Announcement

Collapse
No announcement yet.

Buying custom crystals...

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

    Buying custom crystals...

    Dangit. I keep on running into the same problem...

    Seems for radio use, crystals seems to be a problem. To get a specific frequency out of a radio, I need a specific crystal based on how the circuit is designed.

    Well, I thought I was being smart and tried to design a synthesizer to get around needing these weirdly cut crystals, but I keep coming back to needing weird frequency crystals to simplify the design.

    It seems like in the past, getting custom cut crystals is easier or something? Or perhaps it's typical to go to a crystal company to get specific crystals and have to order thousands of them? Or not?

    Anyway, last night I was reading and found a solution to simplify my nasty PLL problem: divide by 277.7777 (which I did find a digital solution to come close enough to get 0.002% error. Unfortunately it is big and unwieldy, I would likely have to build it into a PLD to simplify things if I find a PLD that would run fast enough. It was unfortunate that simply dividing by 278 was too large of an error.)

    However the solution required a single crystal of 4.551111 MHz (or an even harmonic of that)... except these too seem difficult to find and find cheaply. Any ideas?

    #2
    Re: Buying custom crystals...

    You are getting into a hobby that was for many decades built upon surplus (both commercial and military) and salvaging other electronics for working parts or actually build the needed parts yourself. Now the hobby is an appliance hobby; when it breaks you just threw it away and buy a new one. You should be able to find crystals of 455 Khz in some common case styles just have to know where and how to look for them. Salvage old equipment for a goldmine of parts. 4.000 Mhz crystals are plentiful as well as 4.5 Mhz the latter being used in lots of old tv sets for the sound IF stage.
    Last edited by Sparkey55; 04-24-2018, 12:42 PM.

    Comment


      #3
      Re: Buying custom crystals...

      I keep telling him to learn to grind crystals, like back in the day, but eh.
      Things I've fixed: anything from semis to crappy Chinese $2 radios, and now an IoT Dildo....

      "Dude, this is Wyoming, i hopped on and sent 'er. No fucking around." -- Me

      Excuse me while i do something dangerous


      You must have a sad, sad boring life if you hate on people harmlessly enjoying life with an animal costume.

      Sometimes you need to break shit to fix it.... Thats why my lawnmower doesn't have a deadman switch or engine brake anymore

      Follow the white rabbit.

      Comment


        #4
        Re: Buying custom crystals...

        No, this could not be made from surplus - they had to be explicitly made. The mathematics do not calculate out - I can't actually use random crystals, nor could the original designers. As goontron pointed out, manually cutting crystals is possible, this may be the only option - and could be done by the crystal manufacturers of course.

        As for the crystal suggestions:
        455KHz (AM IF) is too low in frequency. My target frequency is exactly 1111.1111 Hz, and 455KHz would get me close to 111.1Hz but my PLL feedback loop would go too slow. It also is 0.02% error, which is "getting there" and worth a shot to at least try, that is, if the PLL feedback loop was fast enough...
        4.000MHz (general timer) has a huge error and cannot use that.
        4.500MHz (TV sound carrier sideband offset) also has a huge error and can't work around that. 1.123% error.

        Going back to using the 278 divider as an option to do away with special crystals (and use a 1.000000 KHz reference which is easily derivable from a 4.000000MHz crystal), this is 0.08% error and this too is also too much (though I was able to design (on paper) a 277.77 divider, which just about hits the target.)

        My target error is 0.002% and I don't think I can grind with this precision. Seems that 4.5MHz crystals sacrificed to 4.551111 MHz is the best option for grinding however.

        BTW, modern equipment don't use these weird crystals because they're direct driven or only use stages of 2 because...well... yes, computers like 2 and not 3...
        Last edited by eccerr0r; 04-24-2018, 01:30 PM.

        Comment


          #5
          Re: Buying custom crystals...

          Come to think of it, I really need to be precise with the divider, might even need to get beyond the 0.002% (BTW, this is 20PPM right?) because whatever crystal I have, that too has an error and the errors add up...

          Would be best if somehow they cancel each other out but the divider would always sway one way and the crystal error could go either way. What I don't understand, even 20PPM is a fairly large error. At 144MHz, 20ppm error is almost 3KHz error, how could this possibly work...must need a cap tweak or something!

          Ooh this is annoying!

          And oh no... Maybe I'll even need a TCXO for this!!! Aaaahhh
          Last edited by eccerr0r; 04-25-2018, 03:41 AM.

          Comment


            #6
            Re: Buying custom crystals...

            FFFFFFFFFK this PLL... More tradeoffs

            New crossroads: After thinking about it, the problems are disjoint.
            Problem 1: Hitting the right frequencies on target
            Option 1A: 1111.1111Hz reference, requires special crystal, no fractional divider.
            *Option 1B: 1000.0000Hz reference, uses common crystal, requires fractional divider

            Problem 2: PLL frequency step size
            *Option 2A: Selectable prescaler - minimizes high speed logic (50MHz), but no fine tuning at the high range.
            Option 2B: No prescaler - (fractional) divider runs at full speed (50MHz), but allows 5KHz step throughout the whole range (less than 5KHz, I think I will need to resort to a trimcap). This isn't a problem as both high range (for receiver) and low range (for transmitter) have different multipliers in the radio itself and both will end up with a 5KHz step.

            I definitely will need a trimcap for the oscillator.

            This project is tougher than I thought, then again I've never designed a PLL before (and I am also dreading the LPF feedback filter response which I haven't even touched yet)... and the more I look into this I want to see it implemented, even if it doesn't quite work!

            I'm now at 10 or so ICs now *without* the VCO... this is just the digital section and I'm cheating. I do wonder how much jitter there is in the microcontrollers if I use it as a software controlled divider...

            - Microcontroller with two timer/counters, both are used for the PLL (AVR Tiny2313?)
            Unfortunately I think I have to shut off the PLL to write to the EEPROM because of the software refclk ...eeek!!! I don't know how bad jitter is in a microcontroller software controlled output, even if I make this the highest priority interrupt.
            - MAX232 to allow interfacing.
            - 1x74ACT161 for the prescaler (booo... no 74ACT92)
            - 4x74HC161 for fractional divider. But maybe I don't even need the prescaler because the 74ACT161 is fast enough to run at full speed!!!
            - 2x74ACT00 for multiplexer and fractional divider controller
            - 1x74ACT04 for logic inversions...
            - 1x74HC4046 phase comparator (NOT using VCO in this thing...)

            Probably would be great if I could stick all but the microcontroller into a FPGA or CPLD... if it will run fast enough...

            Using a DDS is starting to look a lot simpler, though designing a fast DDS (without simply buying a DDS chip) is equally as tough.
            Last edited by eccerr0r; 04-27-2018, 12:53 AM.

            Comment


              #7
              Re: Buying custom crystals...

              I think International Crystal Manufacturing Company in the States is the only outfit left that makes custom crystals. I used to always get mine at West Crystal in B.C., then they quit making them like 8 years ago. With that I refused to build crystal based repeaters for new sites.

              Comment


                #8
                Re: Buying custom crystals...

                There is a thread here with suppliers of custom crystals with oddball frequencies.
                Instead of copy pasta, https://www.eevblog.com/forum/repair...87/#msg1486587

                Comment


                  #9
                  Re: Buying custom crystals...

                  Well, buying an $expensive$custom$crystal isn't exactly what I want to do, so I'm killing myself with designing a funky PLL instead and using a common crystal... Currently trying PoC of the weird divider in a 22V10 PLA, and I'm still pulling my hairs out getting this thing to count properly... Idea is to hopefully use the PLA to save a couple of downstream counters that don't need to run at 50MHz (but the logic of this portion of the prescaler by itself (not including the control logic) is apparently overflowing the 22V10...ugh!)

                  I'll need to do some timing analysis to see if the 22V10 in my spare parts box is fast enough or I'm stuck with buying faster gates; not sure if my LS161's are fast enough even.)

                  Then again...those cardinal components field programmable oscillators look interesting... if it could be continuously programmed instead of just...twice... Also I wonder if it goes down to 1Hz resolution or not...

                  Comment


                    #10
                    Re: Buying custom crystals...

                    Originally posted by eccerr0r View Post
                    Well, buying an $expensive$custom$crystal isn't exactly what I want to do, so I'm killing myself with designing a funky PLL instead and using a common crystal... Currently trying PoC of the weird divider in a 22V10 PLA, and I'm still pulling my hairs out getting this thing to count properly... Idea is to hopefully use the PLA to save a couple of downstream counters that don't need to run at 50MHz (but the logic of this portion of the prescaler by itself (not including the control logic) is apparently overflowing the 22V10...ugh!)

                    I'll need to do some timing analysis to see if the 22V10 in my spare parts box is fast enough or I'm stuck with buying faster gates; not sure if my LS161's are fast enough even.)

                    Then again...those cardinal components field programmable oscillators look interesting... if it could be continuously programmed instead of just...twice... Also I wonder if it goes down to 1Hz resolution or not...
                    If you want a real challenge try the 1750 Meter Band.

                    Comment


                      #11
                      Re: Buying custom crystals...

                      ^ Nah, 2200 is where its at. And its rockin' from what ive heard.
                      Things I've fixed: anything from semis to crappy Chinese $2 radios, and now an IoT Dildo....

                      "Dude, this is Wyoming, i hopped on and sent 'er. No fucking around." -- Me

                      Excuse me while i do something dangerous


                      You must have a sad, sad boring life if you hate on people harmlessly enjoying life with an animal costume.

                      Sometimes you need to break shit to fix it.... Thats why my lawnmower doesn't have a deadman switch or engine brake anymore

                      Follow the white rabbit.

                      Comment


                        #12
                        Re: Buying custom crystals...

                        I'm technically not that interested in long wavelength communications that much, mostly because it's extremely limited in the number of "channels" available before they start running into each other. Granted trying to receive WWB would be interesting with off the shelf parts - it's at 5KM or so... A lot of the time it depends on the environment and not the electronics, anyway.

                        That stupid fractional divider I have, I already have to do ridiculous tricks to get to fractions of a Hz because of the freaking multipliers in the radio...

                        Comment


                          #13
                          Re: Buying custom crystals...

                          Originally posted by eccerr0r View Post
                          Well, buying an $expensive$custom$crystal isn't exactly what I want to do, so I'm killing myself with designing a funky PLL instead and using a common crystal... Currently trying PoC of the weird divider in a 22V10 PLA, and I'm still pulling my hairs out getting this thing to count properly... Idea is to hopefully use the PLA to save a couple of downstream counters that don't need to run at 50MHz (but the logic of this portion of the prescaler by itself (not including the control logic) is apparently overflowing the 22V10...ugh!)

                          I'll need to do some timing analysis to see if the 22V10 in my spare parts box is fast enough or I'm stuck with buying faster gates; not sure if my LS161's are fast enough even.)

                          Then again...those cardinal components field programmable oscillators look interesting... if it could be continuously programmed instead of just...twice... Also I wonder if it goes down to 1Hz resolution or not...
                          Did you actually call them and ask for a quote on one? I mean some of us ham guys are really cheap and we are still getting crystals from there.

                          Comment


                            #14
                            Re: Buying custom crystals...

                            I was seeing some places asking upwards $10 or so each. Then again, it may still be unobtanium.

                            The programmable oscillator are around $5 each from Digikey.

                            I was thinking $2 as an upper bound, this whole ordeal was meant to be a learning experience and I figure I learn more trying to work around this problem...

                            ... and I forgot the decoupling caps on my divider (hey, why am I seeing those spikes triggering my scope!)

                            Comment


                              #15
                              Re: Buying custom crystals...

                              That 22V10-25 replaced three of the 74x161's which is the guts of prescaler#2... This helps out the wiring glut. The three 74x161s would have been a bigger divider, but ideally I divide as little as possible here and have the microcontroller do more.

                              I found a 16V8-15 that may be fast enough for the high speed stuff that must run at 50MHz (144MHz to 148MHz of the ham band, divided by 3 is below 50MHz; though it would be interesting if I could get it to go even faster to 160-165MHz as well as slower to 88-108MHz to at least hear something instead of waiting or hunting for ragchewers). Now I'm seeing if I can replace the clockmux, prescaler#1, and the control logic for prescaler#2 with the 16V8-15.

                              Unfortunately it looks like I may need two separate PLA's for these functions because prescaler #2 needs to run off a different clock and these crappy PLAs do not support split clocks... It may very well need to be separated to two anyway, I don't think the 16V8 holds enough state. I'm already one state above my prediction to help timing - the 22V10 I predicted using 9 bits, but actually needed all 10. The remaining predictions are 3 bits for prescaler#1 and 4 bits for the prescaler#2 controller. The 16V8 would have only one output/flop left if I could squeeze everything there, but probably best to split to two PLAs.

                              Dang need to stop rambling... but excitement of design.

                              Comment


                                #16
                                Re: Buying custom crystals...

                                Oh so strange... I found a 48MHz crystal oscillator in my parts collection. So, I tried feeding it into my divider... IT worked! Interesting. Not sure if I lucked out with this -25 part or not, or why it didn't seem to work earlier today.

                                Now that I actually have a 48MHz oscillator, I need to see what happens if I feed its output straight into the radio - see what the multiplier does. Unfortunately this apparently is outside of the ham band so I have no idea what I'd be listening to, or whether the multiplier will actually go bonkers with it. Supposedly the frequency would calculate out as 48001.1 KHz (as measured by my freq counter)*3(multiplier)+10700(IF) = 154.703.3KHz ...

                                So, I guess using a 48MHz is actually outside of ham band, so I don't need to go up that high as a minimal design constraint. But all the better, who wouldn't want more frequencies, just need to make sure nobody transmits there. I also suspect this radio to do something unexpected with the signal due to the multiplier, but maybe not... I'm not sure if this radio was specifically for amateur radio, or whether it was designed to be put into aerial or maritime craft. Perhaps it was tailored for these bands as to why it "sucks" for amateur radio as I hear.

                                On another note, I just remembered that I have a broken VCR that I need to dump... Guess what I'm stealing from the VCR.

                                VARACTOR!

                                Maybe more, if there's a chip in there or other parts I could use for a VCO. Unfortunately it was designed for VHF/UHF versus VHF only, then again it's probably direct PLL, I'll have to redesign it. It too has an upper and lower VHF band though for different reasons (due to the air, marine, broadcast FM, AND ham bands sitting in between channels TV6 and TV7).

                                Comment


                                  #17
                                  Re: Buying custom crystals...

                                  The 154.7 Mhz is in the Public Service Band, think Fire, Police, etc. You can not transmit in those frequencies.

                                  Comment


                                    #18
                                    Re: Buying custom crystals...

                                    Is this the radio that has the 10.245 Mhz crystal in it? If so 10.7 minus 10.245 equals 0.455 Mhz.

                                    Comment


                                      #19
                                      Re: Buying custom crystals...

                                      https://www.kenselectronics.com/lists/cbxtal.htm

                                      read the notes about ham crystals near the page bottom.

                                      Comment


                                        #20
                                        Re: Buying custom crystals...

                                        Yes, Kenselectronics is that one page I saw that sold crystals explicitly for the Regency and calls it out by name (search for hr2b) and is the site that I said "NFW I'm playing $30 for a set of crystals they don't even have. IT'S PLL TIME!"

                                        And of course it's not legal to transmit there or 102.5MHz or anywhere at all, for me - even 146MHz - since I'm not licensed (yet). However with the PLL if it reaches 48MHz, it should at least be able to listen there.

                                        I don't think it will transmit or receive correctly outside the designed band, however - the multipliers will crap out on the signal if it's too far away from the design band for both. Point is moot because it still would generate garbage and counted as interference, thus illegal. The frequency counter and dummy load will be my friends, not my oscilloscope (though it is also needed for debugging).

                                        It would be awesome if I could tune the receiver all the way from 88MHz to 162MHz so I could at least hear FM Broadcast and NOAA weather as well...

                                        And yes, there are many VHF/UHF radios that have double conversion superheterodyne for a reason and hence the 10.425 MHz crystal. However it has absolutely no bearing on what crystals are needed to do the actual tuning.

                                        Comment

                                        Working...
                                        X