I'm going to make a few prototype boards with RF DAC and/or DDS components. I want to put down pads for an output reconstruction filter that I would stuff later to cover anything in about 4-50MHz. I've never tried to make an RF filter using SMT components before.
What's a good package size for typical RF inductors? It looks like
0805, 1008 or 1210 would cover the hundreds of nH to few uH range. Is the Q of these inductors a big issue?
Are there some rules of thumb for designing filters in surface mount?
Yes, definitely. The other often overlooked factor is the self resonant frequency (SRF). Best is a virtual tour through the Digikey offerings and looking at lots of the datasheet links.
The usual, such as avoiding mutual coupling. This is easily violated when placing them too densely. It is also often beneficial to rotate neighboring inductors 90 degrees to each other.
What I do a lot is ask my layouter to provide two popular pad sizes for each part, at least for inductors.
One of my favourite suppliers of inductors (especially RF chip types) is coilcraft (there are, of course, many more) because the data is easy to get on their web site.
You'll at least get an idea of what is available and in what packaging.
1008 is a good size for a DDS filter in this sort of frequency range, since there are lots of 1008's in the catalogs. For, say, a couple of hundred ohm terms, you'd be looking at roughly the 2-20 uH sort of range. Q depends on what you want in terms of flatness, but doesn't matter in most situations. I've found that, if you're going anywhere near Nyquist, a combonation of inductor losses and sinc distortion tend to make things roll off at the high end, and you can often compensate that by underterminating (overterminating? too many ohms, anyhow) the typical lowpass filter.
DDSs tend to have a lot more personality than is first obvious.
Check out Mini-Circuits or Synergy (there are others, also) to see if they have any "off the shelf" filters that fit your needs. Their filters are generally inexpensive, too. If you have to roll-your-own, be sure and study micro-strip-line techniques in making RF PWBs, and don't spare the grounding vias.
Note: I do not work for Mini-circuits or Synergy, nor am I a sales rep for them. I have used quite a few of Mini-Circuit's products (including filters), though.
I'm sure you know this, but for the OP: It also depends on (or is determined by) how much insertion loss you're willing to accept. For a DAC or DDS output, you generally have power to burn, but for RF inputs low-Q coils on even 3rd or 4th order filter can often generate 3+dB insertion loss, which is an awful lot of power to throw away on a high-performance receiver.
I'd suggest just designing the filter for a higher frequency than originally calculated might be a better approach -- I've seen Chebyshev and other "rippley" filters *grossly* exceed their ripple specs if they're significantly mis-terminated (e.g., 200 ohms instead of 50).
Although if you're using low Q inductors anyway they'll smooth out all those ripples (which is generally a good thing, since otherwise component tolerates would also tend to create ripple spec violations) and you probably don't have anything to worry about.
If you really wants to get into this, most SPICEs come with "finite Q" inductors, or you can just add your own resistors to simulate it (or get fancy and build a new part).
I started to ask about amplification stages in my original question, but as you say it doesn't really apply when you have a strong signal to begin with. Do receiver front ends in SMT tend to use lower order filters with more gain stages?
Looks like the best plan might be to pick one of their kits with a good range of values, since they group them by size. My Q and size questions were more closely related than I realized. The Q of 0603 inductors doesn't get out of the teens, but 1008 is 50+ up to the 1uH range.
Synergy is a very generic term. Do you have an URL for them.
Mini-circuits has a great selection of LPF and HPF. I always get a kick out of a Mini-Circuits order since they never ship in a mini-box. I've been told never to ship anything in a box that somebody could put in their pocket, but Mini-Circuits takes that a bit too far. ;-)
Zverev's book has guidelines on how to tweak the filter design to account for the Q of the inductors. I can't speak for the effectiveness of the technique, as I've only used the ladders in the book as prototypes for active filters.
The good filter books (like Williams) often have tables for "pre-distorted" or "lossy-L" filters, namely designs that are flat with finite-Q inductors. But that's usually for a small set of responses. DDS filters are often ellipticals, and it's unlikely you'll find a predistorted prototype for them.
IIRC the AD9954 DDS eval board I played around with back in '04... (Analog Devices) used an elliptical filter for the reconstruction output filter. Might want to read their app note on the subject 1st...
The later ADI datasheets, appnotes, and eval boards seem to be pretending that the filter issue doesn't exist. That sort of makes sense, as the filter often costs more than the dds chip and takes up six times the board area.
A 3rd-order bandpass filter using as high-Q inductors as the size constraints allow followed by an LNA is typical. This is generally a tradeoff between getting decent selectivity with a small enough insertion loss to have a negligible impact on the receiver's overall sensitivity. I've done a receiver with a 5th order front-end filter, but it was due to having an awfully wide passband and the insertsion loss (a couple of dB) was painful.
This is for fancy prosumer/commercial/military radios, though -- cheap consumer grade radios often get away with little (e.g., 1st order) or no filtering other than that provided by the antenna/transistor amplifier input load combination! Alternatively, some radios use a tuned front-end with just a single L-C "filter" that tracks the radio's desired tuning; this is often done mechanically -- you've probably seen high-quality old radios where they have 3+ ganged variable capacitors that all tune together for the ultimate in selectivity.
I'm actually too young to know how to build really good receivers, though. :-) These days everyone wants high-speed data channels that are so wide that thermal noise is huge compared to circuit noise; in many ways I believe the RF parts of something like an 802.11 chipset is easier than a really high-performance broadcast band AM/FM receiver; the folks who designed stuff like the Carver TX-11 know a lot more about building receivers than I presently do. Someone like Ulrich Rohde could design circles around me... :-)
These days most people "cheat" and just have an optimizer tweak the component values to account for finite Q's, PCB parasitics, etc... at least if you have a fancy RF design package like Genesys, Microwave Office, ADS, etc.
You still end up doing some tweaking on the physical board... or at least I do, since I'm unwilling to try to spend the time modeling ALL the PCB parasitics in a simulator.
And having his book "Communications Receivers" is indespensible for the serious RF engineer. It's right underneath the bible here (the real one). The only thing I don't like about it is that ugly greenish cover. That sticks out like a sore thumb, messes up the "ambiente" of my office. Oh well, a few more hot summers and it'll blend in.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.