Read Local Oscillator Design Project text version

Local Oscillator Design Project

Mike Suhar, WB8GXB The Midwest VHF/UHF Society is working on a transponder project and they needed two local oscillators. The first LO frequency was 2855 MHz and the second was 3690 MHz. I took on the project to develop the oscillators. The basic design would be a lower frequency oscillator feeding a harmonic generator and picking off the appropriate harmonic. Because of the multiplication factor an intermediate frequency would be selected, amplified then feed into another harmonic generator to achieve the desired frequency. This paper will look into design alternatives and discuss the merits of each. There are two basic designs approaches this project could take. The first would be a conventional oven crystal design. The second would be a PLL design which could use a stable 10-MHz or 1-PPS (pulse per second) GPS reference. This project provided a comparison of the design effort, material costs, and performance. The project was broken down into modules. Each module would be designed and could be modified as needed. The general block diagram of the project became:

82.00 MHz


738 MHz


3690 MHz




Harmonic Generator


Harmonic Generator


114.2 MHz


571 MHz


2855 MHz

Figure 1 ­ LO block diagram Conventional Crystal Oscillator Design We started with a conventional crystal design. This option had the advantage of low phase noise inherent to a crystal oscillator. Because of the frequency multiplication stability would be a problem even with a crystal design. The initial design assumed a 10MHz reference or a 1 pulse per second GPS reference was not available. Stability would have to come from the temperature stabilization of a crystal oven. I chose an oscillator I have worked with in the past. This is a Butler design1, 2 that uses a series resonant crystal. For the crystal module an output power of 0 DBm was the design goal. The harmonic generator would not part of the oscillator module. For the crystal a low phase noise version was selected from International Crystal Manufacturing. The final oscillator schematic is shown in figure 2. The part values shown in the schematic were for another project so they will vary slightly for this project.

Figure 2 ­ Crystal Oscillator All components would be surface mount where possible. As the entire oscillator module would be heated components so U3 and RT1 were included for temperature monitoring and control. Bias components were selected to provide a collector current in Q1 of 5-ma. Q1 would be powered from 9 volts with an on-board voltage regulator. This would avoid pulling the oscillator frequency should the main 12-volt supply voltage vary. The values for L1, C1, C2, C3, and C4 were derived using a few "rules-of-thumb" for this design3. C1 should be selected so Xc is less than the crystal's internal resistance. Unfortunately I don't know the crystal's internal specifications. For fundamental crystals Rs is from 10 to 20 ohms. It is higher for overtone crystals ranging up to 100 ohms4. If ordering a custom crystal this value can be specified. If we keep Xc for C1 below 25 ohms we should be in good shape as I am assuming Rs is higher than 25 ohms operating as an overtone crystal. The rule-of-thumb for the ratio of C1/C2 is between 3 and 4 with C2 being determined by C1. The rule-of-thumb for L1 is to start with an inductive reactance of 100 ohms. From these guidelines we can now determine the values. We have a couple of decisions as to fixed vs. variable components. I can use a Toroid core for L1 but then would have to make at least one of the capacitor's variable. The other option is to make L1 variable and used fixed value capacitor. The availability of

variable coil forms is limited. I found suitable coil forms from Lodestone Pacific. I obtained a few through Amidon Associates. If using the L33 forms take care on the tuning slug. They are very fragile. Make sure you obtain their tool to adjust the coil. My original oscillator for this design was for another project and I used the Lodestone coils. For this project I switched to the fixed Toroid core as I had trouble obtaining the Lodestone components when I needed them for this project. The variable coil design did not use C4 as the coil was adjustable. For the fixed coil C4 was added rather than make C1 or C2 adjustable. Starting with L1 and setting XL to 100 ohms we calculate L1 to be 0.190 uh. (XL=2FL) The resonant value of C is the series combination of C1 and C2. We can ignore C3 as it is a dc blocking capacitor. As such its value will be much greater than C1 or C2 so it will not have much influence on the series combination of the other two. A simple way to calculate LC resonance is to use the equation:

LC = 25330.3 Fo2

where L is in uh and C in pf. Fo is the desired frequency in MHz

The LC value at 82.000 MHz is 3.767. From this value we can determine L or C given the other. L = 0.190 uh so C =

3.767 = 19.8 pf 0.190

This is the series combination of C1 and C2. We can determine the individual values by considering the ratio of the two. Starting with a ratio of 3:1. C1 = 3 * C2. The formula for two series capacitors is:

Cs =

C1C 2 C1 + C 2

Substituting C1 to be 3C2 we obtain:

Cs = (3C 2 )C 2 3C 2 3 = = C2 3C 2 + C 2 4C 2 4


C2 =

4 4 C s = 19.8 pf = 26.4 pf 3 3 C1 = 3C 2 = 3 * 26.4 pf = 79.2 pf

Xc of C1 has to be below the crystal's internal resistance Rs which we estimate is between 25 and 100 ohms. For the give value of C1 1 1 Xc = = = 24.5ohms 6 2FC1 2 (82.0 * 10 )(79.2 * 10 -12 ) This should satisfy the requirement as we are at the low end of our estimated value for the crystal's series resistance Rs. The nearest standard values will be selected for these components. I did not factor in the variable capacitor C4 so select the lower standard values as C4 will increase the total capacitance. To complete the circuit we need a DC blocking capacitor C3. A value much higher than the other two will not change the resonant value of the tank circuit. Just make sure the series resonance of the selected capacitor is well above the frequency of the oscillator. For C3 we will use 1500pf surface mount capacitors on hand. A value of 1000pf or greater has less than a 1pf influence on the total series capacitance. There are a few more components on the schematic to account for. To suppress parasitic oscillations a small value of resistance will be added to the base of Q1. Recommended values are between 27 and 68 ohms. I will use 33 ohms for R4. L2 must resonate with the crystal's parallel capacitance at the overtone frequency. While I don't have the manufacture's value for the parallel crystal capacitance it can be easily measured on an RLC bridge. R3 may be required to control undesired oscillations when using high gain transistors. Values of 1.8k to 3.3k are typical. This may not even be needed in this design. I will set R4 to 2.2k so a pad is provided on the circuit board if needed. The output was link coupled off of T1. The original design just used a capacitor to couple to the buffer. Later the tuned circuit was added to help reduce any harmonics. Don't expect a lot of harmonic attenuation from this arrangement. The loaded Q is very low. The component values were based off of what I had on hand. This completes the oscillator stage of this device. Now we can examine the buffer stage. A transistor or an FET can be used in this application. One might think the FET buffer has a very high input resistance. Modeling with LTspice5 I found the buffer only has about 200 ohm impedances at this frequency. I confirmed this using the TAPR Vector Network Analyzer. Use caution when setting R7. If you don't have enough drain or collector (if using a transistor) current excessive harmonics will be generated in this stage. The output impedance is around 100 ohms of this stage.

The buffer was followed by a low pass filter consisting of C9, C10, C11, L4 and L5. Component values shown are for cut off frequency around 100 MHz6. This was originally the last stage of the oscillator module. Later a very small MMIC was obtained that allowed me to add the output amplifier to the circuit board. I could have place the filter after this amplifier. The output MMIC is a very small surface mount component made by Agilent and sold through Mouser Electronics. This has to be about the smallest surface mount part I have worked with. This is a SOT-363 package. The MMIC is rated at 21 DB gain with a 12.7 DBm 1-DB compression point. Useful up to 3.5 GHz with a 3.5 DB noise figure. The amp requires a 5-volt supply so I had to bring in 5 volts as I did not have room for another voltage regulator on the board. U3, R8, R10, C16 ­ Allow monitoring for adjustment of the oscillator oven without having to insert a thermometer into the enclosure. RT1 controls the heater circuit which was external to the oscillator package. See figure 5 for the heater schematic. The completed oscillator is shown in figure 3 and 4. The board was commercially manufactured by a board house. Today PC fabrication can be done relatively inexpensively through these board houses. Prototype boards are double sided, with plated through holes but no solder mask or silk screen layer. By using this design I could manufacture a number of boards and all I have to do is change parts values for different projects.

Figure 3 ­ Oscillator board bottom view

Figure 4 ­ Oscillator board top view The board was made as small as possible so the entire module could be heated. The board will mount in an enclosure made from rectangular aluminum tubing stock. I cut off a piece of stock large enough for the board. Two pieces of aluminum strap will be cut as covers for the enclosure. The aluminum tubing will in turn be surrounded by insulating material and mounted in a larger die cast aluminum project box. Q2 of the heater circuit is mounted directly on the aluminum oscillator enclosure.

Figure 5 ­ Oscillator heater schematic

PLL Oscillator Alternative Design The crystal oscillator was an interesting exercise but it has the draw back of not having an accurate reference. The design could have been modified to include electronic tuning for PLL control. This design also took time to create and layout the board. What can we do with off the shelf products that may save us time? TAPR has a reference lock board (REFLOCK II) available that looks to fit the design requirements without any PLL work on my part. The circuit was designed by Luis Cupido CT1DMK. The board uses a CPLD from Altera which is programmed to handle all of the frequency dividing. This could be interfaced to the crystal oscillator above but the purpose of this design was to obtain the parts off the shelf if possible. Looking around I found a low phase noise VCO from Minicircuits. The ZX95 series offers phase nose in the -130 dBc/hz range at 100 KHz in the frequency range we are interested in. Documentation on the Reflock board is somewhat lacking at the time of this writing but I am sure that will improve over time. There is a list email server where you can post any questions about the board. The designer frequents the list and answers a lot of questions. The board is only available in kit form at this time but that too may change in the future. This is a surface mount kit using 603 series components. The main CPLD is already attached. This is not a kit for those not experienced with surface mount construction. 603 parts are very small and difficult to work with. My oscillator board used 1206 and 805 SMD parts which are easier to work with. The TAPR board took six hours to construct and I managed to do it without losing any of the parts. A hot-air SMD rework station helped ease the construction. The CPLD is not programmed so you must have a JTAG program board and software. The programming software (Quartus II Web Edition) is available free from the Altera web site7. I managed to borrow the JTAG programmer but the parallel port version could be easily constructed from information available on the web. The CPLD firmware is available from links on the TAPR website. You have to choose between the firmware for an RF reference, i.e. 10-MHz or a 1-PPS reference. I was using a 10-MHz reference so I loaded the appropriate file. Once the firmware is loaded you have some configuration jumpers to set. This is done via 0-ohm jumpers that are also 603 series SMD components. You will have to determine the reference divisor setting and the VCO divisor setting. It may not be obvious from the existing documentation but the final "N" value is actually N+1. I had to replace one resistor in the feedback section of the PLL loop circuit. The tuning voltage of the VCO I used could be as high as 18 volts. The Reflock board was not designed to supply this high of tuning voltage. Fortunately the frequency range I required had a tuning voltage below 12 volts. With the parts supplied the V-tune range could not get to where I needed to go. Changing R22 to 62K solved the problem.

I need to optimize the loop circuit to reduce phase noise but I have not done so at this time. The completed assembly is shown in figure 6. The advantage to this design is it took only a couple of evenings of work from the time the parts arrived in the mail to having a running oscillator. The hardware costs were higher but it saved considerable time.

Figure 6 ­ TAPR Reflock board with VCO attached. The N divider settings are done via the SMD 0-ohm jumpers along the front edge of the board. Getting up to 3-GHZ ­ The multipliers Now we have the base oscillators but we need to get one of them to 2855 MHz and the other to 3690 MHz. From the block diagram in figure 1 we will use a two step multiplication process. Two different multipliers will be used to accomplish the task. Diode multipliers are commonly used for this purpose but MMIC amplifiers can also serve the purpose. A single diode with a lot of RF shoved into it would generate the necessary harmonics. This would be rather inefficient so they need to be optimized. I found an interesting diode multiplier using a full wave bridge8. The circuit in figure 7 will be used for the first multiplier stage. When used at HF the diodes can be 1N914. For the VHF/UHF range the diodes must be something faster such as the 1N5711 Schottky diode.

The reference article details this unusual multiplier circuit so I will not go into the design here. I will cover observations I made while testing this circuit. There is a two diode version of this circuit. I could not get that version to work without bridging another inductor across the two diodes. I also found the input inductor is very critical to the operation. I modeled the two diode version with LTspice and had the same results as the one on the bench. The multiplier provided multiplication but all harmonics were present. This circuit should only contain odd harmonics. I change the circuit to the 4-diode version and got better results but something still was not right. I still had even harmonics, but somewhat lower in amplitude than the odd harmonics. I discovered the reactance of L1 is critical. L1/C1 must resonant at the input frequency but even so if the reactance of L1 is not correct performance will suffer. Figure 7 indicates the reactance values that appear to provide the best results. The value is dependant on the input power. +10DBm is a good target for input power. Efficiency is low as the ninth harmonic will be -30 DBm at best. The proper value of L1/C1 will also result in the best return loss looking into the input port. L1/C1 resonates with the input frequency but select XL for best performance. L3/C3 and L4/C4 resonate with the desired output harmonic.

Figure 7 ­ Bridge multiplier

Figure 8 ­ Single diode multiplier

For the second multiplier a different approach was used. For generating microwave frequencies a single Schottky diode multiplier (figure 8) can be used or you can take advantage of harmonics generated by a MMIC amplifier9. For the single diode multiplier an HP 2835 or equivalent is generally used. I ended up using a mixer diode from a junk radar detector. L2/C2 form a simple low pass filter. C3 becomes a simple high pass filter and serves as a DC block. C1 serves as a DC block and L1 as an RFC for the diode bias current. You may need a resistor in series with L1 if you use very high drive current. As with the bridge multiplier the efficiency is low especially at higher multiplication factors. The MMIC multiplier is much more efficient and requires less drive. As the amplifier is driven into compression harmonics increase. By separating the output of the amp from a reflective type filter the response of the desired harmonic can be enhanced. Filters Filtering is required after the harmonic generator to pick off the desired harmonic. A PC board etched microstrip filter would be a prime candidate. The problem with a microstrip filters is design software is somewhat lacking in the ham budget range. This leads to design by trail and error...mostly error. For this project I went with copper pipe filters as they are not difficult to build and there is plenty of design experience on the web. I have a document on my web site that detail the design used for this project. The down side to the pipe filter is the size and weight. The difficulty in building the copper pipe filter is determining how to construct the coupling into and out of the filter. The mechanical constraints for this design lead to a fixed coupling arrangement. This tends to result in over or under coupled filters. Create your design so the coupling is variable. I achieved this by using a loop at the end of the semi-rigid coax. The UT-141 line was not soldered to the bottom of the pipe filter so I could rotate the loop to change the coupling. Once I found the desired response the line was soldered in place. A tracking generator will simplify the tune up procedure. Return loss measurements should also be made to confirm the loop design. For the 2 and 3 GHz output filters simple copper pipe cap filters were used10. Putting it all together Once the oscillator output is known, input/output characteristics of the multipliers, and filter response have been determined the amplifiers can be specified. MMIC amplifiers are used to achieve the required multiplier input drive and the final LO output. A goal of this design was to create circuit board that could be used for multiple projects just by changing a few components. Since the filters were external to the circuit board this goal could be achieved. This allowed me to use a PC board house to manufacture a number of "prototype" boards I could use on future projects. The first board contains the primary oscillator and the second board contains the harmonic generators and MMIC amplifiers to achieve the required power levels.

The crystal oscillator may have been the lowest cost but it took considerable more time to design and construct as compared to the PLL version using the Reflock board. The crystal oscillator may have lower phase noise but the frequency is not as accurate as the GPS stabilized approach. For updates to this project see my web site11. Wessendorf, Kurt, Sandia National Laboratories, "Oscillator Design Techniques Allow High Frequency Applications Of Inverted Mesa Resonators". [internet][cited June 1, 2006]. Available from:

2 1

Hayward, Wes, W7ZOI, Introduction to Radio Frequency Design, (The American Radio Relay League: 1994) pp.261-340 Philips Semiconductors, "Crystal Oscillators and Frequency Multipliers Using The NE602 and NE5212", Application Note AN1983, pg. 5. Application Note 726: Specifying Quartz Crystals. December 29, 2000. [internet][cited June 1, 2006]. Available from:

5 4 3

LTspice. [internet][cited June 1, 2006]. Available from:


Low Pass Filters. November 27, 2003. [internet][cited June 1, 2006]. Available from:


Quartus II Web Edition. [internet][cited August 6, 2006]. Available from: Wenzel, Charles, Wenzel Associates Inc., "New Topology Multiplier Generates Odd Harmonics". [internet][cited June 1, 2006]. Available from:

9 8

Davey, Jim WA8NLC. "Frequency Multipliers Using Silicon MMICs". In: The ARRL UHF/Microwave Projects Manual. The American Radio Relay League: 1994. p 5-13 ­ 15 Britain, Kent WA5VJB. "Cheap Microwave Filters From Copper Plumbing Caps". In: The ARRL UHF/Microwave Projects Manual. The American Radio Relay League: 1994. p 6-6 ­ 7 Suhar, Mike WB8GXB. "Local Oscillator Design Project". [internet][cited August 22, 2006]. Available from:

11 10


Local Oscillator Design Project

12 pages

Find more like this

Report File (DMCA)

Our content is added by our users. We aim to remove reported files within 1 working day. Please use this link to notify us:

Report this file as copyright or inappropriate


You might also be interested in

Basic Oscillator and Reset Circuit
070415 A Three Transistor Discrete FM Transmitter.doc
A GPS-Based Frequency Standard
Microsoft Word - LaneReport.doc