2.5GHz RF MCU matching network parts 0201 vs 0402 imperial?

[evb149]

I have noticed that in the past couple of years several of the newer 2.5 GHz RF MCU reference designs are now
suggesting the use of 0201 imperial sized L and C parts for the discrete IC to antenna feed line matching parts as opposed to the larger 0402 imperial / 1005 metric sizes.

If I recall correctly the same vendors some years ago often had reference designs suggesting the use of 0402 imperial sized L and C discrete matching / filtering components for previous generations of the same family tree of ICs.

These are not the particularly tiny small size is everything designs e.g. they're not for the WLCSP BGA parts but even the QFN ones are now commended this way.

I get that the smaller size parts will be more ideal lumped components with less parasitic pad and package L and C stray components and that helps the parts behave more ideally / purely at the fundamental and harmonics so in that sense smaller is in theory a little better.

But I wonder if the parasitic reactance advantages are really all that pronounced so as to warrant using 0201 sized parts in designs where there would be no real reason to use 0201 size parts on the PCBA vs. 0402.

Maybe they just assume that in recent years most manufacturing processes are now commonly good enough and comfortable using 0201s that they just specified the best possible thing that will work nearly ideally and be manufacturable on common lines  even if it isn't very advantageous for size or performance?

Then again at 1.x pF level capacitances and 2-4nF level inductances I can see how one could get more precisely predictable / repeatable results with the smaller parts.

I suppose it is possible that the newer parts might also be more sensitive to performance at the 2nd, 3rd, ...+  harmonics than maybe previous generation designs were in which case the higher SRF and such might be actually found to be beneficial.

What are your practical experiences with 0402 vs 0201 and tuning / matching / EMC / parasitics with 2.5GHz low power digital use cases like BLE, et. al.?

[ConKbot]

Set your transmission line width to the width of your component pads, and your CPWG spacing to accommodate your pad to pad clearance in your passive component footprint. Parasitics? what parasitics

Ok there are definitely still parasitics, but if you're not changing your transmission line width for series components, and no ground weirdness for shunt components, it's about as parasitic free as you can get.

[gamalot]

I don't see any issues with using 0402 size components in RF circuits like BLE, but generally 0201 size components are less expensive and easier to route on the PCB.

Even on a non-RF device like the Raspberry Pi Pico, you can see 0201 components around the RP2040 microprocessor.

[evb149]

Yes, that's sensible.

Actually I would think that would tend to work better at 0402 size than 0201 because of the
ordinary good width of the transmission line on FR4 L1 vs L2 probably being more akin to a 0402 pad
size than an 0201 pad size.

Also because sometimes they suggest for 0201 tombstoning prevention things like round pads that are not oversized or in the middle of copper regions without pad exit thermal relief so there is less pad area and thermal heatsink for the 0201 to influence paste reflow / spreading.   Harder to make it a seamless pad to ground plane and pad to transmission line interface if one does those things to isolate and minimize the size of 0201 pads.

Anyway I am not too worried about avoiding or preferring 0201 so I'm happy to use 0201 though it's certainly more mechanically inconvenient to do assembly & rework on if swapping components out during tuning trials etc. than 0402 would be.

Set your transmission line width to the width of your component pads, and your CPWG spacing to accommodate your pad to pad clearance in your passive component footprint. Parasitics? what parasitics

Ok there are definitely still parasitics, but if you're not changing your transmission line width for series components, and no ground weirdness for shunt components, it's about as parasitic free as you can get.

[evb149]

Yes indeed.  If the SOC package was some WLCSP 0.5mm or whatever pitch BGA or one was designing other parts on the board with DRAM and application processors and such one would most likely be using 0201 anyway for decoupling, termination, etc. if there are size constrained areas for fanout / buses where that made sense.

It was just odd to see it suggested in a design where a QFN-40/48 or similar at 0.5mm pad pitch is probably the most dense / complex part on the whole application PCB.

I don't see any issues with using 0402 size components in RF circuits like BLE, but generally 0201 size components are less expensive and easier to route on the PCB.

Even on a non-RF device like the Raspberry Pi Pico, you can see 0201 components around the RP2040 microprocessor.

[E Kafeman]

It is much what space that is available and how demanding needed tuning stability is and how important need for high efficiency is and actual need for impedance ratio transformation.
If the antenna have an very high or very low impedance relative to needed impedance, say in ratio 50 or more will discrete component losses be critical.Smaller is in general not better from efficiency view but it is better in terms of low VSWR.
VSWR looks better but is less effective antenna for smaller inductor.Capacitors have not same type of problem and are in general less critical in its value related to its size.
Antenna efficiency is mainly of interest when available battery power is an limited factor. 50% higher efficiency and  TX power can be reduced in same amount without losing performance.
If available antenna space is small -> increased need for tuning - > more critical doing a good component type and topology selection.Earpice and headset are typical applications where a very small battery should last many hours, have almost no space avilable for tuning components or antenna.

As a comparing example, where selecting type of component have high impact on resulting VSWR did I select a groundplane 80x10 mm with an antenna in short end. Antenna is a 5 mm long wire. Did select some common inductor types from Murata and did select optimized values for each type for a simple network, a indcutor and a capacitor in parallel to ground.
For some inductor types can another type of topology be to prefer and antenna untuned impedance is a bit extreme due to it is short but have seen much worse for real headsets.


 
 
 
Yellow curve in Smith chart  and its table shows measured untuned antenna impedance.

[E Kafeman]

Just to clarity above VSWR curves so that nobody thinks LQP02 is a by me recommended choice.
 It is the most bad choice of above three alternatives and will result in an almost dead antenna!
 LQP02 which is the smallest of above compared inductor sizes. results in nicest VSWR curve but it is due to component losses.
Antenna function will be almost none. If antenna is disconnected will it still shows a decent VSWR.
LQP15, size 0402, is a better choice, but better not select this type of inductor at all for this type of matching.

Probably are practical losses even bigger then this as ground  layer losses are critical due to the relative low inductive value.

Results was shown for Murata LQP-series which are inductors of film type.
It can be ok to use this type of inductors but as a impedance matching network often have function to also be mismatching for harmonics and improved ESD protection, for best result use as big components as possible, preferable wire-wound of good design.LQW15 and even LQW18 (0603) are wire-wound alternatives from Murata but other brands are just as usable as long as it exist reliable S-parameters.Stray losses, both inductive and capacitive as well as transmission line delays are measurable parameters at 2.4 GHz and can affect component values for a matching network.

[rfclown]

If I have infinite degrees of freedom, I use 0402 for everything 2.5 GHz and below (unless I'm wrapping coax around toroids for HF). My 60 year old eyes and shaky hands can still comfortably deal with them for rework. Just made the microstrip width near 20 mils. For 5.8 GHz I use 0201 (I hate reworking 0201, but PCB fabs have no trouble). It's amazing how less forgiving 5.8 GHz is than <=2.5 Ghz. (I've not done higher than 5.8 GHz). The problem for me at my day job is that I don't have infinite degrees of freedom. The PCBs are 20+ layers (RF plus bazillion pin FPGAs, etc), so the microstrip width is narrow, and 0201 is the best choice in most cases for components that connect directly to the skinny transmission lines, even if the signals are < 1 GHz.

[evb149]

E Kafeman: Thank you very much for the excellent data and examples!
Great points about component losses / Q vs. size and how that's especially relevant for small narrow band
radiation inefficient antennas which will have a lot of reactance vs. radiation resistance given their size / structure.
For the moment I was thinking about relatively more long / efficient antennas but absolutely the answers to
this choice / design question of component selection will be much more critical as you said using a really electrically short antenna as some designs like tiny USB dongles or ear buds or who knows what must use.

rfclown: Indeed, I agree 100% about the 0402s being much more eye strain & hand precision friendly than 0201s for manual rework.

I meant to get back to this sooner but I'm busy with many things.  I've probably got most 0201s and 0402s
needed to go either way I want to now but I should get a good selection kit for altering the choices experimentally and also some of the parts were out of stock so I've had to chase alternate parts / vendors too so that has taken
some time.

Given the good advice from E Kafeman I should also go back and see more closely what may be better choices for loss on particular L/Cs than what I may have already picked mostly based on value / tolerance and less so on high performance in case I will benefit from that for alternative designs that are less broadband in their antenna matching.