This is a somewhat long missive on my experience of buying crystals for amateur radio projects. It represents my experience, your experience might be different but I hope this may be of use to some of you. As many of you probably know, International Crystal Manufacturing went out of business in May of 2017. While not cheap (about $35 a crystal), they were one of the few (last?) in the US to make a single crystal to your order. Over the years I purchased a number of crystals from International for land mobile radio, amateur radio, student projects and consulting. When they closed, I felt like I lost an old friend but I have managed to find some potential replacements. What I needed were some 3^rd overtone crystals in the 40-80 MHz range for microwave LO’s and some new crystals for my GE MASTR PE radio. The PE is a 2 watt 2m FM hand held radio made by General Electric in 1970’s.

?

The first company I found is in England: Quartslab: ? Their website clearly states that they cater to the amateur radio market and they will make a single crystal, no order too large or too small. However, I found their website limited technically. It suggested that Quartslab would only calibrate fundamental crystals with 32 pf load capacitance or overtone crystals to series resonance. It was also missing some technical information on available calibration and temperature tolerances. However, I corresponded via email with Dave Hayes GW4AKY at Quartslab. Dave was very responsive, answered my questions, and helped me understand where their cost trade offs are. They had options that didn’t appear on their website. I purchased two 39.625 MHz 3^rd overtone crystals in a HC49u package. They were calibrated to +/-10 ppm at 25C with a 24 pf load capacitance and a +/-10 ppm -10C to +60C temperature tolerance (yes, you can specify a load capacitance for an overtone crystal. I needed to pull the crystal +/-500 Hz around its specified frequency for a PLL). The crystals cost 27.63 pounds Sterling each (~$36) with shipping. Payment was made through PayPal.

?

Many crystal companies have manufacturer’s data for land mobile gear but Quartslab did not have data for the GE PE. They said they would measure the PE crystals and make a new set. I sent them a set of crystals from the PE radio. Their website does say they have data for the crystals in a lot of vintage ham radio gear so, if you need something, ask.

?

Don’t expect fast turn around. They quoted me approximately 20 working days (~month). My order was placed 2/19/2020. The crystals were shipped 4/11/2020 via Royal Mail. Unfortunately, they took 2 months to get here. That was not Quartslab’s fault. The pandemic had significantly slowed international mail. You could probably get them to ship via DHL or Fedex but you would have to pay for it. However, it took them a fairly long time to measure the test crystals for the GE PE. While I sent them with the 2/19/2020 order, the PE crystals couldn’t be ordered until 5/27/2020. They were shipped on 7/13/2020. The Royal Mail was faster this time and they arrived 7/23/2020.

?

Quartzlab made the PE crystals in the same package as the originals. The originals were solder sealed but the new ones have a resistance weld seal. The seal crease around the bottom of the crystal was just a little too big to fit in the very small PE oscillator module so I never got to check their calibration. My mistake.

?

As a footnote, good crystals are a little like fine wine. They need to age. The stresses caused by the cutting and polishing operation need to age out. Crystal frequency drift is logarithmic with time so crystals are often baked after initial cutting to reduce the drift to a reasonable level before final calibration. This operation can take a couple of weeks. I would be skeptical of anyone who offers you a custom crystal in less than a week. If my memory serves, International used to offer a 1 week expedited service (for a price!).

?

While waiting for Quartslab to measure the GE crystals, I contacted Bomar Crystals. Bomar is located in the US and is probably the last company making small order crystals here. They have supplied the land mobile radio market for years. They had the manufacturer's data for the GE crystals and they quoted me directly from the GE oscillator part number. They have a $100 minimum order so each crystal was $50. They quoted me a lead time of 6 weeks. Shipping was an additional $10 via US Priority Mail. The crystals were paid for by credit card and they shipped almost to the day after 6 weeks. The crystals came several days later. They trimmed right on frequency in the PE.

?

Interestingly, Bomar did not use the original sized package for the GE crystals. They used the much smaller UM-1 package. Did they know about the seal crease problem?

?

I corresponded with Minnie Lirio at Bomar via email <sales@...> . She is in sales so there wasn’t much of a technical exchange. She was very cordial and responsive. She must have been working from home as a result of the pandemic as she answered my email in less than an hour at midnight!

?

The second company is in the Czech Republic: . Their website is excellent with good technical information. I could not find any ordering information on the website but I corresponded with Josef Vozenilek via email. His address is on the website. He was very responsive and his English is much better than my Czech! I ordered four crystals, two 59.7000 MHz and two 53.8500 MHz to the same specifications as above. They offered to measure my GE crystals but they didn’t have data. My order was placed on 3/26/2020 and it was shipped 4/17/2020 via post. The post took about 2 months. The amazing thing about Krystaly is the price. The crystals were only $12 each. Postage was an additional $8 for the order. Payment was via credit card. Krystaly provided a secure website to make the payment so your credit card 飞辞苍’迟 choke on an international transaction.

?

Krystaly did something I had never seen before. They provided measured frequency calibration data as well as the measured series loss resistance (Rm), the series motional capacitance (Cm) and the holder capacitance (C0) of each crystal. With Rm, Cm and the frequency, you can calculate crystal Q. The calculated Q was on the order 100,000 which is respectable. The Q of the crystal in an oscillator circuit is the primary parameter that defines oscillator phase noise. While the effect of the oscillator on crystal Q can be controlled to a degree by design, the Q of the crystal itself is usually an unknown. You can measure it but Krystaly provided the data.

?

I subsequently ordered a second set of crystals from Krystaly. The request for a quote was emailed on 9/15/2020 but I didn’t hear back from them until 09/23/2020. It pays to be patient. The crystals were ordered on 09/25/2020. Using the measured data from Quartslab, the order was for two fundamental mode crystals for the PE. The smaller UM-1 package was specified and they were only $10 each. They were shipped 10/08/2020 and arrived via post on 11/05/2020. When installed in the PE, they trimmed on frequency.

?

While the third mode crystals from Quartslab and Krystaly have yet to be incorporated into my microwave LO projects, the series/parallel resonant frequencies were measured in a standard 25 ohm test fixture with a GPS locked Tektronix spectrum analyzer/tracking generator. They all appear to be quite close to their specified frequency. Dave at Quartslab said I should be able to pull the crystal frequency the necessary +/- 500Hz.

?

Quartslab and Krystaly are both good candidates to replace International for crystals for transverters and other ham radio projects. I found their communications very good. You do need to know some of the crystal speak to get what you need but that shouldn’t be too difficult. If you need help to specify what you need, Quartslab seemed more ham friendly. The Krystaly price is very attractive. The Bomar communications were excellent as well but their minimum order makes them less attractive for one off ham projects. However, if you are recrystaling a GE, Motorola or other commercial repeater, Bomar is probably the place to go since they appear to have an extensive database of manufacturers’ crystal parameters. The PE in not the most common of GE radios but the crystals they made for it were right on. Bomar made no mention of temperature compensation, but I didn’t ask. One the gripes about recrystaling classic land mobile gear is that no one seems willing to do the temperature compensation necessary to maintain the temperature stability of the original GE/Motorola TCXO’s. See: .

?

Another source of crystals is surplus. Surplus Sales of Nebraska <> has a extensive list of surplus crystals. They range from $4-$10 each. They don’t provide much information short of frequency and holder type and the latter does not seem to be fully descriptive. I bought a 39.6 MHz crystal labeled as HC18 (the obsolete soldered version of the HC49 holder) for $10. It was more expensive than most of their other HC18 crystals but, when it came, the crystal was in a high stability HC43 cold welded package. They have a number of TO-5 package crystals in the 100 MHz range that should work in the Frequency West/California Microwave type oscillator bricks. Many of the TO-5 crystals in this range were intended for oven operation in the oscillator bricks. I purchased several crystals that multiply out into the 5.6 and 10 GHz ham bands.

?

What has been your experience in obtaining crystals? A number of crystal companies have been bought out or have just disappeared. With the rise of synthesizer based land mobile gear, the market for individual crystals has waned. With the possible exception of Bomar, it doesn’t appear that anyone in the US is making one off crystals at a price attractive to hams. I found crystal houses in Canada, Australia and Europe but it wasn’t clear that they will service the amateur radio market. Interestingly, I didn’t find any crystal houses in China. Maybe they are all high volume or maybe I just didn’t know where to look. What experience have others had?

?

?

?

?

Hello Dennis,

Many thanks for your interesting contribution. I am working on a couple of low noise crystal oscillators actually, at 14.0 and 100.0 MHz, and I use crystals of unknown origin at this moment. So I asked for a quote of some crystals in Czech Republic and I am curious how they will perform.?
With respect to oscillator noise spectra: the crystal is dominant near the carrier, the rest of the circuit further away. I use fets in both circuits. I have a homemade Phase Noise Test Set that is calibrated against a R&S FSWP instrument. In case of interest I can report back here.
Best regards,
Harke

On Wed, Nov 25, 2020 at 05:16 PM, Reginald Beardsley wrote:

ROFL!!!!!!!!!!!!!!!!!!!!!!!!!!

I just sent this link to a friend:

http://www.aholme.co.uk/PhaseNoise/Main.htm

and the next thing I see is this message.

Have Fun!
Reg

On Wednesday, November 25, 2020, 07:03:45 PM CST, Daniel Ricardo Perez via groups.io <danyperez1@...> wrote:

?

?

Dear Harke:
Not everyone has the means to measure phase noise! Could you please share the link from where you built it?
73!
Daniel Perez LW1ECP


25 de noviembre de 2020 12:34:22 ART, Harke Smits via groups.io <yrrah53@...> escribió:
Hello Dennis,
Many thanks for your interesting contribution. I am working on a couple of low noise crystal oscillators actually, at 14.0 and 100.0 MHz, and I use crystals of unknown origin at this moment. So I asked for a quote of some crystals in Czech Republic and I am curious how they will perform.?
With respect to oscillator noise spectra: the crystal is dominant near the carrier, the rest of the circuit further away. I use fets in both circuits. I have a homemade Phase Noise Test Set that is calibrated against a R&S FSWP instrument. In case of interest I can report back here.
Best regards, Harke

Hello All,

Very interesting response! I use cheap crystals indeed and I am just curious how special made ones perform. I like the link by Reginald, though my PNTS is a little less elaborate.?
My PNTS is based on articles by Bernd Kaa in UKW Berichte 4/2015,1/2016,2/2016 and many stuff on the internet. I m working on a series of articles in Dutch for our national magazine at present.?
The attachment gives you an impression of its capabilities.
To be continued, I guess.
Best regards,
Harke

Phase noise seems to be one of those topics of interest but measuring it can be a challenge. I dug into this a couple of couple of years ago with the idea of designing low noise microwave LO’s. Attached is a presentation that I gave at Microwave Update in 2011. This paper focuses on the analog side of the measurement. Hopefully it demystifies some of the calculation associated with the measurement. It turns out to be simpler than I though, but the presentation is the result of a lot of nights wading through HP app notes, IEEE papers and measurements that didn’t work! What you are building is a very sensitive direct conversion receiver and, as they say, “the devil is in the details.” Careful shielding and battery power make a big difference in performance.

?

What is missing is a good display. I am primarily an analog guy with limited computer programming skills. My early measurement used an HP3590A selective level meter and then an HP3561A Dynamic Signal Analyzer (an FFT based spectrum analyzer). Both of these are obsolete and probably hard to find. I attempted to use a Velleman pcsgu250 USB scope. It has a spectrum analyzer function that displays watts/Hz power spectral density. Unfortunately it is not really usable because of its limited 8 bit dynamic range. I subsequently found the Digilent Analog Discovery 2. It is another USB based scope with a 16 bit sampler and a spectrum analyzer function. Unfortunately, while it is supposed to display power spectral density, I never figured out how to calibrate it. Somewhere you need to know the spectrum analyzer bandwidth to do the phase noise calculation. There are some high end sound cards with 16 and 24 bit dynamic range and 196 kHz sample rates. Anyone aware of compatible spectrum analyzer software that will display power spectral density?

?

I have occasional access to a high end Rohde & Swartz FSW spectrum analyzer so continued development was put on hold. Anyone who is interested in continuing it is welcome, good luck and have fun!

开云体育

Very good presentation, indeed, Dennis! Thanks.
With respect to the Clapp oscillator. Please refer to the drawing attached. It is a very simple design but appears to be working very good. I also made a design by prof Jeremy Everard of the York University (please google) with low noise lf transistors, but it looks like the Clapp shows less noise. With the same cheap crystals.?
My Phase Noise Test Set is basically analog by design (by DG4RBF, UKW Berichte 4/2015), I am an analog guy, and the sound card will process the spectrum (very well) by a sw called Audiometer (by DG8SAQ). It is pretty standard in that it consists of a diode ring mixer as phase detector followed by a ultra low noise Wenzel based lf amplifier (60 dB). There is also a PLL to keep the oscillators at quadrature. The amplifier is sensitive enough to show the thermal noise of a 50 ohm resistor. Obviously, you need two oscillators so one of them is provided with EFC. As reference I use a HP10811 at 10 MHz, a Clapp at 14 MHz and at 100 MHz an overtone oscillator with a fet. I hope to order good quality 14 and 100 MHz crystals to see if it makes any difference. The 14 and 100 MHz oscillators are followed by a low pass filter and if necessary an attenuator to get about +6 dBm output power. This is what my PNTS prefers. Moving the DUT power up and down 3 dB moves the spectrum up and down about 1 dB. The convergence of the spectra at the high end (previous post) seem an anomaly of the sound card used. In any case I consider an accuracy of 1 dB in this set up as good (enough).?
So far some more details. Feel free to ask for more info.
Best regards,
Harke

The literature on low noise oscillators is vast and the current state of the art can be a bit exotic. I can’t do cryogenic sapphire resonators! My foray into low noise oscillators was driven by my interest in narrow band communications at microwave frequencies.

?

Here are some papers that I found useful:

?

Michael M. Driscoll, Two-Stage Self-Limiting Series Mode Type Quartz Crystal Oscillator

Exhibiting Improved Short-Term Frequency Stability, IEEE Transactions on Instrumentation and Measurement, June 1973, pp. 130 – 138.

?

M. M. Driscoll has a number of a papers on high performance oscillators and oscillator measurement. My early attempts at phase noise measurement mimicked Driscoll’s system. Variants of his two stage oscillator with the oscillator amplitude externally controlled seems to be the gold standard for a lot of low noise oscillators. For me, this paper was worth the slog but I don’t claim to fully understand it. Most of Driscoll’s papers are in the IEEE literature and are written at an advanced engineering level. The IEEE literature may be difficult to get. If you have access to a university with an engineering school, they probably subscribe to the IEEE Xplore paper database or you can get to know the research librarian at you local library. Research librarians love to ferret out unusual stuff!

?

I have successfully simulated some Driscoll like circuits in LTspice but getting oscillators to oscillate in LTspice can be tricky. My “crystals” in the simulation can’t have a Q much more than about a 1000 otherwise the circuit 飞辞苍’迟 oscillate. The higher the Q, the longer it takes for the oscillation to start until they 飞辞苍’迟 oscillate at all. However, the simulations have been helpful in understanding the loading effect that the oscillator circuit has on the crystal.

?

?

It is my opinion that crystal Q is the dominate factor is determining phase noise. See:

?

D. B. Leeson, “A simple model of feedback oscillator noise spectrum,”Proc. IEEE, vol. 54, no. 2, Feb. 1966, pp. 329–330.

?

The Leeson paper is a classic and, while there are more rigorous analyses, it gives a useful intuitive understanding of the phase noise trade offs.

?

The next paper captures some of the insight of the Dricoll paper and it has a circuit you can actually build. I found it in a web search.

?

Chris Bartman, GW4DGU, Notes on the Driscoll VHF Overtone Crystal Oscillator and a New Low-Noise VHF Crystal Oscillator Topology, Scatterpoint, April 2008.

?

Bartman explores some of the second order effects. Probably the most important is the 1/f or flicker noise in the device used. That tends to dominate the noise spectrum at few hundred Hz and less from the carrier. He suggests devices from the NE688 series. In addition, the NE856 series is one of the few devices characterized for 1/f noise. Unfortunately, both series are no longer available and they are fast becoming unobtainium. I got a life time batch off eBay but I don’t particularly like to use obsolete parts in my designs. See:

?

California Eastern Laboratories AN1026: 1/f Noise Characteristics Influencing Phase Noise

?

Jeremy Everard explores the effect of crystal loading in the oscillator circuit and the trade off between crystal loaded Q and phase noise:

?

Jeremy Everard, Low Phase Noise Oscillators including some Detailed Designs, 2007 IEEE International Frequency Control Symposium joint with 21^st European Frequency and Time Forum, Geneva, Switzerland, May/June 2007, pp 1156-1163

?

J. Everard and K. Ng, “Ultra-low phase noise crystal oscillators,” Proc. IEEE 21st Int. Freq. Control Symp. Joint Eur. Freq. Time Forum,Geneva, Switzerland, May/Jun. 2007, pp. 1246– 1250.

?

The Leeson paper above suggests that the higher the loaded Q, the lower the phase noise. Everard argues that as the loaded Q of the crystal in the oscillator circuit approaches the unloaded Q of the crystal, the phase noise actually degrades. Everard shows that as the loaded Q increases, less power is transferred to the oscillator sustaining amplifier so the signal to noise in the oscillator degrades. At some point, the phase noise starts to increase. This suggests that there is an optimum loaded Q. The Driscoll two stage oscillator may not suffer from this limitation. In the Driscoll oscillator, the crystal is not directly in the feedback path so it is possible to control power input to the oscillator independently of crystal loaded Q. Something to research.

?

The following two papers are quite readable and the give good insight into what controls oscillator phase noise and how to measure it. They also discuss the effect of the oscillator on crystal loaded Q.

?

B. Neubig, DK1AG, An Extremely Low-Noise 96 MHz Crystal Oscillator for UHF/SHF Applications, VHF Communications, 3/1981, pp 135-143.

?

B. Neubig, DK1AG, An Extremely Low-Noise 96 MHz Crystal Oscillator for UHF/SHF Applications Final Part II, VHF Communications, 4/1981, pp 194-203.

?

This paper appeared in QEX and it is the basis of some of my microwave LO work:

?

John Stephensen, KD6OZH, A Stable, Low-Noise Crystal Oscillator for Microwave and Millimeter-Wave Transverters, QEX, Nov/Dec 1999, pp 11-17.

?

Unfortunately, many of the parts used in Stephensen’s circuit are now obsolete but he has a simple example of what Everard is talking about.

?

Using some of the insight of Chris Bartman, I built a variant of KD6OZN’s oscillator and designed a PLL that will lock a 3^rd overtone crystal oscillator in the 40-80 MHz range to a 10 MHz OCXO reference. The PLL uses an Analog Devices ADF4002 whose data is loaded with a PIC microcontroller. The Analog Devices folks really understand their PLL’s. They provide separate power and ground connections for the digital, analog, and phase detector sections of the ADF4002. Managing them correctly is key to getting the best performance.

?

The plot below is the measured phase noise of this circuit with a phase locked 63 MHz crystal oscillator multiplied to 10080 MHz (10080 + 288 = 10368 MHz). The multipliers are all passive Schottky diodes so there should be little excess phase noise beyond the 20log(n) predicted by modulation theory. The PLL bandwidth was set for 100 Hz so you see the phase noise of the 10 MHz OCXO dominate at low frequency close to the carrier and then roll off to the effect of the 63 MHz oscillator. This was my first attempt. I will probably set the PLL bandwidth in the next version to 10 Hz in an effort to roll off the 10 MHz OCXO noise faster. Even at that, this synthesizer is 30+ dB better at 1 kHz off the carrier than the phase locked Stellex YIG I am currently using for my 10 GHz transverter.

?