Saturday, 26 December 2015

RF Bypass and Filtration in a UHF VCO


I've spent >1 month studying frequency synthesis — a big subject that covers topics from writing code, to applying ECL + TTL logic — and finally making VCOs at VHF-UHF.

Some builders will buy a "1 and done IC" like the Si570, and that's great — but through QRPHB, I attempt to dive into more 'organic' design work @ the component level. It's great fun to play with prescalers, charge pumps, different order and styles of loop low-pass filters, phase and/or frequency comparators and the various schemes to divide down the feedback and reference frequencies. The theory and math in PLL loop filter design also challenges you. To me, the PLL = the ultimate challenging circuit to play with and learn about.

Also, over time, I hope to eventually develop up to ~10 GHz frequency synthesizers for my mostly ZIF receivers, but that's a long way off. Today, I'll focus on a UHF VCO and how some of the bypass caps and inductors were chosen to keep it stable.

Above — the discussion prototype VCO. You'll find this JFET VCO topology in newer Kenwood and Icom transceivers that give coverage into UHF. These radios use a PLL chip (such as the LMX2306) with an external PLL low-pass filter and VCO. Moving the loop filter and VCO off-chip presumably allows the engineers to lower VCO phase noise and to optimize the loop filter for various parameters including synthesizer phase noise. Further, for phase noise reduction, designers keep the VCO gain low but get a decent tuning range with higher reverse DC voltage applied to discrete varactors.
Since I'm using this VCO as the heart of a signal generator, my buffer design remains under  development — 1 signal gets low-pass filtered and then boosted to around +7 dBm for the VCO output, while the other gets boosted and goes to a ECL prescaler and then further divided by TTL logic in or outside a PLL chip. I'm still seeking the best strategy, but MMICs seem the easiest way to boost the signal level outside the VCO's metal container.

This VCO topology can easily tune >=30 MHz tuning bandwidth (from a 5-10 VDC PLL output), but in microwave terms it = a low bandwidth VCO. I don't have a need for a 1 octave VCO presently and lower bandwidth allows for the use of a higher Q, single turn, u-shaped loop of wire for the inductor + avoids the need for a frequency doubler for up to ~ 570 MHz from my experiments. I've built this VCO with both leaded and SMT JFETS. Cx varies the tuning bandwidth, however, @ UHF, stray capacitance weighs heavily along with your capacitor choices to determine the tuning range and signal amplitude. Even changing buffer circuit values may affect the oscillator function.

I built on double sided copper clad board with numerous copper via wires connecting the 2 surfaces. RF tight shielding and mechanical stability also factor.

Above —The 2 channels along with the low pass filter shown applied to the (blue) main VCO output channel. The yellow channel gets digitized in an ECL pre-scaler chip and the frequencies tracked well.

Above — Blue or main VCO output channel in a spectrum analyzer to look at the 2nd harmonic @ 808 MHz [ -43.7 dBc ]

Bypass Capacitor

Which capacitor value to choose for RF bypass at ~400 MHz? With a tracking generator and spectrum analyzer, I swept a 100 pF capacitor.

Above — A size 0805 AVX 100 pF capacitor swept to determine its series resonant frequency. SRF = ~361 MHz with a 44.4 dB deep notch.

Wes sent me a bag of these ~ 8 years ago and they deliver a good notch (at UHF) due to a Q >=500 @ 1 MHz. A couple no-name 100 pF capacitors bought on a popular online auction site showed a notch depth of on only -35 to 36 dB due to their poor Q. I won't buy any capacitors that lack a datasheet and always sweep a few to confirm the datasheet specification.

Because they generally go with a series inductor and/or a resistor as a low-pass filter — even though the SRF lies less than 400 MHz — they will work OK.

For pure RF bypass alone, a smaller capacitor value such as 82 pF might make a better choice due to the SRF of this 100p cap @ ~ 361 MHz. Further, to raise the SRF, you can also go to a smaller capacitor package like 0603, but that's getting small for the over 50 builder with presbyopia. One 100pF, Johansen size 0603 cap in my collection exhibits a SRF of ~475 MHz at a notch depth of -44 dB.

Above — A sweep of a size 0805 "ultra high Q"  Vishay cap. Special high Q caps get expensive — and I normally reserve them for bandpass filters and applications where I'm really going for Q. The notch lies 5.36 dB deeper than the AVX capacitors shown above. Normally, you'll see a notch boost of 8-10 dB with a special "high Q" SMT cap compared to a garden variety SMT cap. This shows that those AVX caps Wes sent me are quite good and suitable for my work.

Above — A sweep of a leaded MuRata MLCC 50v C0G capacitor with the leads cut quite short. The SRF dropped by ~ 131 MHz.

In my VCO schematic, the JFET drain DC voltage needs power supply decoupling and filtration. A series 470 nH choke and a 51 Ω resistor bypassed by 2 different cap values serves that function. Here's some experimental details of the individual components and — then finally, the finished network:


Above — A sweep of a through-hole, low Q ( less than 30 @ 1 MHz ) axial 1 µH axial RF choke. You see these a lot in homebrew radio projects. They resemble a color-coded resistor and work ~ OK at HF in some applications.
No deep notch. Note the generally shallow wideband response so typical of these @ VHF - UHF. We seek something with a little more Q up at UHF.

Above — A sweep of a 470 nH wire wound SMT choke that exhibits a 49 dB notch at 292 MHz. This wire wound choke's self-capacitance drops the SRF well below our target of 400 MHz. I've noticed from sweeping many chokes that the nH value often precludes intuition of the SRF. For example, a 68 nH device may exhibit a lower SRF than a 220 nH device. Things really vary from manufacture to manufacturer. For chokes at UHF, I've learned all that really matters = SRF and Q — and a datasheet and/or a sweep is the only way to learn this.

Above — A sweep of a wire wound 220 nH Coilcraft coil from my collection. Beautiful part!

Above — A snippet from a Coilcraft datasheet. Based on the performance of the 220 nH choke shown just above, I ordered 100 of the 330 and 470 nH values today. Note the stout maximum current — these will work well for MMIC decoupling too.

Pi filters

As mentioned, for filtering our DC lines, we usually apply series resistors or chokes plus bypass capacitors.

Above — A sweep of the pi filter: the 100 pF bypass cap on either side of a 51 Ω resistor. The deepest notch lies at ~223 MHz. while this filter may work OK --- how might we boost the notch and filter bandwidth?

Above — A sweep of the pi filter with 1 capacitor increased to 470 pF. This boosted the filter bandwidth + the attenuation at ~ 200 MHz. 

Above — A sweep the pi filter with 1 capacitor increased to .001 µF This boosted the bandwidth + the attenuation at  ~ 400 MHz. Further, we've extended the attenuation to 128 MHz with a notch of 41 dB.

This gets you thinking. If we combine the filter above with a series choke exhibiting a reasonable Q at UHF — can we extend both attenuation and bandwidth?

Above — Yes.  A sweep of the FET drain DC supply filter applied in the 400 MHz VCO shown in the first figure. By combining the series resistor, choke with decent Q and 2 different bypass cap values, a really nice filter emerged. Love this.

Further, 220 µF caps lie in shunt with the 1 nF RF capacitors on the DC rail. I boosted the 1K resistor to 1K to limit FET current and this further drops the filters' lower 3 dB cutoff @ AF + RF.

The end result = a well filtered DC rail. AF ripple can modulate the VCO and RF parasitics could also add unwanted modulation or noise + instability.

One problem still vexes me: my 470 nH choke with a SRF of 292 MHz gets applied as the DC return path for the varactor anode and on the FET drain. I really want the SRF closer to 400 MHz (just above is better).  For the varactor DC return, a high value resistor might work. Further, I could just wait until my ordered Coilcraft 330 + 470 nH chokes arrive. I can't wait to sweep these parts.

Above — A sweep of a source resistor (originally 150 Ω) and choke that goes on the FET source lead. I'm losing a little signal to ground due to less than ideal attenuation @ 400 MHz, however by raising the resistor to 390 Ω, losses were minimized I can adjust the FET current by tweaking the drain resistor when I bias each FET in this VCO design.

Measurement and problem solving on the bench poses my favorite activities in this hobby.

Thank you!

QRP-PosData --- January 5, 2016 ---   Miscellaneous

Above — Synthesizer / PLL books.  For basic PLL math and theory, anything by Roland Best will help you. I also enjoy the blue book shown written by P.V. Brennan.  In Introduction to Radio Frequency Design, Wes, W7ZOI wrote a great chapter on frequency synthesis for those who need a good primer (now out of print).

Above and below— MC145151. I'm always about 2-3 decades too late in this hobby. But the fun never stops with these relics. I only build to better understand electronic design — and not just to make stuff.

Above —My pre-scaler  and MCU collection slowly grows.  You might not imagine how much fun + learning these parts provide.

Above and below— Example discrete circuitry experiments I'm working with to learn the ropes.


  1. Good posting Vasily. The bitter irony that we have thought about is that the best and more appealing posters tend to post the least and the national enquirer quality ego boyz never shut up. Signal to noise ratio is another way to say this succinctly. Post soon because I am eager to learn some more.

    1. Hi Regis – well, I don’t plan to post often and prefer to read books and just measure stuff on my bench. “Blogger” emphasizes a misnomer for me. I’m more a learner -- a hapless student of electronics.

      It’s probable that most people just read blogs for entertainment. I don’t pretend to understand the kinder-digi of today --- I spent most of my life before the Internet and can’t relate to them. But, us analog experimenters needs to move over a little for these keen code warriors -- even if they just copy someone’s code and don’t really understand the math or algorithm – it’s all good ---- for it’s just a hobby.

      I draw comfort in the fact that life proves intrinsically analog -- and math and physics never change at my level. My transistors might be faster (now the fT = 5-9 GHz), but slow cooking, plodding bench experiments and occasional blogging work for me. Best! V

  2. Hi,
    it's necessary to post the datasheet from Si570:
    to take some view from "inside" block diagrams also the others options.
    Waiting to see the PLL circuit and more Vasily......

    1. Thanks Kostas. I really like the Silicon Labs products like the Si570 -- or perhaps their newest Integer-N clock, the Si5380. I also could have given examples like the LMX2470, or LTC6946, or even the now obsolete ( but still fun ) LMX1501 --- or whatever Fractional N / Integer -N synth block the wireless market has delivered to their customers.

      Studying how these chips work makes for incredible learning. Still, too, making your own discrete PLL component synthesizers -- or even if you just ply a PLL chip --- loading the latches and registers with a small 8-bit microcontroller that you coded in ASM might also prove very instructive for learning about synthesizers into the microwave frequencies.

      Really at this point, we're enjoying a great time -- with many options for generating an RF signal. As synth chips get better, cheaper and free code libraries and MCUs blossom, fortunately
      science and measurement however, never go out of style.

  3. Hi Vasily,
    if you apply 8-bit Mcu from Microchip or Atmel you're going to take fine results and I'm waiting to see the way to "build" the PLL blocks.A question Vasily:
    Why are you prefer the asm lang for programming?I think that is very difficult to study.I support C lang it's easy for both Microchip/Atmel Mcu,they have the same "base" in their programming,many libr examples specially arduino platforms..
    Friendly Kostas..

    1. Hi Kostas

      C works just fine. My hurried statements, (who’s got time to type during the holidays?) reflect on my desire to strip away all but the core aspects of the frequency synthesizer process.

      Much of what’s going on in the MCU in a homebrew radio might be performing related functions such as updating the display, or switching a filter or the offset. To learn, I’m only interested in the on-chip functions that set the registers for the reference and feedback dividers, latches and perhaps the pre-scaler divide ratio(s) which might even be dual modulus in more advanced synth designs. Further, perhaps learning the basic frequency functions such as up or down.

      Many basic synth chips have 3 inputs: clock, data, latch enable -- and the programming can be done with just an 8-bit CMOS MCU with itself just a serial interface to flash the EEPROM/EPROM/ROM /whatever as needed.

      Doing fancy I/O lends itself well to C or C++ code but I consider that separate from basic functions of learning how to make good, low-phase noise, UHF synthesizers - at least for me -- at least at this time. Questions such as what phase margin, lock time, loop bandwidth, or how many loop filter poles to choose still trouble me at this point. I realize software is available and helpful, but thinking, crunching math and testing results can’t hurt neither.

      All this is down the road as currently I’m experimenting with ECL and TTL chips like the MC12080 or the venerable 74HC74 -- and a few early PLLs that are programmed via pins. I’m uncertain whether I’ll even post much or any of this since it’s so boring to all but me and the cat. Many would say this is old school dribble --- “the last decades solutions” --- , but I’ll be honest – I’m a total newbie at synthesizer design.

      I’ll be working with the National Semiconductor LMX1501 IC at some point for it only has 16 pins and since I don’t use printed circuit boards, I’ll need to make a breakout solution and so forth. All this takes time and I’m in no hurry.

      (Plus other unrelated experiments are underway always.)

      Many of my inspirational UHF synth builders like Andy Talbot code their MCUs in assembly language. I prefer people refer to the abundant published UHF/ microwave work which is easily available and found with a search engine.

      As far as assembly language goes – there are only 35 instructions for a mid-range PIC chip and even if you don’t stick with it, learning a little low-level programming can only help your understanding of how computers work and how to code better in high-level languages. I remember 1 example in college long ago where – 1 machine instruction took 22 C –coded statements to accomplish.


  4. Thanks for your reply V...
    keep watching your project...

  5. Thanks Vasily. I learnt something from your write up (as I always do). It is ok to sweep capacitors and inductors to learn the SRF using a 50ohm source and termination for the measurement. Even though they may sit in a circuit with different source and terminating impedances (ie not 50 ohms) the measurement is still valid since you are just trying to measure the 'relative' attenuation in dB's. I have this requireent for a circuit I am building. This involves the use of an AD8307 (well shielded) that sits alongside a sweep generator. I need to design a low pass filter circuit on the power lead that will attenuate the sweep generator RF and prevent it affecting the AD8307 measurement capability down to about 80dBm. Many thanks Vasily for your work..73 Dick N4HAY

  6. Hey Mr Vasily Ivankno I need to tell you that your blog is soooooooooooooooo BORING & real hams don’t need a crutchs like a spectrum analyser they just make it happen & I suggest you go to a really exciting blog like SOLDERSMOKE & try figuring out QRP homebrewing by people like N2CQR and his pal N6QW who are authority homebrewers & it will only be a matter of time until Bill gets inducted to the QRP HALL OF FAME which is something you will never have to worry about QRP faN

  7. Can you give us some
    explanations about your opinion:
    "your blog is soooooooooooooooo BORING & real hams don’t need a crutchs like a spectrum analyser they just make it happen"
    Τhe people that you give as example,N2CQR and N6QW are hams like Vasily or me..

  8. is best blog of my tastes
    pozdrawiam serdecznie !eduardo

  9. Anonymous you are not the "average ham" for one thing,hams do not hide behind anonymous. I am proud of my hobby, and am grateful that people like Vasily is willing to take time to share his knowledge. He could VE7simply do his research and leave others in the dark. VE7DGH 73 Vasily

  10. Thanks for the support Dennis and others. My blog is not written to entertain its readers . My ignorance is vast, but I did not get my knowledge at the University of Google: just humble, bench experiments, mistakes and getting advice from my mentors. As always -- I invite these anonymous experts to blog their own findings -- to show us how and why they reached a conclusion. Peace and respect

  11. Wow anonymous on january 16 takes a shot at my GoTo QRP website and then praises Bill @ Solder Stink. He is in good company