65C51 up to his old tricks again

[Chromatix]

If you're building missiles, do it properly - with surface-mount parts.

[GARTHWILSON]

Chromatix wrote:

If you're building missiles, do it properly - with surface-mount parts.

Yep. For the last several years, I've been consulting at a company that makes propulsion units for small satellites (although I recently told them I'm not going to do any more for them, because of the work environment). They do everything in SMT. It all has to be super reliable. There's no human life on the line, but it's very expensive to launch a satellite, and then if something fails, you can't just mail it back home for warranty repair.

I do have ten 4MHz CMD 65c51's in PLCC, but they're not for sale! LOL

[BillO]

Interesting. Why would SMT be more reliable than through-hole? I'm not doubting, just not seeing how it is more reliable. More compact and lighter I can see.

[GARTHWILSON]

I'm not sure they're saying SMT is particularly more reliable than thru-hole, just that it's not any worse, and the size constraints, and the availability of certain parts, dictates SMT. The electronics I was involved with are for the propulsion units' power supplies, controls, telemetry, sensors, etc..

[floobydust]

SMT parts and DIP parts are mostly the same, i.e., an epoxy resin encapsulated chip with leads coming out. Basically, no difference. You could take a DIP part, bend the leads horizontal and solder it like a TSSOP package. The only real difference is the physical connection. DIP parts typically have a thin flat blade pin going into a round hole which is larger than the pin dimensions. When soldering, you simply fill the hole with solder to make the connection. With SMT, you have the part pressed against the PCB making a connection, with a small amount of solder to hold the part in place. As solder (lead/tin) is not really a great conductor, using less of it to make the connection is perhaps, more reliable. Over time, I would expect a lower possibility of developing a cold solder joint with SMT versus thru-hole, but having properly cleaned and prepped parts (the contact pins, that is) help eliminate that in general.

[Chromatix]

I was really pointing out the absurdity of the notion that you'd use something believed to have reliability problems to build guided missiles using hobbyist resources. Anyone with a military-grade budget would be buying milspec parts and assembling them using appropriately professional techniques, whether those be SMT or through-hole, not scavenging 20+ year old PLCC chips and complaining about how fragile the $1 socket was. And anyone without a military-grade budget should not be building guided missiles!

[BigEd]

> they already HAD working chips for decades... they only screwed it up later on. it's a simple matter of sending the old plans back to the factory and telling them to make those.

I wouldn't be surprised to find that the bugs came in when they moved from one process to another, needed new masks, and made a new layout. The old process might not be available any more. Fixing the present design means at least one new mask, which is a major cost probably never recouped by subsequent sales.

Now, if you placed an order for say 50k parts, that might make it worth their while.

[drogon]

cb3rob wrote:

yuk. the 28L92 (like pretty much all things NXP) is smd-only. but then again we could buy a machine to bend the pins down and weld a piece to it but that'd make it even more inferior no lolol.

I was looking at a UART for another project and found an chip more suited to me - also NXP, but it's the SCC2692AC1N28 which is dual uart in a 28-pin DIL package. It has 2 output bits and 2 input bits which (I think) can be configured for rts/cts type functions or independent GPIO.

Sadly, it's no longer being made, but there seems to be quite a few on ebay which I'd normally treat with suspicion, but it includes a .us and .fr based sellers, so probably legit.

The other thing I've done recently it put TSOP2 chips on a DIL carrier which was wasn't hard to do with my usual temp. controlled soldering iron:

thats a 256KB RAM chip, but I think the principle is sound if you want to keep the PCB a much DIL as possible.

(And yes, the 'quality' of my soldering there is somewhat suspect, but it's works just fine - I just used ordinary multicore solder there as I'd run out of liquid flux, also if doing this, try to get carriers that aren't pre-made with the pins already soldered in - that made it somewhat tricky to get the part on)


-Gordon

[BigDumbDinosaur]

drogon wrote:

I was looking at a UART for another project and found an chip more suited to me - also NXP, but it's the SCC2692AC1N28 which is dual uart in a 28-pin DIL package. It has 2 output bits and 2 input bits which (I think) can be configured for rts/cts type functions or independent GPIO.

If you can get it, the SC26C92C1N,602 is a better choice, as it brings out all available I/O connections in a DIP-40 package, plus it has eight-deep FIFOs on both channels, in both directions. When I built POC V1 I initially used the 2692, but soon switched to the 26C92 and then the 28L92, the latter which has 16-deep FIFOs in both directions. The (positive) effect on performance was quite noticeable.

Quote:

Sadly, it's no longer being made, but there seems to be quite a few on ebay which I'd normally treat with suspicion, but it includes a .us and .fr based sellers, so probably legit.

NXP stopped producing the 2692 several years ago and in fact, will soon be discontinuing all of the 26xx series UARTs. The 26Cxx and 28L... series are higher-performing products that can be run on a 65xx bus without wait-states.

[Chromatix]

There appear to be second-sources of the 26C and 28L series. The XR88C92 is equivalent to the 26C92, and the XR88C192 is equivalent to the 28L92.

[1024MAK]

Assuming you are using current production parts and they are of a suitable high quality, through-hole wire ended parts or DIP encapsulated chips, when mounted to the PCB with their legs formed to grip the PCB before being soldered are more resistant to vibrations and can survive higher g-forces.

Whereas some SMD components can fly off, crack or break if they are subjected to high vibration levels especially if the PCB can bend.

Mark

[BigDumbDinosaur]

1024MAK wrote:

Whereas some SMD components can fly off, crack or break if they are subjected to high vibration levels especially if the PCB can bend.

I have to say I have never seen an SMT part dislodge by shock forces. In fact, I daresay the amount of force required to do so would not be survivable by most machines...or people.

[GARTHWILSON]

1024MAK wrote:

Whereas some SMD components can fly off, crack or break if they are subjected to high vibration levels especially if the PCB can bend.

At the company that makes the propulsion units for small satellites, where I mentioned consulting, they do the vibration testing, and never have any trouble with SMT components.

For our aircraft products though, I do not use SMT for connectors, since the stresses of inserting, pulling out, or yanking plugs or daughter boards can definitely tear the foils loose. For those, I use thru-hole. I would comment however that the holes must be plated through. We were slow to go to SMT because in the early years of SMT, our small production volumes meant that the set-up costs were too high to compete with the fact that we had our own thru-hole assembly equipment. Now, we've been using SMT for a dozen years. I have never seen any problem with chip resistors or capacitors, or even ICs for that matter. Remember that SOIC and SOJ IC packages' leads have some flexibility, such that board flex isn't going to be cracking any solder or metalization, or tearing foils away.

I never like to the be guinea pig, and I was glad to let others be the ones to suffer through SMT's early problems which have since been solved.

When I worked in repairs at TEAC in the early 1980's, we were frequently replacing thru-hole relays that seemed to have become intermittent. Somehow we figured out that the relays were fine, but that with single-sided boards with no plate-thru, every time a relay was actuated or releases, it would kind of "jump," and gradually the solder around the pin would crack. So of course when we replaced the relay, the solder was removed and the new relay was soldered in. After we figured that out, we started just re-soldering the existing relays and saving the customer a lot of money.

[BigDumbDinosaur]

Chromatix wrote:

There appear to be second-sources of the 26C and 28L series. The XR88C92 is equivalent to the 26C92, and the XR88C192 is equivalent to the 28L92.

A cursory read of the data sheets seems to support your comment. In fact, the XR88C192 data sheet is more lucid than the 28L92 one. The XR88C192 is stocked by Mouser, but not by Digikey, and is somewhat less expensive than the 28L92. Pick yer poison!

[BigDumbDinosaur]

GARTHWILSON wrote:

When I worked in repairs at TEAC in the early 1980's, we were frequently replacing thru-hole relays that seemed to have become intermittent. Somehow we figured out that the relays were fine, but that with single-sided boards with no plate-thru, every time a relay was actuated or releases, it would kind of "jump," and gradually the solder around the pin would crack.

I recall running into a similar problem with a UPS in the early 1970s. The transfer relay that switched from line to battery was mounted on a single-sided PCB and operated with a noticeable "thwack." Eventually the relay's coil connections went south and when the UPS was supposed to switch from battery to line the relay wouldn't pick and the unit stayed on battery...until the battery went flat. The fix at the time was to re-solder the connections with 20/80 solder, whose high lead content made for a stronger joint.