8
\$\begingroup\$

I want to know what filter circuits look like in the RF (MHz-GHz) frequency range. Are they just like low frequency LC filters, but with appropriately adjusted component values?

For example the VLF-80+ from minicircuits. Why are they so expensive, at around $30 per part? I know that they are high order filters, but still $30 sounds too expensive for a simple LC filter.

I ask this because I tried to design a 200 MHz bandpass filter myself.

I ordered a PCB from JLCPCB but the filter doesn't work at all. It's not easy to identify the problem because I don't have a high frequency network analyzer. I can't even measure the components because my multimeter can't measure capacitances that are less than 100 pF.

Here's my 5th order, 50 Ω output filter.

enter image description here

And here's the PCB layout. I know it's not good enough. I just started designing PCBs. enter image description here

\$\endgroup\$
14
  • 2
    \$\begingroup\$ @MustafaTurhan Do you have controlled-impedance traces on your board? Are you using appropriate capacitors and inductors (at these frequencies, you need C0G capacitors and air-core inductors)? \$\endgroup\$
    – Hearth
    Oct 9, 2023 at 12:09
  • 4
    \$\begingroup\$ Oh, if you scan the VCO, how do you synchronize your spectrum analyzer to that scan? Otherwise, you won't get meaningful measurements: It would fully depend on when the SA "looks" at your oscilator. \$\endgroup\$ Oct 9, 2023 at 12:15
  • 4
    \$\begingroup\$ @MustafaTurhan At these capacitances, most ceramic SMD capacitors are going to be C0G or maybe U2J, which is almost as good. It's the inductors that are the problem, and the trace impedance. Anything above about 50 MHz or so is basically going to require air-core inductors, as ferrite and iron cores become very lossy (there might be a handful of ferrite formulations that can still be used, but i wouldn't recommend it). You also absolutely need controlled impedance traces; right now, the impedance mismatch is going to be reflecting much or most of your power back to the source. \$\endgroup\$
    – Hearth
    Oct 9, 2023 at 12:28
  • 2
    \$\begingroup\$ When working at this sort of frequency, it's usual to wind your own air-core inductors, incidentally. I suggest you get a good LCR meter with measurement frequencies of at least 100 kHz, preferably 1 MHz, or even better, a network analyzer. Then get yourself some sufficiently stiff magnet wire (#16 AWG is a good balance, though #18 and even #20 would work) and something to use as a mandrel (I use the handle of an x-acto knife, a sharpie, and a ¾" wooden dowel for three different sizes), wind some inductors, and measure and tweak as necessary. \$\endgroup\$
    – Hearth
    Oct 9, 2023 at 12:34
  • 3
    \$\begingroup\$ @MustafaTurhan Yes, you certainly can. To get an exact value, you can spread out the windings to reduce their coupling, which lowers the inductance in a smooth continuous fashion. I recommend putting the inductors in-circuit and then tweaking them while monitoring how the response changes; that tends to work better than just using calculated values, as it automatically compensates for parasitic elements. It's very manual, though. You may start to see why the mini-circuits parts cost so much. \$\endgroup\$
    – Hearth
    Oct 9, 2023 at 12:43

3 Answers 3

14
\$\begingroup\$

As you've discovered, once you are up in the 100MHz+ range, little is "simple".

When using discrete capacitors and inductors, you have to consider their parasitic complements. Even basic connecting elements such as traces have to have their characteristics carefully considered, because they too have capacitance, inductance, impedance and skin effects.

It can all be mastered (and there are lots of good texts on practical RF design), but whilst it is simple to model such filters with ideal components, the physical reality is much more complex...

At 80m/3.5MHz filters are still built with discrete components and ferrite cored inductors. (See this ARRL 80m design.)

enter image description here

This link shows the inside of a practical DIY 2m / 150MHz bandpass filter, with screw mount piston adjustable capacitors and hand formed air-cored inductors. The adjustable caps are critical, because you need to compensate for the hard to predict stray capacitances. The inductors are air cored due to high ferrite losses, and low target inductance.

enter image description here

Professionally designed/built filters in the GHz range will typically use transmission line designs created with CAD software which can do EM analysis. You will rarely see discrete components, just oddly bent traces and stubs, but they can work wonderfully.

Have a look here for an example of a 3GHz-11GHz bandpass filter from a master's thesis.

enter image description here

And the resulting performance.

enter image description here

In your case, you'd want to make sure you are using air cored inductors, and that you allow for stray/parasitic capacitances.

A discrete design built in a metal box like the 2m example, might be your best approach in terms of ability to experiment. In fact you could just adapt the example circuit for your use by tweaking the values

\$\endgroup\$
9
  • 1
    \$\begingroup\$ Okay it's not so simple. Then how should I proceed to design the filter. Do I need to do Ansys simulations? How should I test the components? I graduated only a year ago and I always wonder how a experienced engineer design circuits. I always take the easiest route. I know that things are not so simple but I don't know other more advanced ways. For the RF filter, I calculated the component values from an online LC filter design tool. I designed the pcb from KiCAD and ordered it in Jlcpcb. But now it doesnt work. \$\endgroup\$ Oct 9, 2023 at 12:04
  • 1
    \$\begingroup\$ "Then how should I proceed to design the filter" -- If that is your actual question, then start a new submission and ask your actual question. \$\endgroup\$ Oct 9, 2023 at 12:29
  • 4
    \$\begingroup\$ If you are starting RF design from scratch, then personally I'd recommend a copy of the ARRL handbook arrl.org/arrl-handbook-2023. Second hand copies are cheap, and anything dated in the last 20 years will deliver on both analog and digital circuits (from time to time I still reference my 1988 copy). Critically, they contain lots of the practical details which more theoretical texts often ignore. \$\endgroup\$
    – colintd
    Oct 9, 2023 at 13:10
  • 2
    \$\begingroup\$ I've added examples of filters in various frequency ranges to give you a feel for the way practical constructions vary. Hope it helps, but do look at ARRL handbook for more guidance. \$\endgroup\$
    – colintd
    Oct 9, 2023 at 22:51
  • 1
    \$\begingroup\$ @colintd Thank you! This is very helpful. I will study book and learn more about the theory before desiging again. \$\endgroup\$ Oct 10, 2023 at 12:21
2
\$\begingroup\$

The challenge when realizing filters for high frequencies comes from two main issues: parasitic elements and component tolerance. These two factors work against you as the frequencies get higher: the L and C values get smaller, so these deviations have greater influence on your filter.

Why are the commercial ones so expensive then? They’re made with premium materials and use higher-stability, lower-parasitic components. In addition, they are tuned in manufacturing to their target specification.

There are also economics to consider. Filters that have a high-volume use (like diplexers for cable modems and set-top boxes) can be remarkably cheap for the performance they deliver, because their development and tooling can be spread out over a wide number of units. As an example, typical cable TV diplexers have PC boards in a can, use air-core inductors and polyester or mica caps, and sell for a few dollars in volume.

Specialty filters with lower running volumes will cost more as their upfront investments need to be recovered over fewer units. Inevitably, if there isn’t competition then prices will be even higher, because reasons.

\$\endgroup\$
1
  • 1
    \$\begingroup\$ A much simpler explanation for cost suffices: Minicircuits and etc. are in the small quantity specialty market; they charge whatever people are willing to pay. Low quantity may mean higher production cost as well (though I suspect given the simplicity of many of their products, most of that either covers engineering cost, or sweet sweet profit margin). \$\endgroup\$ Oct 10, 2023 at 4:49
1
\$\begingroup\$

I simulated your filter and it looks good. You need to buy affordable NanoVNA to check where the problem is.

enter image description here

\$\endgroup\$
2
  • 3
    \$\begingroup\$ The challenge in physical implementation (as opposed to simulation) relates to stray capacitances, the use of ferrite inductors (which will tend to be very lossy at 200MHz), and trace impedance mis-match. Be interesting to try adding a few pF stray capacitance in multiple places down to ground (and across the inductors), and some resistors in series with the inductors to match their loss characteristics. \$\endgroup\$
    – colintd
    Oct 11, 2023 at 18:32
  • 1
    \$\begingroup\$ This simulation doesn't take any parasitics into consideration, though. The "You need to buy X" makes me almost think this is some kind of sophisticated spam.... \$\endgroup\$
    – Hearth
    Oct 11, 2023 at 20:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.