EEVBlog #408 - Schmart Board 0.4mm QFN SMD Soldering

[EEVblog]

Dave.

[rolycat]

I was intrigued by the GreenArrays chip, and did some Googling.

The GreenArrays company is only three years old and was founded by Charles H. Moore, who invented the Forth programming language back in the 70s.

It makes some fascinatingly weird chips. The GA144 mentioned in Dave's video is available from SchmartBoard together with one of their boards, and as far as I can see it doesn't need any other components apart from a power supply and some way to communicate with it.

No crystal, since the processors don't use a clock. Yes, really - no clock of any kind.
No decoupling caps, since it will work without them (essential for any serious use, though).

Each processor node uses a stack-based architecture and only has 64 words of RAM and 64 words of ROM.
And of course they are designed to be programmed in Forth.

The "Green" part of the name comes because, not having a clock, the processors don't use any power (apart from leakage current) until they are asked to do something.

[Winston]

Apply solder paste to groves on card, wipe away excess with old store shopping card (same size and material as credit card, but typically thinner), place chip, solder in a few minutes on a $20 electric griddle.  Their grooves would function as a sort of reverse template and help to position chip.

[JoannaK]

C Moore has been working on thse Forth-cpable tiny chips for couple decades, he even programmed hs own Asic-development system (with forth) to do those.  I have not seen many hobbyist friedly systems build amongst those, but apparently his chips have found their uses at deep embedded.

Some history of those chips, forths etc.. http://www.ultratechnology.com/chips.htm  As a note, there's talk  about MuP21 (or P21), this new core F18 seems a lot similar, even perhaps a bit more reduced.

In the end, I like to see that some people are still trying to keep Bloat out of computing.

PS: Yes,, it is kinda OT here..

[BigBrother]

[EEVblog]

Well there you go, I was pushing the brown stuff up a hill with a pointy soldering iron!
They could have told me that!

Dave.

[Slothie]

Well you actually worked it out for yourself 

Perhaps Schmart should have put that on their instruction leaflet!

I haven't been able to find a source in the UK for the particular board you used in the Video, but proto-pic have some of the 0.5mm and larger ones, for between £6-12 which is pricey but still cheaper and easier than having your own board made, especially if you are "tinkering" and dont have a finished circuit in mind!

They also claim to have ones for BGA parts - if they work then it would be a major win for prototyping. Although I guess most BGA parts operate at very high clocks and timing/crosstalk might be an issue if you are breaking out to a bigger grid.

Ian

[JoannaK]

Just for note.. that particular chip uses most of the chips bottom as a large ground contact, so it should be soldered too. That assuming Dave ever intents to use that chip.

[EEVblog]

Well you actually worked it out for yourself 
Perhaps Schmart should have put that on their instruction leaflet!

Well yeah, I was just following the instructions in letter provided! As that is how Schmart board are supposed to work. If you have to drag solder it, then it's just a another breakout board, albeit one that works pretty well with drag soldering.
I suspect the ability to solder fine pitch QFN's varies greatly with the actual part, hence them bringing out a video showing the "alternate" method.

Dave.

[Slothie]

And was that "0930" on the chip a date code? if so its been sitting around for 2 1/2 years so perhaps it was badly oxidised...

[k3wals]

Thank you Dave for showing this product from Schmart Boards. This will save me a lot of time and headaches of making a prototype board for reverse engineering/probing a display protocol with 0.4mm pitch connectors. I think it would be wise of vendors to provide a discount code for EEVBlog viewers. 

[EEVblog]

And was that "0930" on the chip a date code? if so its been sitting around for 2 1/2 years so perhaps it was badly oxidised...

Most likely.

Dave.

[free_electron]

yep ,the schmart boards work great. i mention them in my book as well. The trick to those boards is in the soldermask. They use a thick soldermask which i suspect is a dry-film and exposed using UV as opposed to a liquid soldermask. Essentially the soldermask is so thick that the solder stays in the ridges. i;d have to look under a microscope but it may even be a solder mask defined pad ..

i use those frequently for quick and dirty breakout. Another product from 'The valley'.

There is another company her ein silicon valley that makes self-adhesive breakouts. they are essentially flex pcb's with solder resist and adhesive. simply peel from the foil , tack where you need it , solder the hip down and it breaks out to nice big pads you can solder chips on. a bit pricey but works great in a pinch.

[codeboy2k]

The GreenArrays company is only three years old and was founded by Charles H. Moore, who invented the Forth programming language back in the 70s.
[..].
Each processor node uses a stack-based architecture and only has 64 words of RAM and 64 words of ROM.
And of course they are designed to be programmed in Forth.

The "Green" part of the name comes because, not having a clock, the processors don't use any power (apart from leakage current) until they are asked to do something.

I am going to have to do something with these chips I sooooo love Forth. I was programming Forth for the Sun SBus boards back in the 90's.. Sun's OpenBoot was innovative for a firmware and it was cool to work in the monitor when trying to bring up a new piece of hardware on the SBus, and you had a full language at your disposal in the monitor, not just a loader and byte poker.

Basically each add-on board on the SBus had Forth code on it, that loaded as a "library module" into the running Forth firmware during the boot. This basically extended the language with more Forth words specific to the hardware and intended for querying and dealing with this hardware.  This Forth code provided methods (known as Forth "words" ) to discover the board, initialize the board, and put information about the board into the OpenBoot firmware.  Then it was attached into the device tree that the firmware was building for hand-off to the  kernel boot.

Writing these firmware extensions in Forth is about the extent of my experience, but it was such a pleasure to work with. Of course I also use HP calculators with RPN so maybe I was just comfortable using Forth since it's a stack based language.

These green chips seem  worthy of some investigation for embedded systems that can be finely partitioned into multiple cooperating tasks.

And did I mention I love Forth? 

[dopplershift]

I've never used a Schmart Board myself, so I can't really comment on them.

But: I've done pretty well hand soldering 0.5mm pitch TQFNs so far, essentially using the drag soldering method. No special PCB, just a 4x magnifying lamp and plenty of flux. Occasionally, one or two pins need re-work, but usually its fine... just check for issues using a jeweller's loupe after soldering (30x or 40x can be good here).

People really don't need to be so scared of SMT and especially QFN packages! Its easy once you've practised a little. Give it a go! 

[janekm]

Even 0.4 mm QFNs solder just fine with a line of paste, hot air reflow and a quick drag touch up with a decent iron in case of too much paste. Much less hassle than this, I must say...
(and just now I've come across the first 0.35mm QFN, and the pads don't even come out to the side so you can't see if it soldered... That's taking the Micky a tad!)

[HackedFridgeMagnet]

I had some boards made by myropcb.com with 0.5 mm DFNs on them. Unfortunately the soldermask didn't make it between the pads.

I thought I had the spacings right, any tips for getting Itead or Seeed to put soldermask between pads at this pitch? 6 mil process.

I drag soldered the boards ok though but one board took me 5 goes.

Sorry about the thread hijack but I have posted this before to no avail.

[janekm]

I had some boards made by myropcb.com with 0.5 mm DFNs on them. Unfortunately the soldermask didn't make it between the pads.

I thought I had the spacings right, any tips for getting Itead or Seeed to put soldermask between pads at this pitch? 6 mil process.

I drag soldered the boards ok though but one board took me 5 goes.

Sorry about the thread hijack but I have posted this before to no avail.

You shouldn't really need the soldermask between the pads on QFN/DFNs... as you experienced it's a bit of a challenge for the board houses. The Chinese ones are getting better though, most of the boards I've had lately with 0.5mm QFNs were just fine (solder mask very well defined between the pads). Avoid strange soldermask colours, especially white which doesn't adhere as well.
The latest board I had made with 0.4mm QFN (made by vpcb.cn) had about 30% solder mask offset on the pads which was a little disturbing but still soldered fine. The recommendation from some IC manufacturers though is to actually remove the soldermask entirely between the pads for 0.4mm QFN parts.

I wrote a blog post about quality differences between the different fabs recently: http://tempeto.com/?p=56
Based on my experiences, I would expect vpcb.cn, hqpcb.cn, pcbcart and storm circuits to handle 0.5mm pitch QFNs with soldermask between the pads just fine.

It is however much preferable to use hot air soldering for QFNs, especially for the first pass (drag soldering works fine for some touchup).

[HackedFridgeMagnet]

thanks for answering janekm

Some good info on your web page. I think I might try Enig finish with Storm or Seeed next time. I have used Seeed a few times and they are so cheap. But maybe I should tryStorm.

I want to try some Arm Cortex chips maybe TQFP64 next time which should be 05mm pitch. I assume QFP is easier than QFN by hand.
http://www.keil.com/dd/docs/datashts/energymicro/efm32tg/d0071_efm32tg232_datasheet.pdf
Luckily they give the solder mask dimensions too.

The recommendation from some IC manufacturers though is to actually remove the soldermask entirely between the pads for 0.4mm QFN parts.

I seem to recall this, why would this make it easier? Flatness? I guess for 0.5mm I should go for soldermask.
Also are you serious about the different colours, I assume you are, I really don't like green, is black ok?
Incidently I have a small amount of experience with chinese manufacturers I had my own mcpcbs made, they were fine but didn't test the manufacturing limits at all as they were for led bases. https://www.eevblog.com/forum/manufacture/metal-core-pcbs-from-zhejiang-zapon-electronic/

This was the thread I referred to in my last post, not really worth reading.
https://www.eevblog.com/forum/eda/msop-clearances-for-seeed-studio/
In the post I was actually talking about MSOP, although in the final design I actually chose a different chip which was 045m DFN with a heat pad. Much harder. When they came back the solder mask was missing.
You're right I didn't actually drag solder that chip I did use hot air,  and it took me a few tries.

[janekm]

thanks for answering janekm

No problem!

Some good info on your web page. I think I might try Enig finish with Storm or Seeed next time. I have used Seeed a few times and they are so cheap. But maybe I should tryStorm.

Storm aren't the cheapest but I use them for my employer and have had about 6 different boards made by them, all excellent quality. They also communicate very effectively.

I want to try some Arm Cortex chips maybe TQFP64 next time which should be 05mm pitch. I assume QFP is easier than QFN by hand.
http://www.keil.com/dd/docs/datashts/energymicro/efm32tg/d0071_efm32tg232_datasheet.pdf
Luckily they give the solder mask dimensions too.

I'm familiar with the Energy Micro chips, very nice devices! The peripherals in particular are very nicely and cleanly designed. I have only used the QFN and BGA packages from them but they have been no trouble to deal with.

QFP is indeed more practical to solder with a regular iron (I guess the flexibility of the leads ensures that they make contact with the pads on the PCB, with QFN it's easy to get contact with the PCB on one side and floating on the other). All the same, with hot air QFNs really aren't that difficult to solder.

The recommendation from some IC manufacturers though is to actually remove the soldermask entirely between the pads for 0.4mm QFN parts.

I seem to recall this, why would this make it easier? Flatness? I guess for 0.5mm I should go for soldermask.

It's to allow for the solder mask alignment tolerances when making the PCBs; having solder mask partially covering your pads could turn out very bad especially in volume manufacturing, so it's much better to leave the clearance. Leaving 0.1mm clearance results in no solder mask between the pads at 0.4mm pitch.

Also are you serious about the different colours, I assume you are, I really don't like green, is black ok?

I haven't had any issues with black and I do use it when it's needed for cosmetic reasons. The only time I have had real issues was when using white solder mask (there's an example picture on the article on my blog) and I believe the cause is the need for pigment to be added to the paint with white. All the same, green is the standard so if you use green it will go through the factories standard process which will be the one that is most tested.

Incidently I have a small amount of experience with chinese manufacturers I had my own mcpcbs made, they were fine but didn't test the manufacturing limits at all as they were for led bases. https://www.eevblog.com/forum/manufacture/metal-core-pcbs-from-zhejiang-zapon-electronic/

That post is very interesting actually, I have been curious about the cost of those metal core PCBs! Not that I have any need for them currently but it's always good to know.

This was the thread I referred to in my last post, not really worth reading.
https://www.eevblog.com/forum/eda/msop-clearances-for-seeed-studio/
In the post I was actually talking about MSOP, although in the final design I actually chose a different chip which was 045m DFN with a heat pad. Much harder. When they came back the solder mask was missing.
You're right I didn't actually drag solder that chip I did use hot air,  and it took me a few tries.

I have had quite good luck lately using a line of solder paste (which invariably ends up being too much solder), reflow with hot air, followed by a quick drag solder to get rid of the excess solder.
If I absolutely need to ensure reliable soldering (and have the time) I use a different method using solder balls (as used for reballing of BGA chips), effectively turning the QFN part into a BGA part. It takes a lot of time but is very reliable, I should write up the method as I haven't seen it described anywhere.

[HackedFridgeMagnet]

Thanks for the tips janekm
No doubt I will start a thread when I need some advice on my Energy Micro Project. I need to step up from 8 bit.
It will be sort of smart meter + relay controller for quite low power mains. I haven't gotten the go ahead yet, fingers crossed.

[gnif]

Thinking of soldering these small parts, I have not done any solder mask soldering yet, but what about buying a set of masks for common component packages, that are just large enough to use them, then re flow with a hot-air station? Then you could get any breakout board and just do a quick paste application and re-flow it.

It might be nice to be able to buy a nice small set of generic reusable stainless steel set of masks for single chip footprints that you keep in your toolbox just for this kind of work.

[mrflibble]

I had some boards made by myropcb.com with 0.5 mm DFNs on them. Unfortunately the soldermask didn't make it between the pads.

I thought I had the spacings right, any tips for getting Itead or Seeed to put soldermask between pads at this pitch? 6 mil process.

I've got a similar board that I'll send to itead or seeed. QFN-32 with 0.5 mm pitch and a thermal pad. And because I'm a cheap bastard an engineer that does design optimization I'll be using a HASL finish if I can get away with it. It will be just a prototype, so if HASL puts too much solder on the QFN pads I don't have a problem with having to clean up the pads first with some flux + solder wick in a worst case scenario. I don't expect too much trouble with using HASL however, but that may be overly optimistic. We'll see.

You shouldn't really need the soldermask between the pads on QFN/DFNs... as you experienced it's a bit of a challenge for the board houses.

...

The recommendation from some IC manufacturers though is to actually remove the soldermask entirely between the pads for 0.4mm QFN parts.

...

It is however much preferable to use hot air soldering for QFNs, especially for the first pass (drag soldering works fine for some touchup).

About that, what's the general procedure here? Just define a big filled rectangle on the top solder layer? If so, what size rectangle do you typically use? Currently I have a 6.5 x 6.5 mm rectangle on the top solder layer for the QFN-32 part, so hopefully that'll be about the right amount of bare pads. Is that the correct way to go about this?

Edit: mmmh, found some more info in this thread. The idea of still having some solder mask, but using a 0 (or 0.05mm) size solder mask expansion might be another option. The default being 0.1 mm solder mask expansion, which results in DRC violations for minimum solder mask sliver size. If I understand things correctly, 0.05 mm expansion would give me just enough margin to compensate for errors in solder mask registration, while still getting the benefit of solder mask to prevent the chance of bridges. Correct?

[dopplershift]

Hey mrflibble, I got my boards done at iTead and used 0.8mm boards with HASL for 0.5mm pitch QFN parts with no issues. I put a big via (3-4mm) under each part, to use for soldering the thermal pads. This seems to work ok, but as in Dave's recent video, you may be able to do it using several smaller vias to conduct the heat through the board.

As for soldermask, I just used the default footprints from Eagle (except for the big thermal pad vias)... and then I ignored all the DRC errors that came up inbetween the pads. I figured "what the heck, its just soldermask" and they turned out fine. No complaints from iTead either.

Personally, I'd try iTead again... but their free shipping was really slow (4+ weeks to Perth, Australia).

[mrflibble]

Thanks for your reply! Okay, it's good to know you got it working without any special action required for 0.5 mm QFN.

I take it you used 0.8mm (as opposed to 1.6mm) boards to make soldering the thermal pad from the bottom a bit easier? As for vias, I was planning to do 9 smaller vias (0.35mm hole/0.6mm diameter) in the thermal pad. I'll probably use hot air + a dab of paste, depening on what is easier.