Hot Links: Freedom Edition

This Inde­pen­dence Day, I’m going to start a new series called Hot Links, in which I go full neck­beard (read: offer my unin­vit­ed and under­in­formed opin­ions) on stuff I find on the Internet.

Today’s theme is all about that good Amer­i­can stuff.

Texas Instruments (TI) wireless modules

As a hob­by­ist who will soon start a career in cre­at­ing hard­ware, I see TI as folks who “get it.” They’re already at the top of most engi­neers’ lists just by hav­ing great prod­ucts with (more impor­tant­ly) excel­lent doc­u­men­ta­tion. But it’s more than that. They real­ly go the extra mile to make the design­er’s life eas­i­er; they’ll cut down on exter­nal parts for their inte­grat­ed cir­cuits (ICs) when­ev­er pos­si­ble and try to use few­er pins to get the job done. They even try to make the pinouts on their chips easy to lay out a print­ed cir­cuit board (PCB) for.

So I was­n’t sur­prised to find a TI Blue­tooth mod­ule cater­ing to my “lazy bum” use case of lever­ag­ing the Blue­tooth Ser­i­al Port Pro­file (BT SPP) to emu­late a con­nec­tion between a micro­con­troller UART¹ and a desk­top ser­i­al con­nec­tion: the LMX9838. My lack of sur­prise lacked hard­er when I dis­cov­ered that it’s small­er (10 mm × 17 mm × 2 mm) and not cost­lier than what I had been using due to its sin­gle-mind­ed­ness of pur­pose in doing only that. It’s got that per­fect “no-frills” prod­uct feel. I mean just look at the appli­ca­tion diagram:

This part is a bru­tal­ly sim­ple “by-engi­neers-for-engi­neers” prod­uct that also hap­pens to be per­fect­ly con­veyed by their tech­ni­cal writers/illustrators. Com­pare that schemat­ic to the cur­rent hob­by­ist-favored² Rov­ing Net­works RN42:

Here’s the thing, guys: the folks who look at some­thing like that and think, “hmm, I’d like to incor­po­rate that stuff into my prod­uct” are not the fuck­ing cus­tomers for mod­ules like this because they’re design­ing a prod­uct. It’s noobs like me who use an LMX9838 or an RN42. Unless your “prod­uct” is “hand­made” by “pro­duc­tors” in “Brook­lyn³,” you’re not using some off-the-shelf plug’n’­play mod­ule like this in it. It’s not just a Blue­tooth radio; it’s a Blue­tooth radio on its own board with an anten­na and a micro­con­troller run­ning a whole BT stack and it con­nects to com­put­ers all on its own. Engi­neer­ing teams⁴ design­ing prod­ucts are per­fect­ly capa­ble lay­ing a sim­ple BT radio down, rout­ing RF fron­tends to anten­nae, drop­ping some ven­dor’s yucky Blue­tooth stack on their own micro­con­troller, and then get­ting it FCC cer­ti­fied. Every­one else just wants a don­gle that hooks her “data device with sen­sor” to her MacBook.

Now, tin­ker­ers and pro­fes­sion­al engi­neers would prob­a­bly rec­og­nize that the LMX9838 part num­ber does­n’t look very TI-ish, and they’d be right: this is orig­i­nal­ly a Nation­al Semi­con­duc­tor part. How­ev­er, it got dis­con­tin­ued in 2010 and TI wise­ly brought it back, prob­a­bly as part of their Sim­pleLink move. How­ev­er, the only “every­thing onboard” Blue­tooth solu­tion that’s actu­al­ly part of the Sim­pleLink fam­i­ly, the CC2564-BLUEBRIDGE, is from two (?) 3rd-par­ty devel­op­ers (???) and suf­fers from pret­ty much the same immense fea­ture-creepi­ness I just whined about. It’s also big­ger and more expensive.

Texas Instruments Half-Bridge Modules

Along anoth­er branch of TI’s awe­some “shove many great­ness­es into one mod­ule” path of prod­uct evo­lu­tion, I’m real­ly dig­ging how they packed a half-bridge dri­ver and a half-bridge into a 5 mm × 6 mm pack­age (think SON‑8 except not 8) I nor­mal­ly asso­ciate with pow­er MOSFETs.

Every­one had already been cram­ming two N‑channel Field-Effect Tran­sis­tors (N‑FETs) into one 5×6 block for a few years now, but now we can have it with dead­time-matched dri­vers built in? This is hard­core Amer­i­can patri­o­tism right here.

Sad­ly for motor con­trol fans⁵, this is designed for mul­ti­phase buck DC-DC con­vert­ers in com­put­ers, where a bank of these help drop 12V down what­ev­er your CPU uses. That means that the FETs inside are high­ly asym­met­ri­cal, as the high-side FET is pass­ing cur­rent much less of the time than the low-side FET does. For exam­ple, the r_{DS (on)} of the high-side FET on this dual FET (no dri­ver) is more than 3 times that of its low-side FET. So this is no bueno for con­trol­ling motors on a robot, where all the FETs are engaged for basi­cal­ly the same frac­tion of the time. Not only that, these mod­ules are rat­ed to only 16V input—not enough for many.

I can only hope that TI can bring this minia­tur­iza­tion and ultra-low r_{DS (on)} tech to their motor con­trol pow­er stages, which are cur­rent­ly utter shit.

Intersil Full-Bridge Drivers

Mov­ing on to anoth­er Amer­i­can semi­con­duc­tor com­pa­ny, Inter­sil in good ol’ Sil­i­con Val­ley. They’ve got three dif­fer­ent dual half-bridge dri­vers in itty-bit­ty 16-pad 4 mm × 4 mm QFN. If judg­ing only by pad count, they’re even bet­ter at the “sim­ple is bet­ter” thing than TI is: just think, you need at least

• Four pads for each of two half bridges
• Two pads for each of log­ic and pow­er supplies
• Two pads for the inputs

Oh wait, that’s it. The oth­er two pads on the ISL6610 and ISL6614B are both No Con­nects (NCs). Heck, the ISL6610 datasheet even says that it’s two 6609 sin­gle half-bridge dri­vers crammed togeth­er with enable func­tion removed. The PWM inputs are three-state (high/low/disable for 1/0/Hi‑Z), the boot­strap diodes are built-in, the shoot-through pro­tec­tion is auto­mat­ic (?!), and the only exter­nal com­po­nents are two boot­strap capac­i­tors for the high-side dri­vers (whose val­ues, of course, depend on what FETs you drive).

It’s so Amer­i­can, it’s as if Apple designed and sold inte­grat­ed cir­cuits and this were one of their three semi­con­duc­tor products.

At any rate, either of these would be real­ly nifty dri­vers for brushed DC motor con­trol in robots. I’d use the ISL6614B for the big­ger FETs that require more than 5V to ful­ly enhance (it does require a sep­a­rate sup­ply though) and the ISL6610 when 5V gate dri­ve would suf­fice. Now the caveat here is their absolute max­i­mum volt­age rat­ing (no rec­om­mend­ed is giv­en) of 15V, with 30V for 100 ns spikes. I think that rules out both chips for a lot of robot folks. Those folks should then look at the ISL6210, which has dou­ble the volt­age rat­ing and some fea­tures you might actu­al­ly want (I don’t know what diode emu­la­tion is, but appar­ent­ly that pad dou­bles as a dead­time set resistor).

The final catch? I can’t find any vari­ety of the ISL6210 in stock any­where, and the oth­er two only as 14-SOICs, which are mas­sive. Rawr.

Maxim Integrated Charge Pumps

Speak­ing of using high­er gate dri­ve volt­ages, where were you about to get that gate dri­ve volt­age? For smart but small ana­log solu­tions where the con­di­tions aren’t extreme, I look to anoth­er Val­ley com­pa­ny, Max­im. I’ve known them as the com­pa­ny that sounds like a machine gun and the men’s mag­a­zine, and which has the best sam­ples pol­i­cy for hob­by­ist engi­neers, hands down. I’ve been grub­bing free chips off of them in lit­tle acrylic cas­es lined with anti-sta­t­ic foam bed­ding since mid­dle school (I can hear the geezer engi­neers scoff­ing already :P).

Side­note: they recent­ly rebrand­ed à la Microsoft (nice but­ter­fly logo) and retooled their web­site in such a way that changed zero func­tion­al­i­ty. Wat.

For gen­er­at­ing gate dri­ve volt­ages, how about a MAX1680/1 or even a small­er MAX680/1? Both are charge pumps. The for­mer han­dle 125mA and the lat­ter 50mA, and both use just three exter­nal capac­i­tors to dou­ble volt­age input volt­age for com­pact, sim­ple DC-DC con­ver­sion. Nifty.

Sneak Peak

Hmm, I def­i­nite­ly did­n’t intend to do the whole thing on just elec­tron­ics, but oh well. Hap­pens. Here’s what I’m work­ing on right now:

1. A hard­ware periph­er­al or sub­sys­tem on most micro­con­trollers that han­dles asyn­chro­nous and syn­chro­nous ser­i­al com­mu­ni­ca­tion. Oth­er periph­er­als can talk on the SPI or I²C bus­es, gen­er­ate PWM, etc. [▲]
2. And by that I mean sold car­ried by Spark­Fun, one of the self-appoint­ed tastemak­ers of the “mak­er” com­mu­ni­ty. [▲]
3. Since when was Williams­burg ever rep­re­sen­ta­tive of Brook­lyn? Since when was it even part of Brook­lyn? [▲]
4. So again, we’re drop­ping Williams­burg from the race here. [▲]
5. As in peo­ple who like motor con­trol, such as a Fan, not fans designed for… con­trol­ling motors…? [▲]