What is the maximum frequency that can be practical to build a filter with discrete components?
I need a band pass filter for 1090Mhz.
1
-
1$\begingroup$ Above HF, filter characteristics can depend more on how you build the filter than on the value of the discrete components. Huge difference between 0805s on an 8 layer PCB versus similar value leaded components plugged in to a solderless prototype breadboard. $\endgroup$– hotpaw2Commented Aug 4, 2021 at 16:27
Add a comment
|
3 Answers
$\begingroup$
$\endgroup$
3
Lumped elements are theoretical constructs from the lumped element model. So you can't really build a filter from lumped elements at any frequency.
But, you can ask at what frequency discrete components stop being approximated by the lumped element model. As a rule of thumb, anything that is physically bigger than 1/10th of a wavelength is definitely not well approximated by a lumped element. At 1 GHz, that's about 2.7 cm, so a discrete component filter at that frequency is not entirely out of the question.
However, all but the most basic filters can be quite sensitive to parasitic elements, which will be significant at sizes even less than 1/10th of a wavelength. So the filter must be significantly smaller than 2.7 cm.
I'd say it's not entirely out of the question, but you will probably need to build this filter on a PCB, with a good layout, and surface mount components, and small ones at that.
answered Aug 3, 2021 at 15:49
-
$\begingroup$ thank you @Phil, I rephrase the question replacing lumped elements for discrete components $\endgroup$– mpradioCommented Aug 3, 2021 at 16:03
-
1$\begingroup$ And, of course, at some point the parasitic values exceed the discrete values, and then you can build circuits almost entirely out of PCB traces with almost no discrete components. $\endgroup$ Commented Aug 3, 2021 at 22:00
-
$\begingroup$ @Phil In this paper they mention filters built with "lumped elements" A New Design of Lumped Element Bandpass Filter in L-Band $\endgroup$– mpradioCommented Aug 4, 2021 at 14:11
$\begingroup$
$\endgroup$
2
Two datapoints that might help:
A previous company we had a 20-6000 MHz diplexer with a changeover at about 900 MHz, using 0402 parts. Some trial and error to get it to match the simulated design.
A current bandpass (highpass+lowpass) filter for 570-1050 MHz is made entirely from discrete parts, including hand-wound wire inductors instead of chip inductors. Good for tuning.
Capacitors work comfortably up to many GHz, but most inductors SRF is only a few hundred MHz. Coilcraft and others make conical inductors, which start off on an incredible 0.1 mm pitch, to give some high frequency inductance, and go up to perhaps 1 or 2 mm to give the necessary total inductance.
At 1090 MHz a quarter wave is only 40 mm on PCB, so you are entering stripline filter territory. Here is a very nice filter by W1GHZ for 1296 MHz which you could probably scale and trim. A VNA for this frequency range is now under $50!
-
$\begingroup$ you have some example of filters with discrete parts for MW? Thanks $\endgroup$– mpradioCommented Aug 4, 2021 at 14:19
-
$\begingroup$ No but there are tons of calculators/designers on the web, including ones that will stick to standard values. Start with some specifications though - what would be an acceptable passband loss, and how much attenuation do you really need at 960 and 1710 MHz. $\endgroup$– tomnexusCommented Aug 5, 2021 at 3:29
$\begingroup$
$\endgroup$
To design any filter, you need s parm specs.
Then realization will determine the requirements of stripline and or discrete components based on tolerances and layout.
The higher the Q, the greater need for precision stripline reactances and controlled impedance thru board-shop “electrical testing” by TDR methods on test coupons to validate the Dk assumptions and tolerances for quality control.
here’s an example of a GHz VCO using discrete parts https://uniteng.com/index.php/2021/05/02/2-45ghz-to-2-85ghz-vco-design-with-infineon-bfp420-wideband-silicon-npn-rf-bipolar-transistor/
