FR-4

My assembly guy says he prefers 0402's with 0.6mm round pads on 1mm centres.

This topology allows them to be on the underneath BGAs (the pads can mirror the BGA pads and leaves room for 0.5mm vias with 0.25mm holes).

He says the way the solder behaves on the round pads gives them a much more reliable process and he convinced me to use this footprint throughout with 0402s.

Nial.

What my people really like is 0805s. We do boards with all 0805s when we can.

John

"John Larkin" wrote in message news: snipped-for-privacy@4ax.com...

I suppose I might go 0603 if I get in the habit of dense, commercially produced boards, possibly with multiple layers. But that'll be a while.

Tim

do they also use a coal fired reflow oven ? :P

almost everything we do is 0603, last project did have few caps in

isn't 0805 starting to be harder to find?

-Lasse

Those are *huge*. We generally use 0603s and are moving to 0402s for common values (1K, 10K, 10pF, 82pF, .1uF, and a few others). The only 0805s we use are some bulk caps, and a few for headphone current limiting resistors.

Yup. Sure is nice.

John

Waste of space. Sometimes it matters, so we're going to 0402s. Production doesn't mind them at all and they'll be nice for moderately dense FPGAs.

Between your ears.

Now Dimmie, I've told you over and over that you're *not* my type. I see you've been in mommy's hamper again. It's time to put her computer, and undies, away and go to bed now, Nymbecile.

In answer to my own question;

I cut out a piece of FR4 2.45" x 2.45" to equal 6 sq.in. The capacitance of 6 sq.in. of this FR4 is 131pf. Capacitance per sq.in. is 21.8pf. It measures 0.062" and the copper is 0.001" on each side, that makes the dielectric 0.060".

I measured the Q of an air core inductor on a Boonton 260A and found it to be 390 at 893 khz. I substituted the FR4 for the capacitor in the 260A and found the Q now measured 68.

I picked a tap on the air core inductor and found a Q of 238 at 2.85 Mhz. I substituted the FR4 for the capacitor in the

I used the following formula to calculate the Q of the FR4. Qc=(C2-C1) (Q1*Q2) / (C1(Q1-Q2))

The Q of the FR4 at 893 khz measured 82.4 and the Q at 2.85 Mhz measured 83.8. MikeK

Neat. I don't have a simple way to measure Q.

So FR4, as a capacitor, has terrible tolerances, a huge positive TC, and relatively low Q.

John

I'll try to get energetic, (after honey do's), I think I have some 0.032" FR4 and I know I have some Duroid Teflon (5880 ?) 0.032". I'd like to check the Q's of those. Any logic to having the same capacitance or the same physical size of my new test subjects? I just happened to have a decent Q coil handy and the 131pf worked, but I could go as high as 50 mhz, with it's own parasitic problems. The Boonton internal capacitor rotates down to 0pf so the FR4 was the only capacitor. I think I'd like to stick at low mhz. MikeK

Same C might keep the Q measurements more honest (he guessed.)

I think FR4 losses go up seriously with frequency. That matters when thinking about ground+power planes in multilayer boards. A plane pair forms a very low impedance (like, 1 ohm) transmission line that is the first level of bypassing for ICs. Added bypass caps help, especially at low frequencies. The thing that keeps this from being the oft-published Spice model of a zillion resonances is the losses of the copper and FR4 (he guessed.)

John

If anyone cares; emclab.mst.edu/documents/TR00-1-041.pdf

"The design of high-speed circuits requires a knowledge of the dielectric constant (?) and loss r

parameter (tan?) of the substrate. This is particularly important when using FR-4 material because of the wide variations encountered among different vendors and even between different orders from the same vendor. It is therefore desirable to develop a method allowing the design engineer to personally determine the ? and tan? of a substrate material. " MikeK

...and that is why Rogers and other PCB makers have formulated high performance materials - and why they are so expensive WRT FR-4.

Depends on what you're doing. If you're not building microstrip filters, or pushing 20 picosecond stuff around, commodity FR4 PC boards can work fine. I do GHz/picosecond stuff all the time on regular boards. If the traces are short, dielectric losses usually don't matter.

RF front ends can lose a bit of noise figure from board loss. Again, keep things chort.

John