Advice for capacitors selection for RF filter

[Sp4]

Hello!

I am planning on building a 40M-10M band pass filters similar to the design from qrp labs website: https://qrp-labs.com/bpfkit.html and trying to understand how to pick the capacitors. I do understand about C0G, but SMD / through hole is an open question. I am planning on using Multilayer Ceramic Capacitors. At 1st I want to build the board, tune it with the trimmers, extract the trimmers, measure them and replace with the 1-2 constant ones in parallel with the prepared pads one at a time to reduce the element of surprise. I am not trying to accommodate any size constraints, just don't want to have any trimmers in the final build I put in the use in the field.

I am not planning on pushing through them much power, maybe 10W and TBH not really planning to use in transmission at all. For transmission I will be using low pass filters, but the questions will remain the same.

So the questions I have:
1. are MLCC capacitors the right choice for the job?
2. do I have to use through hole or SMD will work just fine?

Looking at mouser - MLCC in SMD can go into much smaller capacitance way cheaper making more precise substitution described above much easier task. As for max voltage - SMD capacitors come in high values as well, good enough to push 50W if I use 200V+.

Appreciate any advice.

Thanks.

[T3sl4co1l]

MLCCs are fine, maybe air on the larger side (0805, 1206?) just in case.

C0G has extremely low loss; they're about as close to ideal capacitors as you can get, short of vacuum.  Check of course, but typically Q is in the thousands, easily handling this power level.

High Q inductors are a lot harder to come by, probably using air core of much larger size, or powdered iron toroids (#2?) of modest size (T37?).  Litz isn't much use at these frequencies; you'd need strands about as fine as available, but special build (hollow core surrounded by strands?) may still be required.  Q for round wire coils isn't bad though.

Component Q limits overall filter efficiency (insertion loss, particularly at the band edge, necessitating further change), and sharpness/bandwidth (probably not important for L/H pass, but BPFs are limited in BW directly).

MLCCs are better for having lower stray L, but THTs are plenty usable at these frequencies.

Tim

[Solder_Junkie]

The QRP-Labs filters use "high-quality RF ceramics of the C0G type (a.k.a. NP0 - near-zero temperature coefficient)" as mentioned on their web page.

The QRP-Labs kits are not bad value, but beware that while the type of band pass filters they use are suitable for most home constructors, they have a fairly shallow response to reject out of band image and IF signals, around 40 to 50 dB, depending on chosen IF, etc. With more elaborate filter types you can easily obtain 60 dB to 70 dB figures. If you also use a low pass filter, as in a typical transmitter or transceiver, depending on the IF and local oscillator frequency, you can achieve really good image rejection. Also note the QRP-Labs band pass filters are not suitable for use on the output of a transmitter, except at very low power levels of a few mW.

It is interesting that QRP-Labs use NP0 fixed capacitors, then add foil trimmers... Overall, I don't think temperature stability is an issue for home use, although Elecraft did have some trimmer temperature related issues with the filters used in their K1 and K2 kits.

As the QRP-Labs article mentions, a wider filter has less loss and will be less critical on exact component values (and temperature variations). You can play around with filter design using the free Elsie program https://tonnesoftware.com/elsie.html Note the default Q values in Elsie may be higher than typical components you are likely to be using, so can give misleading results... Try a Q of 200 for the inductors and perhaps 500 or more for capacitors. Murata data sheets show a Q factor of >1000 for some of their multi-layer SMD NP0 caps, but not all manufacturer data sheets show Q. T50-2 and T50-6 cores are likely to result in Q factors in the range of 200 ~ 250. This page gives details:
https://elnamagnetics.com/wp-content/uploads/library/Micrometals/Iron_Powder_Cores_for_High_Q_Inductors.pdf

This table was produced from Elsie. https://www.qsl.net/g4aon/pdfs/g4aon_bpf.pdf myself and two other hams have built 7 sets of the these band pass filters. The receiver was the first I built and the SSB TX is the later one with some updated ideas, construction details: https://www.qsl.net/g4aon/

73 SJ (G4AON)

[TimFox]

Remember that "MLCC" is a type of construction.  NP0/C0G is a dielectric that can be assembled in MLCC or disc construction.
In both SMT and THT, other dielectrics (e.g., X7R or Z5U) give higher capacitance in a given size, but have higher loss and non-linear capacitance, so not a good choice for signal applications.

[fourfathom]

I'm using surface-mount 0805 NP0/COG caps in the filters and "filter-combiners" (similar to a multi-band diplexer) for my WSPRSONDE multi-channel propagation-study beacons.  The power level is about 1W per channel, and the frequency range is 1-50 MHz (ham bands, but not necessarily).  Inductors are the hard part, the caps are easy.  I use surface-mount inductors when I can, but when I need a higher Q or higher power capacity I use hand-wound iron powder toroids -- usually the "2" or the "6" mix.

It is interesting that QRP-Labs use NP0 fixed capacitors, then add foil trimmers... Overall, I don't think temperature stability is an issue for home use, although Elecraft did have some trimmer temperature related issues with the filters used in their K1 and K2 kits.

I've built some QRP Labs stuff (the QDX mostly), but nothing with a trimmer cap.  I would guess that this cap is to compensate for winding and core material variations in customer-wound toroids?  As you say below, yes, the narrower the filter the more critical the tolerances and "Q" are.

As the QRP-Labs article mentions, a wider filter has less loss and will be less critical on exact component values (and temperature variations). You can play around with filter design using the free Elsie program https://tonnesoftware.com/elsie.html Note the default Q values in Elsie may be higher than typical components you are likely to be using, so can give misleading results... Try a Q of 200 for the inductors and perhaps 500 or more for capacitors. Murata data sheets show a Q factor of >1000 for some of their multi-layer SMD NP0 caps, but not all manufacturer data sheets show Q. T50-2 and T50-6 cores are likely to result in Q factors in the range of 200 ~ 250. This page gives details:
https://elnamagnetics.com/wp-content/uploads/library/Micrometals/Iron_Powder_Cores_for_High_Q_Inductors.pdf

I've only had tempco issues when designing oscillators.  These days I use a synth chip referenced to a TCXO or OCXO or GPSDO if I want stability and accuracy, but back in the day I had a full assortment of capacitors with various temperature coefficients.

This table was produced from Elsie. https://www.qsl.net/g4aon/pdfs/g4aon_bpf.pdf myself and two other hams have built 7 sets of the these band pass filters. The receiver was the first I built and the SSB TX is the later one with some updated ideas, construction details: https://www.qsl.net/g4aon/

73 SJ (G4AON)

Thanks for those links!  I didn't realize that the filter topology I'm using for my transmitter filter-combiners was called a "mesh capacitor-coupled bandpass"!  This has some nice characteristics when used in a combiner -- the input / output Z goes high out of band so you can connect multiple filters together at the common output with minimal interaction (depending on filter bandwidth of course).  Also, I drive my transmit filters with 1W squarewaves, so having the series inductor input reduces loading of the third and higher harmonics, which reduces the current (and heating) in the output stage drivers.

I also use SMT C0G caps (0603 mostly) in various filters and preamps used at the input of SDR and other receivers.  None of these are extremely narrow, and of course the power levels are low.  I use surface-mount inductors for these, and these inductors are the limiting factor (Q, self-resonant frequency).  I use 5% tolerance inductors and capacitors, and design filters that can tolerate this.

I wish I could answer the OP's question about power levels when using these capacitors.  I don't think that loss is much of an issue at the 10W level, even with 0603 parts (the published loss specs for NP0/C0G are very good).  Voltage ratings are of course important, but the small-value caps have a good available range.

If you're curious, I have a document on my website where I give a simplified review of these filter topologies and design issues:
https://turnislandsystems.com/wp-content/uploads/2024/04/FC-2-1.pdf
And here's the site: https://turnislandsystems.com/

[TimFox]

Power rating:
30 years ago, when I was using Ceramite NP0 disc ceramics in RF applications, a factory engineer told me to calculate the power dissipation from the ESR (computed from the Q value) and the RF current, and gave me some power values for different disc package sizes.
Perhaps you can estimate a power capability from those quoted for SMT resistors?

[fourfathom]

Power rating:
30 years ago, when I was using Ceramite NP0 disc ceramics in RF applications, a factory engineer told me to calculate the power dissipation from the ESR (computed from the Q value) and the RF current, and gave me some power values for different disc package sizes.
Perhaps you can estimate a power capability from those quoted for SMT resistors?

That would be my approach, I just haven't needed to worry about this at my low power levels.  If I start designing for higher power I'm definitely going to do the calculations.

[Sp4]

Wow, people, you are super awesome!!!!

Thank you all soooo much!

My take is - I can use SMD capacitors with NP0/C0G with enough margin for high voltage rating with 0805+ size to be on the safe and comfortable side and inductors are much bigger of a trouble I have to deal with 

I will order a few of each and run some tests. Now at least I know that I don't have to be that crazy about favoring one package type over another and can freely combine them based on what I have in my parts box

And some links you gave me - I have never seen before, which is super useful as well.

You are amazing.

Heaps of thanks.
Sergey

[LM21]

Can  Elsie  or other filter design software  tell the  currents in filter components at  some power level? Or only in LTspice?

By the way. making coils and links to formulas
https://hamwaves.com/inductance/en/index.html#input

[fourfathom]

I like Elsie, but use LTspice to get a decent idea of the currents in these filter components.  And here's a very nice on-line filter calculator:
https://markimicrowave.com/technical-resources/tools/lc-filter-design-tool/
I'm working in the HF and low-VHF region, and LTspice has been quite good at evaluating my filter designs (especially if I put in some guesses for stray PCB capacitance, etc.)  I often use CoilCraft SMD inductors because they have good specs and they provide Spice models that are pretty accurate.  This helps a lot.

[Solder_Junkie]

Kyocera manufacture SMD NP0 caps that are used in the Elecraft KPA500 (500W) linear amplifier low pass filter. Manufacturer part no 1206GA680JAT1A
An alternative may be muRata GRM31A7U3D680JW31D, although that is not NP0.

These are 1206 size, 2KV rated, 68pF.

Data sheet: https://datasheets.kyocera-avx.com/AVX-HV_MLCC.pdf
And: https://www.mouser.co.uk/datasheet/2/281/1/GRM31A7U3D680JW31_00B-3156863.pdf

SJ

[ftg]

The muRata GRM31A7U3D680JW31D seems to have U2J dielectric, which is a new encounter for me.
According to Kemet U2J does not have major change in capacitance from DC bias, so it should not distort at least.

muRata's Simsurf parametrics also make it seem decent.
https://www.murata.com/en-eu/products/productdetail?partno=GRM31A7U3D680JW31%23
 

[Sp4]

Kyocera manufacture SMD NP0 caps that are used in the Elecraft KPA500 (500W) linear amplifier low pass filter. Manufacturer part no 1206GA680JAT1A
An alternative may be muRata GRM31A7U3D680JW31D, although that is not NP0.

These are 1206 size, 2KV rated, 68pF.

SJ

Agree, but that muRata is already pushed into the consumer grade part. While GQM22M5C2H680JB01L is GQM (high power High Q) is x5 the price, the question we start asking ourselves whether that price difference is justified with respect to the fact that inductors are much bigger contributors of the Q reduction.

muRata caps are amazing, no questions there, for high-end devices they "a must" for sharpening the filter skirts. Do you think I will be able to appreciate that high Q in a home brew implementation?

[Sp4]

Remember that "MLCC" is a type of construction.  NP0/C0G is a dielectric that can be assembled in MLCC or disc construction.
In both SMT and THT, other dielectrics (e.g., X7R or Z5U) give higher capacitance in a given size, but have higher loss and non-linear capacitance, so not a good choice for signal applications.

great High Q capacitors with NP0/C0G go only up to 150-ish pF. Sometimes they are required in the filters according to the online calculators. I presume using a few of them in parallel (or stacked) is a better approach than "downshifting". Never had such experience TBH

Thanks,
Sergey

[TimFox]

Kyocera/AVX, following normal industry specs, measures their C0G surface mount caps up to 1000 pF at 1 MHz (1 kHz for > 1000 pF) with a minimum spec of Q > 1000 for C > 30 pF.
For large C values, self-resonance is probably more important.
Another good dielectric is “porcelain”, such as ATC’s 100 series (now part of Kyocera), which have Q values up to 10,000 at 1 MHz and high SRF, with P90 tempco.
They are available up to 5.1 nF with specified SRF.

[joeqsmith]

ATC former engineering and sales started Passive Plus. 

https://passiveplus.com/

[Sp4]

ATC former engineering and sales started Passive Plus. 

https://passiveplus.com/

Thanks.

It looks great, but no idea about their pricing or distributors....

[fourfathom]

Remember, at the OP's 10W level there's probably no need for fancy dielectrics.  Here's a test I just did, running a 20V P-P 10 MHz squarewave into a 50 Ohm load (2W), passing the signal through two 0.1uF X7R 50V 0603 surface mount caps.  The caps are on a small PCB I use for DC-blocking (both ground and signal).  These X7R caps have a Dissipation Factor of <10%, and the NP0's we've been discussing have a spec of <0.1% -- 100x better.  I used a thermal camera attachment on my cellphone:

My 1W power amp:


The DC-block board:


The DC-block thermal (I can barely see a very small amount of heating on the caps):


For comparison, the thermal image of the dummy load:

[T3sl4co1l]

The muRata GRM31A7U3D680JW31D seems to have U2J dielectric, which is a new encounter for me.
According to Kemet U2J does not have major change in capacitance from DC bias, so it should not distort at least.

muRata's Simsurf parametrics also make it seem decent.
https://www.murata.com/en-eu/products/productdetail?partno=GRM31A7U3D680JW31%23

AFAIK, Q and bias are comparable to C0G; the traditional type is N750.

Tim

[TimFox]

The various “N” dielectrics are also Class 1, with reasonable Q and higher dielectric constants, so you get more capacitance in a given package.
They were popular in disc ceramic, since the largest NP0 devices were too low for many purposes.
An interesting circuit used a “differential” variable capacitor with an N220 and an NP0 or mica capacitor to adjust the temperature co-efficient of an LC oscillator.

[T3sl4co1l]

Remember, at the OP's 10W level there's probably no need for fancy dielectrics.  Here's a test I just did, running a 20V P-P 10 MHz squarewave into a 50 Ohm load (2W), passing the signal through two 0.1uF X7R 50V 0603 surface mount caps.  The caps are on a small PCB I use for DC-blocking (both ground and signal).  These X7R caps have a Dissipation Factor of <10%, and the NP0's we've been discussing have a spec of <0.1% -- 100x better.

Note that this isn't very representative, because 10MHz is well above the frequency where any voltage drops across the capacitance at all; series resistance and ESL are dominant.  With ~no voltage on the capacitance, no amount of current can cause loss due to dissipation factor (shunt loss); and then current isn't very high to drop power across series loss.

In a filter, voltage on the capacitance is significant, particularly in the transition band where Q of the filter multiplies applied voltage.  Typically the dissipation factor of type 2 capacitors is several percent; in addition to nonlinearity and tempco, they're not very suitable for signal filtering (or, more than basic/crude purposes: EMI, power supply, DC, low precision analog, etc.).

A coupling capacitor is also a filter, but generally the cutoff occurs out-of-band, so losses, distortion, etc. aren't important, and this is also a good application for "poor quality" capacitors -- includes electrolytics for low frequencies, etc.  Indeed they're a common sight in audio and video circuits (video because the bias level needs to be consistent from frame to frame; bias is reset at vertical refresh (sync pulse)).

Tim

[TimFox]

Also note that the capacitor current in a resonant circuit can be higher than the input/output currents.
Coupling capacitors only need to exhibit a low impedance to do their job—it need not be constant for many purposes.

[fourfathom]

Tim, Tim-
Yes, you are both right. power dissipation and circulating currents aren't the same as in what is essentially a bypass application.  But still, current is current, and current² times ESR is still power.  2W into 50 Ohms is 200 mA.  I thought my little test was instructive.

[T3sl4co1l]

I thought my little test was instructive.

It is-- but just to make clear of what and why it is.  Since you posted it as example in a thread on RF filters, it would be improper for someone to get the impression that that type will work nicely for any kind of filter -- it is a good example, of a coupling network when signal response in the transition band is irrelevant; but that much narrower application needs to be clear.

Tim

[Solder_Junkie]

Generally, band pass filters are used at low levels of RF, maybe up to +10 dBm. Low pass filters used on the output of a transmitter can have potentially damaging currents and Voltages, it is interesting that the Elecraft KPA500 (500 Watt) HF/6m linear amplifier uses 1206 size SMD capacitors in the low pass filter.

Unfortunately, there is no service manual for the KPA500 and only sketchy details can be found on the web. Details of other transmitters/amplifiers can be found from more customer focused companies, such as Icom. You can download the service manual for the Icom IC-7610 (100W HF/6m) transceiver from https://www.manualslib.com/download/1542112/Icom-Ic-7610.html
The part numbers for at least the low pass filter capacitors are 1206 size SMD by Murata, C844 is GRM31A5C2J560JW01D (56pF, 630V, NP0, around $0.25 from Digikey). Take a look at the manual, it is refreshing that manufacturer part numbers are listed.

SJ