center of the slope and in 10 years be on other end of the PPM limit.
Thank you for this very interesting presentation.
15:25 To have an idea what happens in filters with multiple resonators one might start with two resonators. One resonator stores energy from the input power and decays it slowly over a time of many periods depending on its Q factor.
If it is coupled to a second resonator it also exchanges energy with the coupled one which has identical parameters in our case. With a loose coupling the passed energy is so small that the resonant frequency is not affected. This works up to the so-called critical coupling which is in effect not critical at all.
With tighter coupling the energy is passed over and back as it can be observed at coupled pendulums ua-cam.com/video/3c73d5GJ2V0/v-deo.html. With this passing forth and back (which is also getting faster with closer coupling) each of the resonators gets an amplitude modulation with complete phase inversion. In this way its frequency moves to both sides of the original resonant frequency by just the frequency of the energy transfer.
As a result, a filter curve is formed with a near rectangular shape with ripples at the top. That is just the principle, for more resonators it will be a longer story.
Thanks for your very interesting videos, I really enjoy them!, 8:10 perhaps the resonance is shifted towards lower values with respect to the quoted value due to parasitic capacitances/inductances of the probes/leads that you used in this test? Best!
The peaks you see come from an impedance mismatch. The filter is not destined for 50 ohm. Once the impedance is correctly matched they will disappear
Yes, I was going to comment on this also. Normally filters are not designed for 50 Ohm I/O but something a bit higher like 500 Ohms maybe. You just design matching networks on both ends. Typically this would be a split capacitor where the in or out is at the tap point (like tapping down an inductor). Lots of crystal design software out there will have this as an option as you are specifying the filter parameters.
yes i made exactly the same comment in his LC inductance measurement video yestersday. The ceramic filter manufacturers usually specify the input impedance. You will need a matching L- or T-network for both input and output of the through measurement.
Thanks for adding marker->ref level
it gives a nice picture but here is my use case:
peak search (carrier)
marker->ref
now all other signals are dBc
it needs to be at the top of the display
would be nice if labels on the Y-axis are dBc and not dBm ( zero at the top)
HP decided never to label the Y axis as they thought SA is more of a relative measurement and not absolute.
I suppose I can just mentally ignore the extra Y division at the top and ignore the labels.
There's also other designs of xtal filters that use some of them diagonally-connected, beyond stagger-tuning. A bit beyond me. I must admit I'm lazy type and buy.
There's another great IF filter much cheaper than crystal and that's Ceramic filter + Mechanical filter combination, one for selectivity the other for stop-band and used together a rectangular passband, great for NBFM.
Thanks for the presentation.
Homebrewing an SSB TRX requires building a ladder filter and we used to buy a bunch of 10 MHz crystals from the market and choose the ones with frequencies close to one another on a 8 digit counter. The 8th digit will always be jumpy. We adjust the LO frequency to get 7 MHz and we had live with whatever the audio we got...hi..hi..
Adjusting the input and output capacitors or the shunt capacitors didn't bring about different changes in the pattern, I can see.
And it was not a nice single sharp peak at all there.
I want to know the input/output impedance of the filter, ie, if it's 50 ohm.
There are so many questions on the ladder filter; it has been a mystery.
At 1.40, instead of intermediate frequency you said internal frequency.
At 11.14 you said, parallel resistance instead of parallel capacitance.
At 26.26, you said, input and ground and input and ground instead of output and ground.
Let me read about the filter (about 3 decades later) and come again.
Thanks again.
I had sold my 4 inch NanoVNA but then found out that my new V2 couldn't do a really narrow sweep like you'd need for a crystal filter so I've ordered a 2 inch model which is supposed to do the job.
The crystal manufacturer assumes you'll be using a 10pf load. Any amplifier you make that's designed to oscillate a crystal will have to overcome the input capacitance of the amplifier. Since the input capacitance can vary, the worse case scenario would be 10pf. Since most oscillator amplifiers are designed to have the lowest input capacitance you're expected to pad in more with a trimmer capacitor. Your test setup technically doesn't have an input capacitance (I assume you neutralized the signal before you tested the crystal).
You've gone n done it ay' you!
A project I'm going to make for my ssb, when I looked I choked at how much they cost so doing it the home brew way lol.
My icom uses 9.0115 hz I found on ebay a bag of 50x 9mhz for like £5! Just gotta find some of the high running ones and out of 50 I have a chance lol
Good luck!
Well I got them today, shame none of them are on the "high" side, nearly every single crystal is near spot on and some are lower!
Nearly every crystal as me just off the lowest spot for the resonant frequency I need, 9.0115 don't feel that far away from 9.000 but what a shame lol
I don't suppose you know a way to drag it over? A mix of inductance perhaps?
As I have so many I'm going to have a play and open the can and "grind" it and see what I can do, it's all good fun lol
This was very interesting indeed! I wish you had finally put the commercial one on thje analyser and showed us a comparison. I am very curious now how big a difference there is between a commercial (mass produced) and a custom built and tuned filter. Did you ever test the commercial one yourself?
I'm sure I did. But many years ago now and the commercial one is soldered in the radio. They were not apples to apples. I put in a narrow CW filter and this one was more of a narrow SSB one.
@@IMSAIGuy ok then nevermind. 🙂
All crystals are specified with their defined burden capacitance to be found in the data sheet of the part. Typical values are between 15 pF and 25 pF. Would this solve the issue?
What if you put parallel L with each Crystal? Around 68uH.
Great video. I have only Nano VNA, I don’t have a spectrum analyzer.
Comments:
- the band pass frequencies is lower than the specified IF frequency, maybe it is good for LSB. Actually it depends on the radio design. For instance if the carrier frequency is 9.0015 MHz and the filter center frequency is exactly 9 MHz with 3 kHz BW then the filter is acting as an LSB filter. Or the other way around, if your filter if centered at 8998.5 MHz and your carrier is 9 MHz then it is an LSB filter. Sorry to put 9 MHz for easy calculation😊
- Do you think the source and load impedance will vary the characteristic of the filter? I think your spectrum Analyzer is 50 ohm input impedance. I have read some filter must be loaded with the impedance ranging from 1 kohms - 2 kohms
People tell me the loading is closer to 500 ohms. While that will change the out of band some I don't think the actual peak would move but I'm no expert.
@@IMSAIGuy thank you 🙏
the reason is they specify a load capacitance for the frequency
About the four peaks:
The crystals will load the other crystals. They have their own capacitance and they'll interact with each other.
you might like this video: ua-cam.com/video/oMFalh3MDS8/v-deo.html
Is using the smaller form-factor, still through-hole crystals still good or should we use the full-height ones?
they are all the same but different. the Q-factor might be better with a larger crystal, I'm not sure.
Hi i am studying for my amatuer radio ticket, and in the book it shows the top peak as being parrel residence and the bottom series residence, but in this video your saying its opposite, am i missing something.
what is the y-axis of the plot you are looking at. I've seen many plots of the y-axis being reactance. the measurement I show is amplitude of S21. Please use the information in your book as that is what you will be tested on.
The square root of (8.827 * 8.837) = 8.832
I'm no expert, either, but I've learned that crystal filter construction is not a trivial task. It's not complex, but it requires care and no skipped steps in the process. It is not a 30 minute task.
I think there were 2 problems with those sharp peaks of 10 dB or so in the crystal filter passband you saw on your analyzers - input and output impedance matching of the filter, and the need to more closely match the crystals to start with. It takes a lot of crystal sorting to get a really good, close match, generally as many as 10 crystals from the same lot for every 1 you use to make a quite good filter. So, for a 4 crystal filter, 40 crystals tested. And to do the sorting you will need to build a good crystal test jig with 50 Ohms in and 50 Ohms out, as well as 4 to 1 impedance matching transformers in and out to get down to the crystal motional input impedance of roughly 12.5 Ohms. There are a couple of online references about how to do it. It really boils down to the fact that the crystal testing is fussy, and once a filter is built, variable capacitors can't fix everything. They can only optimize a filter that has already been built with crystals carefully matched and optimized for the expected circuit load impedance.
Some circuits will need 200 to 1500 Ohm filter , and that could be another problem to deal with. Perhaps having to buy and test all these crystals is why commercial filters cost so darn much.
I've been studying up on the subject because I was lucky enough to snag well over a thousand 12 MHz crystals from one manufacturing lot, so now the fun begins. :)
I have this video if you have not seen it: ua-cam.com/video/oMFalh3MDS8/v-deo.html
@@IMSAIGuy , thanks. I watched it a few weeks ago, but I'm watching it again. I'm sometimes not too bright, and I find it helps if I watch the tech videos 2 or 3 times. I think you said you spent years working at HP. I spent a couple of years there in Colo. Spgs, but left for better pay. But I sure did enjoy the company scrap sales.
Hi,
Really strange adjust only on the right side...thanks
Begs the questions, how does your analyzer get a BW resolution filter of 300 Hz in your demo? Just a fancier crystal filter topology?
in many designs, the instrument down converts to lower frequencies where it is easy to have narrow filters. 300 Hz is hard at 21 MHz IF but if you down convert that to 455 kHz then even a 30 Hz filter is easy.
How well did this filter actually work? Especially in comparison to the Kenwood filter. Thanks for the video, Barry, VK2FP/AG7VC
It worked OK. It was off center as I said so that was a pain. It was a bit too wide and I replaced it when I found a cheap Kenwood filter on eBay. It was great fun though and proud to have a home made one.
Crystal Marked Value Is Actually Simple To Understand, They Mark The Resonant Frequency Value Of The Crystal On The Slope That Goes Down From The First Positive Wave And It Can`t Be Lower Then The Neutral Wave And It needs To be "Back In Time" (Becaause Of SOme Calculations And Laws) From The Middle/ Brake And It Can`t Be Highter Then + - 5% Of The Slope If You Invert It And Resonate It So Imagine If The Both Up And Down Would Be Just resonating, So The Dot From Where The Value Comes Is A Strange Point From The Full resonant Wave. You Showed Prety Correct Where It is And I Told Why Its There, But Im Tired And Sick So Maybe My Explanation Is Muddy, Sorry of TThat If Thats The Case
Rigorously?