Coupling capacitors for microwave frequencies

[LukeW]

Suppose that I want to build a basic reference design implementation around a Hittite HMC474 or similar MMIC gain block, for operation at 5-6 GHz.

DC blocking capacitors are required at the input and output. Supposing that we want the magnitude of the capacitive reactance to be say <= 1 ohm at 5.5 GHz, therefore the capacitance needs to be at least 29 pF.

http://www.hittite.com/content/documents/data_sheet/hmc474mp86.pdf

The datasheet actually recommends 100 pF as the capacitance of choice, for operation in the gigahertz range. So let's call it 100 pF.

However, the series resonant frequency of the capacitor needs to be >= 6 GHz. So my question is this: Where do I source a capacitor of an appropriate capacitance range with an appropriately high SRF? It seems to be unobtanium.

If I do a quick search on Digi-Key for high-Q high-frequency 0402 100pF ceramics, for example, here's one typical result:

http://www.kemet.com/kemet/web/homepage/kechome.nsf/file/E8E10945B57BE93285257989006DD493/$file/KEM_C1030_CBR_SMD.pdf

But we can see in the SRF plot that the SRF hits 6 GHz at a capacitance of only about 2-3pF, which is typical for a couple of other datasheets I've looked at. Even for 0402 package ceramics, to bring the frequency up into the range we need it the capacitance needs to come down well below 10 pF which is too low for this application.

So, any recommendations about where to get appropriate parts? Thanks.

[branadic]

Well, as you already found SRF is a critical point. Kemet is one of the minor companies that have full spice models of there different capacitance series. You can study them in detail to find a type and package size (also one of the values that influences SRF)  that matches your requirements.
For example a 100nF ceramic capacitor in 0402 will have lower resonant frequency (~20MHz) compared to 0603 and 0805, so maybe a 0805 package is the better choice.

In any case you should check the website of American Technical Ceramic:

http://www.atceramics.com/

They have ultra-broadband capacitors e.g. with value of 100nF and resonant free range up to 40+GHz. I have build a programmable attenuator with parts from Hittite and such capacitors from ATC, they are perfect, but expensive.

[VK5RC]

For small quantities of good capacitors may I suggest
1 Down East Microwave in USA, they normally sell kits but are very helpful     http://www.downeastmicrowave.com 
2 RF Electronica  in Europe   http://www.rf-microwave.com/eng/catalogue.html
3 Mini Kits  in Australia,  Mark is very helpful  https://www.minikits.com.au  Mark also  has some "universal" boards that are made of the appropriate substrate with 50Ohm stripling, earthing vias etc for fair prices
Regards Rob

[KJDS]

I'd loosen my requirements up to 5 ohms of impedance unless it was critical. Then a standard 0402 can be used up to a decent frequency. Whilst you'll lose a little signal in a 50 ohm  system it's rare that it's that important. If it is then I end up using one of ATCs series or for narrowband applications it's sometimes possible to print a suitable structure or make use of the inductance in the cap as part of the matching circuit.

[Rufus]

However, the series resonant frequency of the capacitor needs to be >= 6 GHz. So my question is this: Where do I source a capacitor of an appropriate capacitance range with an appropriately high SRF? It seems to be unobtanium.

The SRF doesn't need to be >= 6GHz. At resonance it has the lowest impedance and still blocks DC. Above resonance it looks more like an inductor and still has low impedance and blocks DC.

The SRF is telling you the series inductance and it seems you are stuck with 5-600pH. At frequencies where that is significant I guess you have to take that into account with a matching circuit - but I am an rf noob so don't know much.

[G0HZU]

Be careful with ATC caps and resonances...
I've been using various cap packages from ATC for many years and you can easily get minor resonances when they are mounted on a typical PCB with microstrip. These show up as sharp dips in the passband response. Sometimes the dip can be >0.2dB.

The dips don't show up in their s parameter data partly because of the grainy step sizes in the data and also because the dip frequency will be affected by the PCB dielectric/thickness etc.

If you look at their tape/reel options ATC offer horizontal and vertical versions of the same cap. This is because you can shift the resonance (with the PCB) up in frequency by mounting the part 'vertically'. This really means you mount it on its thin side rather than flat.
Generally, I avoid ATC caps in broadband dc blocks for these reasons. By broadband I mean something covering 20MHz to 2GHz or maybe 1.5GHz through to 6GHz. Basic/cheap COG ceramic caps are better for use with MMIC amps etc. But you can use ATC if you use a value that keeps any resonance out of a narrower band.

It's the resonance mode the ATC cap sets up with the PCB that will sting you if you want no subtle dips in the response anywhere. Sometimes cheaper is better

If you do use a lower cost cap and it shows a resonance with the PCB then try mounting the cap on its skinny side to see if it shifts the problem out of band

[LukeW]

So you're saying that any old cheap C0G (NP0) ceramic may be usable? (Even though in most cases they don't have SRF plotted in the datasheets at all, or if they do it's well below 6 GHz).

OK, and I should aim to get the SRF right at 6 GHz, not above 6 GHz, for optimum results.

"For example a 100nF ceramic capacitor in 0402 will have lower resonant frequency (~20MHz) compared to 0603 and 0805, so maybe a 0805 package is the better choice."

That's funny, I always thought it went the other way - in other words using smaller components is generally what you want at high frequencies because of these parasitic effects.

[SArepairman]

can you mount the capacitor at a 45 degree angle so you get the best of both worlds??

[G0HZU]

I think you can experiment with ordinary ceramic caps in 0402 package. The caps will have a tiny bit of loss up in the GHz region but nothing to be concerned about in a typical MMIC application.

I'd start with 100pF 0402 but then experiment with various cap values (even quite large values) to see which works best on your PCB layout. However, your real design concern at 6GHz will be how well you ground the MMIC device on your PCB. You will need a lot of fine (plugged?) via holes around and under the device in order to preserve performance (and stability) up at several GHz.

can you mount the capacitor at a 45 degree angle so you get the best of both worlds??

For a one off test circuit I guess anything goes (if it helps to shift a resonance out of a wideband circuit) but it isn't a viable solution for mass production

[G0HZU]

To illustrate the issues with the PCB design/grounding etc, I had a quick rummage through my test amplifier stash and I can show you images of a couple of typical MMIC PCBs from Hittite.

See below for the official eval boards from Hittite for the HMC474 and the HMC311. Sorry about the poor image quality but you can see that the PCB uses lots of via holes and also has mounting holes for a heatsink.

I'm pretty sure the 474 board has 100pF 0402 coupling caps on it. Sadly, I only have a 3GHz VNA here at home so I can't test these up at 6GHz.

What isn't shown in the images is that the PCB is multilayer, with a thin low loss dielectric used for the top RF layer and a thicker FR4 backing layer to add mechanical strength.

[SArepairman]

I think you can experiment with ordinary ceramic caps in 0402 package. The caps will have a tiny bit of loss up in the GHz region but nothing to be concerned about in a typical MMIC application.

I'd start with 100pF 0402 but then experiment with various cap values (even quite large values) to see which works best on your PCB layout. However, your real design concern at 6GHz will be how well you ground the MMIC device on your PCB. You will need a lot of fine (plugged?) via holes around and under the device in order to preserve performance (and stability) up at several GHz.

can you mount the capacitor at a 45 degree angle so you get the best of both worlds??

For a one off test circuit I guess anything goes (if it helps to shift a resonance out of a wideband circuit) but it isn't a viable solution for mass production

my post was meant to be a troll but im glad to know that microwave circuit builders have the option in some cases to put parts at a 45 degree angle!

i could see some guy, with his 150,000$ a year salary being presented with a problem. He takes out his butane soldering iron, wedges a piece of wood under a decoupling capacitor and then solders it down. then a bus load of unskilled labour is brought in to do this on a mass scale.

[G0HZU]

 

Seriously though, my experience has shown that ATC caps are generally a poor choice for coupling caps for ultra broadband design. I wouldn't use the ATC600S series in the Hittite boards above because I'd expect to see various suckouts even with 100pF caps.

I can only measure accurately up to 3GHz here at home but even a cheap 100pF cap in 0805 is a far better choice than the ATC 600S (fairly similar size to 0603) 100pF cap as a broadband coupling cap to 3GHz.

For example, I have various component test jigs here and one of them tests on a PCB with Rogers 4003C 0.02" dielectric.

I tested a few caps this evening...

A 600S 100pF cap shows a -0.2dB suckout at about 1700MHz and a nasty one at 3GHz.

By contrast a cheap and basic 100pF 0805 cap taken from a 20 year old dev board has no suckouts up to 3GHz even on its flat side. It has about -0.15dB loss by 3GHz but it is a nice smooth and progressive loss all the way up to 3GHz.

I also tried a cheapo 1nF 0805 cap from the same 20year old board and it was slightly better than the 100pF 0805 cap with only 0.1dB loss at 3GHz.

That's why you need to experiment with coupling caps. At work I have a library of plots taken of various caps up to 6GHz on various PCB thicknesses and dielectrics to speed up the design process.

Sometimes cheaper is better...

I can remember a fair bit about which caps are good and which are bad because I've been doing fairly critical RF design for nearly 25 years so I cover this ground fairly regularly

[eb4fbz]

For cost sensitive microwave designs under 10GHz i use Murata GJM1555 series. Murata has a very nice online tool with basic plots and s2p export: http://ds.murata.co.jp/software/simsurfing/en-us/index.html You can even look at the series S21 and S11!

For coupling capacitors you want the lower possible series impedance along your frequency range. Unfortunately, the SRF limits the maximum capacity before the capacitor start to act like an inductor. Working near the SRF is not a problem, but usually the capacitor parametrics have the board effects de-embedded, and in practice the SRF will be a bit lower due to ground plane and PCB substrate underneath. That's why i usually select capacitors with low enough impedance but "bare" SRF at least 10% higher than my maximum frequency.

For 5-6GHz i would use a 1.8pF capacitor like GJM1555C1H1R8CB01 from Murata with a bare SRF arround 6.5GHz. |S21| will be under -0.05dB with better than -25dB RL.

Sometimes you have to be carefull with out-of band impedances too, because it could turn an amplifier unstable.

ATC has some nice broadband and Hi-Q capacitors, but they are very expensive. HiQ: I only use them for very low loss LNA input circuits or high power amplifiers. Broadband capacitors (those 100nF with SRF higher than 20GHz) are very cool, but their loss is higher, although tolerable in some circuits like very high frequency laser modulators or some special uW amplifiers topologies for ensuring it's stability or even power supply bypass in multiple stages amplifiers with stability issues.

[G0HZU]

Yes, Murata make some good capacitors
 
ATC do make some low voltage (eg 16V) broadband caps in X7R dielectric that would probably be OK with a low voltage MMIC. However, I would still recommend testing them on the chosen PCB type.

However, there are also quite a few cheap X7R caps around that can deliver good broadband performance.

Here's some plots from my quick tests on my test fixture at home. This is on Rogers 4003C 0.02" material. You can see the issues with the 600S capacitor. I still use the 600S series for filter design and also over smaller bandwidths because they have very low loss and can be produced with very tight tolerance and can work at high voltage.

The plots are quite noisy but I didn't bother with any averaging or narrow IF bandwidths in the VNA. Rather than save them to floppy etc I imported them into Genesys via GPIB so that's why the plots are in Genesys format.

NOTE: If I turn the 600S cap on its side then the suckout at 1700MHz disappears but the one at 3GHz is still there. For applications up to 2GHz the 100pF 600S is an excellent performer when mounted on its side (vertical) but it isn't suitable for ultra broadband applications.

[G0HZU]

The other thing to be wary of if you try the big value X7R coupling caps (eg >100nF) is that they can cause issues with MMIC switches coupled to MMIC amplifiers as they can cause problems with the switching time if you want to switch really quickly (eg in ballpark 10us)

This is because of the high capacitance of the coupling caps and the changing dc levels each side of them as the MMIC switches are operated. The 'effective' RF switching time can slow down dramatically because of this but it depends on the MMIC switch type. IIRC some Hittite MMIC switches are seriously prone to this effect.

[branadic]

Noone mentioned Mica caps up to now, what about them instead of X7R caps?

[G0HZU]

I have an old 6GHz HP8753ES on my bench at work at the moment so I measured a cheapo 100nF 0402 cap on a text fixture yesterday lunchtime. This is 100nF 0402 and not 100pF 0402.

You can see below it is pretty good up to 3GHz and the match then degrades and the loss then goes to just under 0.4dB by 6GHz. I also measured various ATC 600S caps and the one with the lowest loss at 5-6GHz was a 0.9pF cap. This had about 0.08dB loss at 6GHz according to the VNA but there will be a degree of measurement uncertainty associated with that figure. Also the loss was only low over that range of frequencies so this cap would be useless as a broadband solution.

The obvious risk with using a cheapo 0402 capacitor is that the performance will vary according to manufacturer and the various specs of the capacitor. But the plot below shows you can get reasonable performance to 6GHz from a cheapo cap.

It also works down to about 100kHz before the loss dips below about 0.1dB.

 

[Rufus]

I have an old 6GHz HP8753ES on my bench at work at the moment so I measured a cheapo 100nF 0402 cap on a text fixture yesterday lunchtime. This is 100nF 0402 and not 100pF 0402.

And what was the SRF of that cap? About 20MHz?

[KJDS]

I have an old 6GHz HP8753ES on my bench at work at the moment so I measured a cheapo 100nF 0402 cap on a text fixture yesterday lunchtime. This is 100nF 0402 and not 100pF 0402.

And what was the SRF of that cap? About 20MHz?

Have a think about what the SRF actually means in terms of thruogh loss. Then have a think about how the Q of that SRF affects the through loss and you'll understand why the plots look as they do.

[Rufus]

I have an old 6GHz HP8753ES on my bench at work at the moment so I measured a cheapo 100nF 0402 cap on a text fixture yesterday lunchtime. This is 100nF 0402 and not 100pF 0402.

And what was the SRF of that cap? About 20MHz?

Have a think about what the SRF actually means in terms of thruogh loss. Then have a think about how the Q of that SRF affects the through loss and you'll understand why the plots look as they do.

I asked about SRF because others in the thread seem to think it is relevant and needs to be above 6GHz.

[G0HZU]

There's some good info about the significance of SRF and PRF in the link below:

http://www.johansontechnology.com/technical-notes/rf-capacitors-a-inductors/srf-a-prf-for-rf-capacitors.html#.Usc7ebmYZY4

You can see they mention the 'vertical mount' trick that reduces the number of resonant suckouts. In reality, something as complicated internally/externally as a PCB mounted 0402 100nF cap can't be judged like this up at several GHz.
 
Experience has taught me to experiment with caps for broadband dc blocks. Sometimes you get a surprise how good/bad
some caps are and the broadband 'winners' won't always be the ones with the highest SRF on a datasheet.

I always test caps for resonances, even a tiny 0.1dB resonance can be a problem in a transmitter. However, it's no big deal with a low power, low voltage MMIC amplifier unless you want to use the amplifier as a reference gain block
with no unwanted dips in the response.

I also tested an ATC 600S 100pF yesterday. You can see the plot below. The cheapo 100nF 0402 looks to be a far
better choice as a broadband cap for a low power MMIC.

It would be interesting to compare against the $$$ ATC equivalent of the 0402 100nF cap but I don't have any. I'd
hope it would be better than the cheapo 0402 cap but I'd want to see it for myself on a typical PCB based test fixture
I'll see if I can get some samples at some point. It's a shame they have such a low voltage rating though...

See below for the 100pF ATC 600S tested on 0.02" Rogers 4003C PCB material.

[G0HZU]

For completeness see the plots below for a cheapo 0603 100pF cap and also the plot for the 0.9pF ATC 600S.

In the case of the 0.9pF ATC 600S cap the package inductance (and my less than perfect soldering?) has placed the Series Resonant Frequency very close to 6GHz and this therefore gives the lowest loss at around 6GHz. Note: I did expect a 1p8 to give the best result and I initially tried a 1p8 then a 1p5 and the next one available to me was 0.9pF. I may not have soldered the 0.9pF cap down very well but it gave the best response at 6GHz. But it at least shows that you can get very low loss at 6GHz over a narrow band with the ATC600S series caps

[LukeW]

To illustrate the issues with the PCB design/grounding etc, I had a quick rummage through my test amplifier stash and I can show you images of a couple of typical MMIC PCBs from Hittite.

See below for the official eval boards from Hittite for the HMC474 and the HMC311. Sorry about the poor image quality but you can see that the PCB uses lots of via holes and also has mounting holes for a heatsink.

I'm pretty sure the 474 board has 100pF 0402 coupling caps on it. Sadly, I only have a 3GHz VNA here at home so I can't test these up at 6GHz.

What isn't shown in the images is that the PCB is multilayer, with a thin low loss dielectric used for the top RF layer and a thicker FR4 backing layer to add mechanical strength.

Great, thanks for that.

Personally I look at the pricing of those sorts of evaluation boards and wonder how the hell they set the pricing at about 100x the IC price sometimes. Maybe the fancy hybrid PCB stackup accounts for a lot of that.

Here's a picture of one of my experimental builds, still a work in progress, that I'm mucking around with.

I do have some small plugged groundplane-to-groundplane vias surrounding the ICs and the 50-ohm coplanar waveguide lines. However my via density is a bit more spaced out than on the Hittite reference boards.

The coupling caps are generic NP0 0603 100pF ceramic, and the inductors are some "well that looks like it might be OK" candidate 0603 wirewound RF inductors I could find on DigiKey. But I thought I'd just build it with easily obtained, cheap but probably suboptimal parts and see what happens. But I think on the next board revision I should change the 0603 footprints for the caps to 0402 and use 0402 parts.

The main thing I want to decide what capacitor package size I should use for the PCB layout - 0603 is currently used but I think I should move to 0402. Then I can fabricate PCBs and keep experimenting with different caps. Different caps are cheap and easy to change, different PCB footprints are a little less so.

I haven't finished building this yet, haven't tested it, I still need to order the GP2X+. (Other than the stock carried by Mark at MiniKits in Adelaide, a lot of Minicircuits stuff is hard to get your hands on Down Under, I think. They wanted to try and charge me about $100 postage for about $100 order, ordering directly from the US website. lol, not bloody likely.)

I don't have any access to a VNA, so in terms of testing things it might just be a matter of pointing it at a neighbour's 5.8 GHz WiFi AP retroreflector target at a certain range, with the VCO sweeping, and see what sort of IF signal comes back and see what happens.

The board is 2-layer FR4 and the CPWG geometry is set accordingly, aimed for about-50 ohm. Yes, FR4 is a bit lossy at microwave frequencies, however a gain block MMIC or three to offset the loss is probably much cheaper than finding a board house to fab the boards (on a small hobbyist scale) out of Rogers or PTFE or whatever unobtainium RF dielectric. I'm not an expert in design at these frequencies by any means, still learning a lot, but the cheap, ubiquitous access to FR-4 in the PCB fab industry seems like a big advantage economically if it is possible to get a design to work successfully using it.

I see on the Hittite reference board the smallest of the 3 parallel caps in the DC bias circuit is a 100pF 0402. I assume they've just used an identical cap from the same batch used for the signal coupling caps, tested and confirmed to behave well at the frequency of interest. I did the same thing on my board, using a cheap 0603 NP0 100pF, as well as a 1nF 0603 and 2.2uF 10V 0603. (Hopefully substituting the tantalum used on the Hittite ref design for a ceramic does not make any difference.) On a future design revision I guess I'll change the 100pF cap on the power rail to 0402, consistent with the coupling caps, and keep the other two as 0603.