EMRFD Message Archive 15332

Message Date From Subject
15332 2019-06-07 17:31:39 Vern VanZ Crystal, crystal...on the wall...
Oh wait...I got that a bit off...

Anyway, I'm wondering if anyone here is still building crystal filters (ladder type), and if so, would love to know where are you sourcing your crystals?

The last time I tried this (a few years ago), I purchased about 100 units from Mouser, but the quality was so poor that I only found (maybe) a half dozen that were close enough for 'minimal' performance (wide 3db passband, wheelchair ramps for skirts).  Certainly not good enough for CW...  Nice steep skirts like grandma wore back in the roaring 20's (when she was still 20 and single), would be nice, but maybe that's a thing of the past...(erm, the filter's skirts, not grandma's...).

Any thoughts?

Thanks, 73,
Kennewick, WA.
15333 2019-06-07 21:11:32 Ashhar Farhan Re: Crystal, crystal...on the wall...
I have been busy building ladder filters for some time. With the luxury of having a choice, I probably use the worst of the lot. Anything that I can get from local hobby shops. They are so cheap that I am sure they are rejects.
The trick is in four tricks - 
1. Sort. Take the entire bunch sort it frequency wise into bunches that are within 20% of the bandwidth. While you are plugging each crystal into your oscillator, watch for those with low output, toss them out.
2. Now, measure the motional parameters. Don't use the motional parameters of just one crystal from the bunch, measure more, find the average, select those with near match of motional parameters. 
3. Design and build. 
4. Measure again. This time plot the transfer function.

What you lose, in broad terms, is the attenuation. Worse crystals have greater loss, but you can almost always get the bandwidth you shoot for. Skirt may vary.
The other trick in the bag is to make multiple filters with just 2 or three filters with buffer amplifiers in between. Jeff Damm makes those, he doesnt say so, his filter pictures do.

73, f.

15335 2019-06-08 08:46:38 Alan de G1FXB Re: Crystal, crystal...on the wall...
It never gets mentioned, So much so I doubt it's existence these days.......

I'm certain we spotted a gotcha....
??when they introduced the reduced height crystal packages it required a different cut of the crystal, that changed it's properties.
(Not talking Fundamental / Overtone properties.)

If I had a choice I would choose full height, through hole HC49.
Even HC6 if I wanted something extra special, and could find them. (Miss Daisy)

regards Alan

15336 2019-06-08 11:02:56 chuck adams Re: Crystal, crystal...on the wall...

You are right. 

The HC-49S vs HC-49U question seems to be in the background.

I mentioned it in one of my videos on YouTube that the HC-49S (the
small form factor) is taking the place of the HC-49U.  Two factors
here (IMHO) are due to the Chinese and others going with reduced
costs and smaller amounts of crystal material required.  The lower
profile also seems to be desired for space issues with modern
designers and board layout desires in manufacturing.

The S form factor crystals operate in the flexure mode while the
U form factor crystals operate in the shear mode.  The latter gives
higher Q.  I have yet to get around to studying how much you give
up in pulling the crystal off frequency for VXO operation.  Maybe
by the end of the year.

I have not tried to get a quote on 1,000 crystals lately for the
two form factors, but I imagine it is a significant cost differential.

Humbly submitted,

chuck, k7qo

15337 2019-06-08 11:23:30 w7zoi Re: Crystal, crystal...on the wall...
Hi Chuck, and gang,

This question of the tiny crystals versus the larger HC-49 parts is one that has been discussed often.    It has come up on this GROUP several times.    A couple of years ago I wrote a piece that addressed the issue.    It ended up on my web site as a PDF.  See http://w7zoi.net/xtalmod.pdf     Bottom line is that the tiny crystals are still AT cut, just as we have always had.   However, the parameters (Lm, Cm, EST, C0) are much different than what we have used for other designs.

It does us little good to write about a problem in this forum without also including data.   Rather than talking about a experiment not working, we need actual numbers.   What was the spread in center frequency?   What was the unloaded crystal Q?   Were measurements even done, and if so, what were they?    Once we have actual data, a design can be done.    Of equal significance, we can tell if a viable design is even possible.    Sometimes it is not, but that is still covered by the fundamental theory.

73, Wes

15338 2019-06-08 11:48:53 jim_amos_n8cah Re: Crystal, crystal...on the wall...
Hi Vern,

I would agree with Farhan completely with regards to measuring and sorting, except I use a bit higher end method of sourcing my crystals.    I enter all of my crystals into an Excel spread sheet and use that to sort, group, and create the average parameters for designing as Farhan mentions.

I'm working on a CDG-2000 variation as a long term project and wanted to use a 9 MHz IF.    The CDG-20000 design uses discrete crystal filters for the post mixer roofing filter, and the IF Noise filter.   The original crystals used were no longer available but Mouser sells 9Mhz crystals that are rated at 10 PPM at the Industrial temperature spec from IQD.  I bought 50 of these and, like Farhan, measured all of the crystals, this allowed me to easily group the cystals for building different filters.  Since I started with relatively closely grouped crystals, I was able create groups that were within 20 - 30 Hz of each other for making different filters.   I've measured the average Q of the crystals at about 127000, not bad, but better crystals can be found, which helps with attenuation and shape factors.  

One factor in deciding to go ahead and try to use the IQD crystals was that I saw someone else using them successfully.   Otherwise, I agree that finding crystals for filters is a bit of a crap shoot.

The IQD have worked well so far for the roofing filters, and also a CW filter.   I will say though that those crystals are not the bargain crystals that everyone is hoping to build crystals out of.   They are $1.22 each, or $0.78@100, they were $.83 when I bought them, so they've gone up a bit. 
15339 2019-06-08 12:43:44 Vern VanZ Re: Crystal, crystal...on the wall...
Hi Wes and group,

Sorry to all if I've started something here, or inspired any unpleasantness. 

I looked for the measurement data (excel file) that I generated for the 10 Mhz filter I attempted to build (around 2016 I think), but haven't found it yet.  The lot was 100+ HC-49, "+- 20 ppm" (so says the bag's label from Mouser), and I'd be willing to measure them again, record the data and make available if folks think this would be an adequate group size.  I've been out of touch for too long on this, so do need to take stock of my test fixtures, etc.  But I'm willing to help answer the question...

73, N7GTB
Kennewick, WA.

15340 2019-06-08 14:37:37 Vern VanZ Re: Crystal, crystal...on the wall...
...just found the spreadsheet I built way-back-when of the 10 Mhz crystals... It's in .ods format, but should be compatible with either Microsoft or open source products.  For what it's worth, it's attached. 

DISCLAIMER: There's been a lot of water under the bridge since I did these measurements, so please be understanding if I don't have a lot in the way of detailed answers to questions.  For instance, I don't recall now if I used the KA7EXM digital power meter, DDS VFO, and HP 8514A counter combo; or my NAT-SNA to measure them.  I tend to think it was the later and not the former...Again, sorry.  Too much time has passed...

15342 2019-06-09 06:41:00 Nick Kennedy Re: Crystal, crystal...on the wall...
That's certainly interesting about the relationship between thickness and Lm & Cm and how that relates to HC49S (short) versus HC49U (standard height). I had some measurements that show this effect fairly clearly but at the time I didn't have an explanation for the difference.

Specifically, a group of 4.9152MHz HC49/U crystals measured Lm = 91.35 mH average and two groups of HC49/S crystals for the same frequency averaged 233 mH and 313.5 mH.


Nick, WA5BDU

15344 2019-06-10 08:12:53 Vern VanZ Re: Crystal, crystal...on the wall...

I hope this isn't a stupid question, but do you recall if there were noticeable difference(s) in Qu, series resistance, etc between the two (hc49/s vs hc49/u)?

73, N7GTB 
15345 2019-06-10 08:26:01 Vern VanZ Re: Crystal, crystal...on the wall...
...sorry, please read that as "significant measurable difference(s)"

15346 2019-06-10 11:46:29 Nick Kennedy Re: Crystal, crystal...on the wall...
Well, my stupid non-question was that it didn't even strike me as significant that the Lm for the short crystals was two to three times larger than for the normal size HC49/U crystals or how it might complicate filter design. Until I read Mr. Hayward's paper.

The Q measurements were pretty good, compared with others I'd measured. Samples from two mfgrs of 4.9152 MHz crystals were 250,000 to 350,000.  This use of crystals was for the ATS-3 transceiver. Use of the short crystals was no doubt related to the challenge of putting a sophisticated transceiver into an Altoids tin.

I also used HC49/S crystals in a "measurement receiver" I built to use with PHSNA. I had a bag of 100 5.52 MHz crystals so I wanted to use them for the filter. The Qs were in the 150,000 range after I'd cherry picked the good ones. LADPAC said the Q was too low for a 200 Hz filter so I bumped it up to about 400 Hz. The response curve looked pretty good but maybe the ultimate attenuation of about 55 dB wasn't the greatest.


Nick, WA5BDU

15347 2019-06-10 15:14:01 rogertraylor Re: Crystal, crystal...on the wall...
I just went through 10 HC49U, and about 40 HC49S crystals.  These were 4.915mhz units. The tall ones
were fresh from Mouser.  The short ones were at least 10 years old and I got them in a bunch of parts from
a ham who is now in a senior facility. Oddly, the tall crystals on average had lower Q, higher Rs values.
The tall crystals were fairly consistent with regards to all parameters however.

The short crystals were all over the map with regards to Q, Rs.  About 8 of them had Q's less than 80,000.
I discarded these. Many were fairly good with Q's in the 150k range.  I selected a group of 24 crystals
with Q's often exceeding 200k. 

Here is some data done only from rough eyeball average of the values. I haven't had time to enter
all this data into a spreadsheet.

Tall crystals:
ABRACON +/-20ppm, 4.9152Mhz, part no. AB-4.9152MHZ-BZ
       Fs            Fp             Q       Co       Rs        Lm           Cm
4.913615   4.925071  170033  3.7pF  10.9   60.2mH    17.4fF

Short crystals:
Marking on the top was 4.91ECSR, no other info on these parts.
       Fs            Fp             Q       Co       Rs        Lm           Cm
4.914873   4.922297  200111  1.6pF  32.2   217.2mH    4.8fF

Measurements were taken with a SARK 110 analyzer in its crystal parameter mode.  I have no idea of the
accuracy or correctness of its readings but at some point intend to see how it compares to methods used
in EMRFD.  I made these measurements mostly out of interest. Using these values in a SPICE simulation
yields very predictable results. The SARK 110's data was also quite repeatable over several days and
across several crystals.  So, if its readings are bogus, they are consistently bogus!

Hope this is helpful.

Roger Traylor
15356 2019-06-11 08:42:57 Nick Kennedy Re: Crystal, crystal...on the wall...
I did a few more measurements. Those Q values I reported nagged at me as maybe a bit high and my measurements were done several years ago. (The way we do it now is always better, right?)

Of course we're discussing the effect of size on Lm & Cm, not Q. But I was still surprised at the Q values of the short crystals.

I retested some short and standard height 4.9152 MHz crystals with this result:

For the short crystals:
Lm average: 183.3 mH. Q average: 204,733

For the standard height:
Lm average 63 mH, Q average 162,129

Comparing the short Vishay-Dale crystals with standard height FOX units:

Lm of HC49S is 2.9 times higher. Q is 1.26 times higher for short crystals

I also retested some of my surplus 5.528 MHz HC49/S crystals, FWIW, but don't have standard height ones to compare with:

Lm average = 120.8 mH, Q average = 144,300

I use my PHSNA system for measurements. It seems to compare fairly well with other systems on my workbench and at one time in a "round robin" test where crystals were mailed from ham to ham and results shared.


Nick, WA5BDU

15358 2019-06-11 10:25:21 Vern VanZ Re: Crystal, crystal...on the wall...

Not surprisingly I also found room for improvement... It seems my crystals (when properly measured) have smaller -3db bandwidth, series resistance and much higher Q.  I also 'played' with the data sort functions in Excel, and found 7 (5 of equal center frequency, and two that were only 3 Hz away...). 

Making changes (improvements?) to my resistance measurement fixture resulted in values less than half of what I had recorded earlier. Using my (now ancient) KD1JV DDS/VFO, KA7EXM PIC Power meter and HP 5314A counter setup, I also found that the bandwidth was significantly narrower than what was originally reported.

So while my Qu figures are (so far) still turning out to be less than 100k, making the above changes resulted in higher Q (more than double). I can still improve this, but requires building a new fixture...

Thanks Nick, and everyone for sharing your crystal sources, measurement data, suggestions and advice. 

Please keep sharing your data...!

73,  N7GTB