All posts by Pete Soper

Plain Diodes are terrible voltage regulators

Sometimes we’re faced with a situation where a single DC supply is all that’s available but it’s is too much for part of a system and we really wish we had a lower voltage too. I knew diode forward voltage drop is a function of current (and temperature), but I didn’t appreciate just how variable this was until I got lazy and failed to look at a datasheet before I strung some 1N4148 diodes in series as a quick and dirty way to knock 12 volts down to nine. I figured that with the “standard” voltage drop across a silicon junction being .6 volts then 6x.6 = 3.6, and I knew if the current variation gave me around nine volts that would be OK for my application. Except when I tested this I found the variations at the currents the application involved would be all over the map. Here’s a table of what was measured across the six diodes at different values up to the absolute max current for the device:

Forward voltage drop across six 4148 diodes vs current

Here is the same data but for a single diode:

So the per-diode voltage drop is only close to the .6 volts cited for silicon junctions at around one milliampere of current. At the upper current limit of the diodes the drop is very much higher. My application spanned a wide range of currents, so this diode string was hopeless (the client might have noticed a slight behavior change depending on operating mode, and I couldn’t tolerate that.) In my defense over the years I’ve treated diode voltage drops as something I wanted to minimize, making me not just a schottky diode bigot but one who would spend a hour finding the absolute lowest forward voltage drop for a given current. But I obviously developed no intuition at all for the more general cases.

Luckily I found a 7809 linear regulator in my junk box and I can simply carry on without having to order a part, but I thought this might be interesting to share. A high enough power zener would have been another solution.

As mentioned at the beginning, diodes change their behavior with temperature changes too and this is why you sometimes see a diode clamped under or near a power transistor: it’s change in behavior is leveraged with circuitry to keep the transistor from operating outside its safe temperature range. And some folks are able to use tables or perhaps a Taylor series with firmware to monitor voltage drops and use their simple diodes as thermometers.

Keeping YOUR POWDER^H^H^H^H^H^H ICs DRY

Storage for parts that must not absorb moisture
Dry air storage

This is about assembling five PCBs yesterday that shared an IC and how, despite meticulous application of solder paste, reflow profile, yada yada, two of five didn’t pass FAT (final assembly test via a simple computer program and microprocessor). Late in the day a little light bulb glowed with faint neon letters spelling “popcorn.” This is the slang for when trapped moisture in an IC explosively changes to steam in the reflow oven, physically damaging the part. The part in question had been away from desiccants for months, the humidity in the shop is far from Arizona levels, etc. I don’ t have the equipment to determine if this damage happened, the budget involved can’t cover sending the board off to be studied, etc, so I just have to assume this is a “maybe” cause of my failures.

So I followed the specs for “baking out” the moisture and set up these chips with some others that need to stay dry or be in line for baking out and thought I’d share a snapshot. (Click on it to get a better version, and click on that for full resolution. The inner card color dots are distorted: they are actually blue). You know how we get desiccant bags and moisture card from places like DigiKey and toss them? I saved enough over the years to combine them with a cat treat container and Indian sauce jar to make a two level dry air environment. All the bags and some of the chips went through the bake out process (I used 125C for 12 hours). The outer card will let me judge how well the cat treat container lid fits and the same for the card inside the sauce jar with its tighter lid. Access is pretty easy and there is room for a “next gen” larger inner container down the road. The parts are in plastic boxes with labels facing out so I can quickly determine what’s there, and a few notations are on the box stickers to indicate their status wrt baking. (My parts database uses US town names starting with “A” for containers, thus the “AMORY” sticker).

I’m assembling several more of the same PCBs today and it would be nice if Whammy Central allows for pop-corning to have been the cause and the new boards to all pass FAT!

Where some CPU CYCLES COULD BE USEFUL

Years ago at one point I had maybe 200 cores of Sun Microsystems servers searching for gravity waves several days at a time through out the year as part of the “LIGO at home” project. Our Java release tests ran in waves and I could borrow the machines to use the troughs to crunch on this project.

Now there’s “Folding @ Home” that is for doing the very hard work of figuring out protein details or other life science computing, and apparently there is some work going on relevant to the virus pandemic. The link below is to an article about this and it contains a link for taking part in the last sentence. If this works the way the LIGO one did, the app automagically gets out of the way of you using your PC, such as running when the screen saver is active. I suspect it’s even more flexible now. So consider donating a bit of energy with your PCs to pitch in and make this go faster.

https://decrypt.co/23064/folding-at-home-now-has-400000-users-fighting-coronavirus

Direct link to Folding @ Home: https://foldingathome.org/

Key CAD Developer Interviews

Image result for tomato

(OK, I say tomatoe; my wife says tomahto.  But out of respect, we don’t say we use Lienix, or Pythone, right? Although it took from  1991 to the late 90s, we did eventually stop saying Lienix, Leenix and Linooks.)

Chris Gammell, the expert interviewer half of The Amp Hour and long time KiCad supporter, organized a fantastic first KiCon conference in Chicago. Chris conducted some execellent interviews, including one of some of the KiCad developers, including project lead Wayne Stambaugh.  Wayne is now employed full time to develop KiCad, a new milestone for this tool that seems to be on track to become the gcc of PCB CAD in coming years.

 

Shop Talk

1.54

I’ve been listening to The Amp Hour podcast for some time and want to point out one episode that was particularly pleasant to listen to.  This is episode #412 titled “3 Cent Micros and 1000s of LEDs”. It’s Dave Jones of EEVBlog and Mike Harrison of Mike’s Electric Stuff talking shop, sharing interesting news items, and just yacking about a very wide variety of topics. I found it just the thing while standing in line to vote, rummaging through my old ham radio gear hoping my LMR400 cable was still there (wasn’t) and struggling with the failing Digikey web site (first time every: could not do a parametric search for a SMD cap without getting a weird “page not found” web server error).  Dave is free and easy this time around and the only problem folks will have is that when a Brit and an Aussie talk rapidly American ears can struggle to recognize all the words in real time, so have your “replay the last 10 seconds” button ready. Still, well worth it, and good on you Dave, for an excellent interview!

Arrows for your Quiver

Image result for quiver

Two events reminded me recently that you can’t have too many choices for solving a sourcing problem.

A client and I had finished a simple PCB design, got it off to OSH Park and OSH Stencils, and turned attention to the BOM. All passives but one: in my shop already, so check. The remaining one and some connectors: Digikey has them, so check mark. The central IC that is the point of the board:  Digikey, zero. Mouser, zero. Other suppliers we’d heard of, zero. Anyplace in the USA, zero. China, only on breakouts. Ouch.  Europe: “Rutronik”. Who? After a day of thrashing I noticed this is the chip manufacturer’s favorite distributor in Europe, and I should be able to trust them. I had to establish a business account but all was well until we got to shipping: Big ouch. And this huge minimum shipping wasn’t for  Startrek teleportation beaming from their warehouse to my bench, as the fee implied. They predicted 72 hours, but later we found that was if you’re in the EU. Oh, right. When it wasn’t at my door in 72 hours I dug up an Excel spreadsheet  via a well disguised link on my account page and on one row was a red “warning dot”. More digging and I decoded this warning to translate to English as ” six days estimated delivery from order date”. But the chips should be here in time for the reflow oven and all is well even if the client’s wallet is thinner than desired.

Just days later another client and I were deciding what parts of a new board really needed breadboarding to give us confidence in a major respin and this forced an immediate order of a few piddly parts (note to self: as soon as the part is on the short list for a design, order three of them.) I was inhaling to commiserate about shipping when the client reminded me that Arrow isn’t charging anything for shipping. Period. Zero, no matter what size the order is. I  was startled, as this has been going on for months and I just assumed it was a one shot, short time thing and had let it fade from my memory. Instead I was able to throw the part number and quantity  for a special FET I want to try out into the client’s order, smiling at the idea and my share of the shipping will be zero. Nice.

Meanwhile, the US rep for the big Euro distributor contacted me and we’ve agreed I will call her before risking another big wad on shipping that isn’t even fast in relation to the cost.

So, one arrow moved to the front of the quiver and a new one added for Euro parts that the US hasn’t discovered or that got too popular for supply here to keep up with.

All git repo hosting sites are not created equal

GitHub, GitLab, and Atlassian BitBucket are all sites that offer git repository hosting. But if you create a repo named “Test” with BitBucket, then you copy/paste the slug URL  and feed it to “git clone” as usual,  the URL is forced to all lower case  “test” and you end up with a clone of your repo in directory(folder) “test”. This stinks, for example, with common library directory naming conventions.

After thrashing around in the team settings, thinking I’d just missed the way to override the default URL, it occurred to me this might be a feature, not a bug. Sure enough:

https://jira.atlassian.com/browse/BSERV-3148

(“Stash” is simply Atlassian’s downloadable repo hosting system, effectively giving a customer a “Bitbucket” system within their enterprise)

This is just a classic bug report pattern. Who knows what caused the Bitbucket architect to decide not to honor the name spelling when creating the slug (the text box holding the URL you copy from to do a clone). But, the fact that one can simply modify the URL and end up with a target directory with any pattern of upper/lowercase makes it clear that they COULD offer an “honor name capitalization” radio button to override their sacred cow default, but they choose not to, with the lamest of lame excuses:

“Given that it is possible to modify the clone URL to include a camel-cased repository name (which creates a camel-cased repository directory when cloning), this is unlikely to be a priority in the near future.”

But this latest submission was started in 2013 and “resolved” in 2015, two years ago. Maybe it’s time to submit it again.

 

ESP-32 Developer Interview

A recent edition of “The Amp Hour” podcast had Chris and Dave interviewing “Sprite_tm” (aka Jeroen Domburg) of Espressif. Amongst all the other things this interesting Dutchman has accomplished, he added symmetric multiprocessing support to FreeRTOS on the ESP-32 (which has two identical 32 bit processors). I came completely unstitched hearing this toward the end of tantalizing news about the software tool chain available for this ESP8266 follow-on. Transparent multithreading with complete control over processor affinity (i.e. either bound or freely scheduled) was the final brush stroke painting a picture of superior infrastructure, open source accessibility and the usual unbeatable price point of the Espressif hardware. Wow.