Components on the back side

t,

It only makes a difference if you are doing auto pick and place. Hand assembly won't. How big is your order?

"Tim Wescott" schreef in bericht news:546dndP0-LFMG-HSnZ2dnUVZ snipped-for-privacy@web-ster.com...

Hardly see any boards that are *not* populated on both sides these days. With SMDs that is. It may have been little more expensive in the past but for todays automated assembly it makes no difference. As for testing, I suppose you to know your test equipment and you have to place the testpoints at the appropriate places. You always had. Only boardlevel repair and reverse engineering become more difficult. Do you care?

Maybe old(er) board designers who are not accustomed to two side population did not like it?

petrus bitbyter

We do it when we have to. It takes another pick-and-place and reflow pass if you put many parts on the back, but it's not a big deal.

Larger parts have to be glued, so they don't fall off during the 2nd reflow pass. Again, not a big deal.

One compromise is to just put bypass caps on the bottom, which simplifies life a little. If you're only 4.9 out of 5, that might work.

The more-layers thing gets nasty. Trace impedances get out of control when you have many thin layers, more than 8 roughly, and eventually via density gets you anyhow. Use less parts!

--

John Larkin         Highland Technology, Inc

jlarkin at highlandtechnology dot com
http://www.highlandtechnology.com

Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators
Custom laser controllers
Photonics and fiberoptic TTL data links
VME thermocouple, LVDT, synchro   acquisition and simulation

Our assemblers say it's no problem. An extra pass through the screen printing and P&P machines is all. I have avoided it too in general, just doing the first one now with a significant number of back-side parts.

On two-layer boards it helps quite a lot with decoupling caps, that would otherwise block track routes from QFPs and the like.

--

John Devereux

If I put all the little passives on the back side of this board, the effort required goes down tremendously.

--
My liberal friends think I'm a conservative kook.
My conservative friends think I'm a liberal kook.
Why am I not happy that they have found common ground?

Tim Wescott, Communications, Control, Circuits & Software
http://www.wescottdesign.com

n,

ast,

ut

e

No problem of course, just more cash ;-)

Actually, i would like caps inside the PCB, so free up space for other ICs.

We're increasingly using resistor packs, 1206-size quads mostly. It makes value selection harder, and makes changes harder, but production sure likes it. Our placement cost exceeds most parts cost by a big factor.

--

John Larkin         Highland Technology, Inc

jlarkin at highlandtechnology dot com
http://www.highlandtechnology.com

Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators
Custom laser controllers
Photonics and fiberoptic TTL data links
VME thermocouple, LVDT, synchro   acquisition and simulation

,

st,

t

We have a board with quite a few big part on one side they are not glued I believe they just run that side last

We also have a board the has smd on top, some dpaks on the back and lots of through hole connectors

the topside is first reflowed, then back with dpacks and connectors is wave soldered

you can always go crazy and use laser drilled micro vias, blind and burried too

-Lasse

What's a "two layer" board?!!

--

John Larkin         Highland Technology, Inc

jlarkin at highlandtechnology dot com
http://www.highlandtechnology.com

Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators
Custom laser controllers
Photonics and fiberoptic TTL data links
VME thermocouple, LVDT, synchro   acquisition and simulation

yes probably :)

Well that is Larkins powerplane theory isn't it? I.e., one larger cap with powerplane = lots of little caps without one.

--

John Devereux

Really cheap^H^H^H^H^H cost effective! Hey, at least they are through-hole plated these days. Really helps stop all the parts falling off :)

With small-medium microcontrollers, I can dedicate one side to groundplane and do almost all the routing on the other. Taking advantage of the ability to remap I/Os to simplify routing where possible. For power rails I can use local decoupling caps and series inductance/resistance instead of a powerplane.

I do use 4-layer for the next step up - anything with an external (parallel) memory bus.

I looked at 8-layer once for a BGA I wanted to use but the cost was prohibitive.

--

John Devereux

We started doing it in the early 80's (with thru-hole work!) when it wasn't mechanically possible to get various components to coexist on the same board. Back then, it was a bit of a chore finding a fab that would take on the work as most weren't geared up to testing that sort of an assembly.

Now, I routinely move components to the backside (or, hide them "under" other components :< ) to make things fit. Memory tends to be a "lots of bang for the buck" choice for this: either put an entire array on one side of the board *or* put half on each side and use "reversed pinout" devices for the other side (locating devices directly opposite each other on different sides of the board).

Aside from decoupling, my designs tend not to have many passives intermixed with the actives -- any field interfaces tend to drift off to separate areas where such devices can more freely comingle.

Look at testing and repair costs before you decide. How many defects (and what types) are you likely to encounter during assembly/initial test? How will you handle warranty repairs? etc.

Also, mechanical issues tend to bite you when you least expect it (perhaps because you tend to forget about the back side of the board when you are thinking about it, in general).

And, finally, thermal issues (probably not as big an issue with

*most* passives but can be significant when you are dealing with highly integrated digital devices).

It's rare that I've done a board with components only on one side for the past 20 years. Assembly houses have no problem dealing components on both sides. You should be able to put 5.5# off stuff in your 5# bag.

There are specialty materials out there that let you have an innermost layer that's all of one mil thick with a reasonably high dielectric constant, so in many cases you can get rid of a bunch of bypass caps.

Of course, John Larkin is correct in that even with a standard board (maybe 10mil inner layer thickness, FR4 material), even get enough capacitance (with very low interconnect impedance) that in most cases you don't need nearly as many bypass caps as some app notes suggest. E.g., one bypass cap per power pin on a large FPGA is very much overkill, IMO.

But it's not just by-pass cap. LCD controller wants storage cap. Serial drivers wants +/- charge pump caps. Max3232, for example, needs 4 caps for 1 chip.

I have used cap arrays, although they are rarer and more expensive than resistor arrays.

--

John Devereux

We just bought a few Mastech bench power supplies. There's a big transformer inside, about 6 PCBs, dual LCDs, nice steel box. Amazing for the price. The main board is single-side phenolic with jumpers.

A VME board like I pictured, 6 layers, BGA capable, is about $80 at 15 pieces.

--

John Larkin         Highland Technology, Inc

jlarkin at highlandtechnology dot com
http://www.highlandtechnology.com

Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators
Custom laser controllers
Photonics and fiberoptic TTL data links
VME thermocouple, LVDT, synchro   acquisition and simulation

Well, you can expect killer prices from a company name Manson.

We average three or four caps *per voltage* on FPGAs. I've seen boards with three caps *per power pin*, hundreds of caps per FPGA.

I've never done a multilayer board that had too little bypassing, so even three caps per rail must be overkill.

I saw a board once with about 3000 bypass caps, turret-feed p+p machine. I wonder what the world record is.

--

John Larkin         Highland Technology, Inc

jlarkin at highlandtechnology dot com
http://www.highlandtechnology.com

Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators
Custom laser controllers
Photonics and fiberoptic TTL data links
VME thermocouple, LVDT, synchro   acquisition and simulation