Use bleeder resistors in audio opamp design?

Looks like standard double sided, guillotined PCBs to me. Could've specified milling them (or whatever they do for multi-layer) like the main board?

Grant.

I don't recall the fab notes. Sometimes we say "boards may be sheared to size" for small array type things; makes them cheaper.

John

We use "v-score" - the boards are scored top and bottom but are strong enough to be handled, carried through the assembly machine etc. Then when complete they go through a machine with to rotary "cutter wheels" top and bottom which splits the boards without stressing them. (You can do it by hand but risk breaking delicate parts like chip ferrites).

Last design we used routing all around 3 sides to get a smooth finish, but left the fourth side attached to the panel with v-scoring. That was for very small boards, 5x10mm, 50 to a panel that was only ~3 inches square.

null

inaccurate?

It looked to me like internal planes were peeking out.

We do the same, but I don't recall them looking that ratty. It may be the photograph or even the solder mask color contrasting with the FR-4 matrix.

I have one coming up (should go out Monday morning) that is quite small (.6"x.75", or some such) and I'm a little concerned about the separation process after SMT. It has a 68-ball, .5mm pitch, BGA on it as well as a couple of SC70s, and a bunch of 0402s. The first pass has 7 of these attached to a board that will be used to program and test them. The final board will just have piles of these "stamps".

null

inaccurate?

Here's a tiny breakaway adapter..

ftp://jjlarkin.lmi.net/Break2.jpg

ftp://jjlarkin.lmi.net/OnBoard.jpg

that we had to do to fix a Maxim problem, about 3000 of them.

John

null

inaccurate?

I remember you talking about these. The edges of these don't look as ratty as the others. How did you cut them? It looks like the router cut through plated holes to make the mounting tabs. Isn't that rather dangerous? How did you separate the chiplets? The "mouse-bite" tabs don't look like they made it to the final product.

I'm not sure something that flimsy will work, placing .5mm BGAs. I don't think we can leave the tabs on the boards, either. They have to fit into a DE-9 connector housing. We've already stolen every square mm we can. I'll file these away as a suggestion to the layout guy and ME. Thanks.

null

inaccurate?

view

Seemed to work fine. The half-moons soldered to the old SO8 footprint on the board. Our QC lady suggested the idea!

How did

We have some sort of shear blade, in a press, that chops them up after they are assembled. As you can see in the second picture, they wind up looking pretty clean. I can get details from my production guy if you're interested. He's a genius at processes.

John

null

inaccurate?

view

attached

I did notice that they were cleaner than even the "filters". At first, I expected the mouse-bite tabs to be evident on the final product. We use them to separate boards but they're not nearly as clean as the second picture indicates. If you think of it, sure, I'm interested.

n, and

=A0Since the unit

output

se

in

ter.

al.net

lip

Hey, some people value the filter shape, and some don't. SK is fine for junky filters. Just don't expect much out of them.

unit

That's a canard. You can make a reproducible filter with a S-K, just not a reproducible _sharp_ filter.

Cheers

Phil Hobbs

Mine has this really delicate RH sensor chip, that's really why the carrier PCB is routed around 3 out of 4 sides. So the cutting machine only has to cut one short edge farthest from the chip.

I was thinking of having test tracks too, but since the separation process could potentially cause the fault there's not too much point. Plus its hard to run test tracks over a v-score :)

I might be doing something like that for the sensor chip I mentioned before. I.e. make a "carrier" that itself ends up as a "component" surface mounted to another board.

That first photo looks like the plating is done after routing, do you know if that is how it normally ends up? That is, the cut cross-section of the through holes looks shiny (and solderable).

No, those are ordinary PTHs, sliced in half by the cutter when the breakaway pattern is routed. Works fine.

John

unit

net

Nice point, as I'm planning to make a nice blunt 4 pole Linkwitz-Riley filter, two 2 pole S-K filters in series, unity gain and .7 Q -- hardly a demand on the circuit topology? It's claimed that summing the high and low pass outputs results in a flat response again.

Just a background project I've wanted to do for decades. Got most of the parts now, lay out a PCB... Relearn CAD tools for schematic and PCB layout, using free stuff since this is a hobby for me now.

Right at the start I pointed out my query was to do with audio, hardly precision design compared to the wonderful stuff some of you guys are doing.

Grant.

as

it

I've seen it done many places, most recently a Samsung P&S camera does that for a right angle PCB join, from the main board at the back of the cam, a dozen or so connections to the top board under the shoot, power and mode select switch mount PCB on the top of the cam --> cheap, and no space for connectors in there. Made in China, so Korea exports jobs too?

Grant.

son, and

ps. =A0Since the unit

he output

no

to

oise

n in

ilter.

So

ical.net

clip

lly..

.net

Isn't the canard that you are expecting less out of the SK filter? For instance, if the filter has a sloppy shape but an insanely tight phase spec, then the SK would still suck. I've done sonar filters where the spec is sloppy but the phase matching between filters has to be tight.

And let's leave Burt Rutan out of this!

the unit

output

net

No, the canard is that you implied that I didn't care about the shape of the filter. What I usually care about is the shape of the ultimate filter, i.e. the product of the analogue and digital transfer functions (with suitable frequency normalization, of course).

The ticket is to use a gentle, but nice and repeatable analogue filter with low noise gain, whose job is primarily to preserve dynamic range at the digitizing step, followed by an appropriately chosen digital filter to turn the analogue one into whatever filter transfer function I actually want.

It's all bobbing for decibels, like most signal processing jobs.

Cheers

Phil Hobbs

Bob them with your hammer, and hear them ring... decibels!

Ring!

Ring!

Ring!

I STR an SNL skit with RRD (of course they didn't mention decibels).

No! It was the 'guru' 'priest' guy in "All of Me".

Every time the phone would ring, he would go flush the toilet.

Hilarious movie.

ike

eason, and

amps. =A0Since the unit

the output

or no

et

ors

k to

noise

ain in

filter.

=A0So

ptical.net

hen

ou clip

ch

is

g a

tally..

t- Hide quoted text -

A colleague sent me this...

They use two LP 11 pole filters in the signal chain. It's implied that they are mostly L-C discretes. Corner frequency at 650kHz and down 120 dB by at a 1 MHz, or something like that.

George H.