Better, cheaper breakout boards!

Breakout boards for the most common SMT footprints are everywhere. SOIC and TSSOP? Everyone and their dog has those for sale. We're not going to try to compete with those. Unfortunately, these days the most exciting parts don't always come in those convenient easily soldeed packages that we all love. They come in "no-lead" packages. To make matters worse, while the name still identifies the overall shape most of the time (though many manufacturers have in-house names to refer to largely standard packages), and the number of pins, for sizes smaller than TSSOP, the package name is no longer specific to identify the actual footprint! There are, for example, at least 4 different packages called QFN12, QFN20, and QFN28! It's a common misconception that the added first letter relates in any way to the package footprint. A QFN20, VQFN20, WQFN20 and UQFN20 may all exist, but the initial letters like T, V, W, and U have nothing to do with the footprint. Those stand for Thin, Very Thin, Very Very Thin (VV looks like W, and so that's what's often used), and Ultra Thin)...

However, an M (as in MSOP) does say something - though less than you might like: an MSOP is a "Micro Small Outline Package", smaller than an SSOP (Shrink Small Outline Package) which is in turn smaller than an SOP ("Small Outline Package") - though MSOP does not identify the package uniquely; you know an MSOP-16 will be smaller than a TSSOP-16, and that it has leads, unlike a DFN-16. But will the pin pitch be 0.65mm? Or 0.5mm? You'll have to refer to the package drawings in the datasheet to figure that out...

The picture as relates to DFN (Dual Flat No-lead) is similarly murky. There are DFN's available with many pin pitches, though only three different pitches are commonly seen in the wild - 0.65, 0.5, and 0.4mm. Each of those might be found with or without an exposed pad between them, and where present, the exposed pad width (and hence the width of the whole package) varies between devices.

It is, to use a technical term, a total shitshow, and an inexperienced buyer can easily end up with a breakout board (or worse, a custom made PCB) with a footprint that doesn't match their part. Sometimes the same part is even made in multiple packages with the same basic "name" (ex, VQFN-20), but different pin pitches. I ran into this the hard way, when I failed to notice that the ATtiny841 was made in two different VQFN-20 packages: One with 0.5mm pin pitch, and another with 0.4mm pitch. My boards came in, and the chips arrived, and I eagerly started trying to assemble one... and then realized that I'd used the 0.5mm, 4mm x 4mm footprint... but ordered 100 pcs of the smaller version. Ooops. (That's what happened to Rev. D of the tiny841 breakout....)

As mentioned above, the waters are further muddied by vendor-specific names: Some companies like to call DFN parts things like SON/TSON ("[Thin] Small Outline No-lead"), or refer to some of their QFN packages as MLF. "MLF?, huh?" That's the "Micro Lead-Frame", which describes the internal construction most QFN parts. I'm sure if one happened to know the inside development story for those vis-a-vis various technologies, a company's naming patterns would make some level of sense - but for everyone on the outside, this is just more muddying of the water. Nobody cares about the precise details of the manufacture of their chip as long as the chip works. Do I care that this chip is made using a micro lead frame? (no!) From the outside, though, it looks pretty random.

So - breakout boards

Unsurprisingly, there are specialist suppliers who manufacture breakout boards for a great variety of these wacky parts. I have been profoundly disappointed with their product. The best known of these is a certain manufacturer who I won't mention by name, with what seems to be an incredible selection of breakout boards. At the exorbitant prices they charge (sometimes as much as $10 per breakout board), you'd think you would be getting carefully designed boards, with attention paid to the details that make the difference between an easy to use board, and something that, well, maybe you could solder it with a good hot air reflow station and a microscope.

You'd be wrong - what they (and the other major manufacturer of obscure breakout boards) sell is actually pretty lousy. Exposed pad? If they bothered to include it, it's solderable by hot air or reflow oven only: there's no provision for getting heat to it otherwise, and if the manufacturer specifies that the exposed pad has to connect to something, you're out of luck: even when present, they didn't have the courtesy to connect it to any sort of termination: It's just an island of exposed copper in the middle of the footprint. The footprint itself? Well, that's going to be the canonical one given in datasheets (usually without the EP, as noted). This footprint is designed to be used with parts placed precisely by machine, and soldered in a reflow oven. You're not going to be using a pick and place machine to place the parts for your prototype onto a breakout board, and there's a good chance you won't be using a reflow oven either...

That ground plane that the manufacturers of parts often describe as a necessity? Keep dreaming. The manufacturing quality of the boards? Far from inspiring. Really nothing about them stands out - they're mediocre boards with pads prone to lifting off the PCB (more so than my boards, at least) - At least they (mostly) have bothered to use ENIG surface treatment. It's my opinion that these breakout board manufacturers are taking advantage of their customers, who have few options - either buy their breakout board, buy a likely overpriced eval board, or hire a PCB designer to make a custom board to mount the part on. Until now...

Introducing Azduino prototyping breakout boards

We've reimagined the essential features of a breakout board to address the common challenges of working with parts in wacky, fine-pitch, no lead packages, to facilitate rapid assembly and minimize the misery that comes with no-lead parts.

• The pads have been extended modestly outwards (depending on available space), generally by at least 1mm beyond the edge of the package, often more. This aids in soldering with an iron, as it helps to conduct heat to melt the solder under the chip, and provides a place to "drag" excess solder to away from the chip.
• Most of the pads are originally specified as being wider than the pin or or what passes for a pin on no-lead parts on the package. This makes great sense in automated assembly, but is a liability when hand-soldering. 0.4mm pitch pads on the chip are normally 0.2mm wide, while the standard footprints use 0.25mm wide pads on the PCB. So now to form a bridge that shorts two adjacent pads, the bridge needs to span only 0.15mm, instead of 0.2mm. That's not an improvement! We keep then no wider than the terminations on the chip, making bridges easier to fix and less likely to occur in the first place.
• All QFN packages that may have an exposed pad have one - and there's a hole in the middle to the other side of the board. That provides a thermally conductive path between the back of the board (which you can get to and rest a soldering iron on) and the pad upon which you need to have solder melt to connect to the EP.
• Similarly, DFN footprints with an exposed pad extend that pad outwards, beyond the edge of the chip, making soldering this pad easy.
• All boards have a copper pour on both sides of the PCB, connected to any exposed pad, so you can do what the manufacturer say, and connect the EP to the ground plane.
• Many chips require external components to be connected between certain pins, and often warn that these parts should be placed close to the chip. That's very difficult do do when theres a single hole for each pin on the chip... The most common example are decoupling capacitors, the importance of these capacitors to ensure proper functioning is hard to overstate, and I've read studies where the authors with fancy measuring tools demonstrated essentially identical results when they were omitted and when they were just inches away, and the drastically different (better) results when they were right next to the part. It's not public domain, so I can't reproduce their graphics here, but it was dramatic. So you really would like the decoupling capacitors on the same circuit board, right? And if a crystal is needed, you need both loading caps and a crystal. Crystals are a whole other level of The distance is not always as critical, such as for R's and RC's for timing or setting current limits, but it could still neaten wiring to put all the related components onto the same PCB. Our solution - each pin is brought out to two holes. Typically you'd have one of these connected to pin header, and the other connected to any external components, There's abundant exposed copper around these holes, so surface mount passives can be used too.
• In the case where crystal is used, we strongly suggest not putting pin header on those pins. One set of holes for the crystal, one set for the two loading caps - perfect. Nothing should be connected to a crystal pin that doesn't need to be. In the quest to lower power consumption, the power used to drive crystals has fallen. And without that extra energy to throw around,
• To further facilitate connection of external components, except where noted, parts with less than 28 pins have an exposed ground trace (tied to the ground plane) between the holes and the edge of the board, giving a convenient way to connect (say) capacitors or resistors to ground which so many parts want. Be sure to also connect the the ground pin(s) to these ground stripes too, as they'll otherwise be floating planes instead of ground planes. And that's no good. The presence of these stripes is constrained due to the dimensions of the board - the pricing structure of my board house is highly non-linear, such that widening the board to permit the stripes on the wider boards would significantly raise the cost.
• Any breakout board without ground stripes has several connection points to connect the EP to ground. Do remember to do this, if the parts require the center pad to be grounded
• The mounting holes are (except where noted) tied to the ground plane, so you can solder a wire there, or even use a brass standoff and conductive, grounded enclosure if that's appropriate. If instead you need the mounting to not be grounded, nylon standoffs and screws are readily available.
• All boards except where noted are ENIG (Non ENIG boards are being sold at a lower price; the next order is going to be ENIG, and possibly with a ground stripe shoehorned in. I'll certainly try to find a way to pull that off.
• Some combination of the two rows of holes can be used to match the 0.6" spacing of wide DIP packages, so it's easy to match up with holes in prototyping board and solderless breadboard, allowing for rapid design iteration. Note that there are some exceptions where the board was too narrow for 0.6 - (But.... these can use narrow DIP sockets!) or a design mistake was made and the spacing is off by 1/20th of an inch (these are discounted while supplies last, and to be replaced with corrected ones).

Products

DFN-like

Note that if the exposed pad is about the same size, and the pin pitch is the same, you can use a smaller DFN than the board was designed for - that is, a 0.5mm pitch DFN6 could be mounted on the DFN16A (in fact, two of them could even be mounted on the same board, unless they are very odd shaped DFN6's.

The following shorthand is used: 0.5p, = 0.5mm pin pitch, 0.5EP = 0.5mm wide exposed pad.

The pads extend outward farther than would be needed to accommodate the width of any plausible part. The catch comes from the EP. If the EP on the chip is much wider, then the pads on the PCB for the pins will short to the exposed pad on the underside of the chip, and if the opposite is true, the the pins will reach the center pad of the PCB with the same effect. Luckily the widths do not vary continuously - there seem to be a few widths that get used by many manufacturers, to within 0.1mm of each other - for the most part. There are always a few companies that just have to do things differently.

Other names for these packages include WSON (Wide Small Outline No-lead - often for wider packages, though dimensions vary), TDFN (Thin DFN), and sometimes more exotic names like XSON-8 (often a gigantic 8-pin no-lead package used for large capacity flash chips - they definitely couldn't call this a DFN, as the pin density is quite low; non-leaded package is used for reasons other than a need to make the overall product smaller) and USON (not sure what the U stands for. "Unusual" would be accurate, but not great marketing. DFN breakouts are also suitable for MSOP (Micro Small Outline Package), VSSOP (Very-thin Shrink Small Outline Package), and SM8 packages. None of these packages uniquely specify one pin pitch - there are 0.5mm and 0.65mm MSOPs, for example, and the presence or absence of an exposed pad on the parts with leads depends on economic decisions and the design requirements of the manufacturer.

Notes:

DFN6 has two footprints on side A, a SOT-23-6 and 0.04mm DFN-6 with no EP.

DFN8 Side B has a DFN8 0.65mm pitch with no EP. This also fits a SOT-23-8.

* DHVQFN16 is NXP's pet package. It is a strange hybrid of a QFN and a DFN, 2.5x3.5mm in size, and is not used by other manufacturers. Bizarely, the EP is most often tied to Vcc. To cater to this possibility, there are 8 bridges in the corners - cutting the 4 next to either set of pins will sever the connection between the ground plane and the stripe on that side, and the mounting hole is not tied to the ground plane.

** Due to a design mistake, the current stock of these boards has a row spacing not divisible by 0.1, and so does not connect cleanly to 0.1" spaced prototyping board.

*** this cannot be fit into a wide DIP socket... but it will fit a narrow DIP socket! Also, on SOTDFN6, there rows of holes have 6 pairs of holes on either side, and each footprint is by default connected to only one side. You can connect the two by bridging the solder bridge pads if desired, or if you happen to need both a DFN6 and a SOT23-6 part in the same project, you can use both sides and not bridge the pads (and really, what project doesn't need a SOT23 part for something?). Bridging the pads may also be useful if making some sort of an inline adapter that would go in the middle of a cable with a buffer, or similar device being mounted on the board, where you might want to connect wires to both sides of the board (pro tip: connect the wires to the board on the edge opposite where they came in, and hold them in place against the back of the PCB with liberal application of ~hot-melt glue~ thermoplastic adhesive (never say you're going to "glue" something, always "utilize adhesives" - glueing something sounds unprofessional, but adhesives? You must know what you're doing!) But seriously, the glue-as-strain-relief method is used very heavily in industry for exactly this purpose - open up most old VCRs and things of that sort, even kids toys with lights on widely distant parts of the toy, and you'll see wire soldered to the PCBm and then stuck down with glue and/or looped through a slot in the PCB. This will prevent them from fatiguing and snapping during normal handling, which it is normally prone to, as many of you will know from experience (the solder wicks up into the wire which makes the wire rigid, but there is a sudden transition from flexible to non-flexible some distance up the wire and that is a natural stress point. Hot Glue is very good at preventing this, as it has a fair amount of give to it that it won't just create a new stress point, but will prevent any flexing from happening at the most vulnerable point).

QFN

Where there's more than one model for a given pincount, we try to maximize the sum of the difference between sizes of the parts on the two boards. So given pitches of 0.4, 0.45, 0.5 and 0.65, we'd pair the 0.65 and 0.45, and the 0.4 and 0.5. This minimizes the chance that one might try to use the wrong side and not notice until you'd made a mess of it. There's not much more maddening than using a reflow oven to solder a a 0.45mm pin pitch package onto a 0.5mm footprint. You'll think you're so close to being able to make it work, but even though it looks so close... no amount of heroic rework efforts short of removing the part and putting it on a proper footprint will be successful (speaking from bitter experience). By not putting the most easily-mistaken-for-eachother sizes on the same board, as long as you ordered the right board, you are guaranteed to not have two packages of near-identical size on opposite sides, begging you to get them mixed up.

* this cannot be fit into a wide DIP socket... but it will fit a narrow DIP socket!

** Due to a design mistake, the current stock of these boards has a row spacing not divisible by 0.1, and so does not connect cleanly to 0.1" spaced prototyping board.

A few notes on specific boards

QFN28C is has the rectangular package with 8 pins on two sides and 6 on the other two. Analog Devices likes it.

QFN32B has a QFN footprint on one side, and a QFP footprint on the other; that's the most common 32-pin flatpack package. Breakout boards for QFP-32 are hardly difficult to find, though most of them are worse than these. What makes these unique is that they provide an exposed pad. You don't often see something in QFP32 that requires an exposed pad, but when you do, they generally specify that you need to have it connected (otherwise they wouldn't have put it there, like all the other QFP parts).

QFN10A, QFN10B. (3 footprints for QFN10) plus a DFN-variant sold as USON-10. There are two shapes (2 pins on 2 sides and 3 on the others, and 1 pin on 2 sides and 4 on the others for a QFN10 and one of them is available in two different pin pitches. Both have been described as having slightly different physical dimensions depending on the manufacturer. The USON is yet another whackjob package that is shipping on production parts. It's like a DFN, 0.5mm pitch, with extra wide middle pin on each side. No, I cannot fathom why, either). Planned pricing $5/6

XSON/UQFN8 - XSON with exposed pad on one side, UQFN (3 pins on one 2 sides, 1 on the other 2 sides, 0.5mm pitch, 1.5x1.7mm). XSON8 is a much larger package - 0.8mm pin pitch - but it has an EP that you're supposed to solder down in the middle of it. I encountered it in the context of Winbond flash ICs, and it appears to be popular for such parts (particularly high capacities - it may be forced upon them by the die size. )

DFN6 - DFN6 with 0.4mm pitch one one side, and 0.65mm pitch on the other. Both are in common circulation. There was extra space on the board, so I put a SOT-23-6 footprint on as well. Since we sell them by the six-pack instead of the single board, you'll probably have some extras, and there's always something in some SOT23-6 or SOT-23-5 package that you need for something

DFN16D "miniQFN" 2mm x 3.2mm 0.4mm pitch. the shorter sides are completely filled by the pins, while on the longer pins, the pins start only once they clear the pins on the thinner ends. I happened to need some of these myself too; the only package in which this analog switch that was perfect for my needs wasn't backordered to absurdity in was this silly package. (several manufacturers use it with slight variation) with no EP, with two 0.4mm pitch DFN-8 packages on the back). It was quite possibly the most unpleasant package I've ever soldered!

SOIC16C - This has a TSSOP16 on the bottom, SOP/SOIC16 on the top. But it also has pads to solder 1206 decoupling caps real close to the chip. Obviously, not all parts have power and ground positioned on the same pins - but most either have it to the two corner pins (hint: you can rotate the part soas to use the right opposite-corner pins), or two pins one one end or the other. Those are the pins we had room to give a decoupling cap to.

More breakout board varieties will be made available in the future

There is definitely room for a more thorough treatment of the larger DFNs, but I'd like to get some feedback on these from customers and let that inform my decision of what to do next on this front.

Part mounting service

I'm considering offering the service of mounting parts onto these PCBs, as I've gotten reasonably good at it soldering these successfully, and I know many people are not comfortable soldering these sort of packages. Message me if this is something you would be interested in

Unavailable

Have we missed any? Please give suggestions for underserved packages!

There is only one real constraint here:

We will pull the datasheet for the part. If it is obvious that the part has a footprint that is deeply specialized will not make breakout boards that won't be fit for purpose, or won't provide a level of safety that I'd be comfortable distributing. Usually these whackjob parts come from companies that seem to be suffering from a bad case of not-invented-here syndrome when they decide which packages to use, and create novel unique packages that are different from everyone else's (and then, these days, often fail to provide reliable supply of the parts at hand) - but sometimes there is a very concrete reason that the part has such a strange package. This (far and away) most frequently seen in small PMIC parts that nonetheless are designed to deal with mind-blowingly large currents for the size of the package. TI, which is well known for using wacky packages, seemingly just for kicks, sometimes actually has a very good reason for it. Their ""Hotrod QFN10" for example, a tiny QFN16 that looks like the sides got squished closer together so the 4 corner pins merged, and with two of the pads extending the whole width. A - it's that way because of the high current it carries. It needs a dedicated board designed specifically the application. This is one of the times you'd probably want to use the company's Eval board.

I was recently asked about an EFuse chip from TI. The package it had was probably the weirdest package I had ever seen, with the exposed pad divided in half and a line shaped like interlocking fingers between them. It didn't take much consideration to realize that a) the part, if aligned even slightly off of the perfect alignment, the pads would be shorted to each other under the chip. since those pads were the two sides that the e-fuse was supposed to disconnect under fault conditions, that would be a Bad Thing, because it would not be detectable in non-fault conditions, and b) even TI wouldn't make a package that weird for no reason (and they love weird packages). So I start paging through the datasheet trying to spot some reason for such a strange package. It doesn't take long to find it: That would be because of the maximum current rating... THIRTY-FIVE AMPS, through something a few millimeters on a side! But something handling such extreme currents presents obvious hazards, which is why you would put a fuse or circuit breaker or efuse IC into such a circuit. But that wouldn't be able to do it's job if it was soldered slightly cockeyed or skewed, since the two sides it's supposed to disconnect would be shorted by the solder; such soldering errors could be invisible without x-ray imaging (which is sometimes done for critical systems industrially, and is common in electronics assembly process development). Beyond the fact that the part in question is a protection device which needs to work consistently, which won't be likely with any sort of home soldering process or readily detectable either, at that kind of current, even normally straightforward things, like connecting the wires, are no longer trivial. This is the sort of part that needs to be designed into the board, and may not be practical to solder at home at all (it would require some empirical testing to see whether a shorted one behaved differently in some measurable way without putting >35A through it. I will not make breakout boards for packages only used for parts for which a generic breakout board is not going to be viable (there must be parts that will draw, at most, a few amps (more than that, and you need to design a custom board anyway to accommodate the connector(s) you plan to use to handle such large loads (2-3A is most you can get through a 1p DuPont connector to pin header - and the dupont (knockoff DuPont) pins we normally see in hobby electronics are considerably worse), or sufficiently sized wires - 35A for example would require 13 AWG wire to follow typical design guidance figures), and it must not be a package used only by parts that are unusually layout sensitive and which would be unlikely to work without severe performance degradation (such RF parts, switching DC/DC converters, and similar).

Outside of that, pretty much anything is fair game, as long as it there is at most one termination (the EP) that is hidden when the chip is installed. No BGAs, no dual row QFN, or other things like that - things that can't plausibly be assembled with high yield without a stencil and reflow oven. I'm particularly interested to know which of the many rectangular QFNs that can be described mathematically (for N=Pincount/4, and a given pitch, there are N-2 possible shapes of the package - the 2 that are subtracted from the total simply being the extremes of square QFN and a DFN) actually exist in production hardware (you often cant spot them on parametric searches, because generic images of a QFN are used most of the time, rather than photos, the shape of the package is not always listed, and even when it is, the accuracy of digikey/mouser/etc's parametric search information is a bit of a crap shoot), and of those, which are actually packages that people end up wanting breakout boards for? I suspect there are some packages that are not all that exotic statistically - but which are only found on parts that aren't going to see use by hobbyists for one reason or another ("Yes just ask your sales rep and they'll get a copy of the required NDA to you, we can send you the full documentation as soon as we get that back" (you have to wonder what it is in the docs that they're so petrified about getting out that they'd take such measures. Those are a turnoff even to many business customers too - if they're small, they don't have inhouse lawyers to review it and will be uncomfortable about signing it's ), "... is only available as a bare die" (ie, those black epoxy blobs you see on cheap electronics instead of normal chips; that's a die sitting directly on the substrate, with the bondwires between the die and board, then covered with a splash of black epoxy for protection), "....10gbps....", "Radiation hardened" (aka "radhard" - for outer space and niche uses), "Only $163/each", and so on) - there are lots of reasons a part might not be a good match for hand assembly by individuals.