Nick wrote:
I donāt think you need to be using tantalums. Small electrolytics are fine.
Pete
Huh.
I just finished reading the diode chapter in on of my textbooks (the class is going way too slow for impatient Nick 
) and the atomic-level perspective makes a lot of sense about the non-sharing of the diodes.
Breaking down the barrier (which is where the dropout voltage comes from) is hard work.
Those lilā electron buggers are too lazy to break open a new path, especially when they can easily cram themselves in the already opened hole.
Nice to hear that I can get rid of these tantalums.
Methinks those were proprietary to a regulator which I have since tried and ditched.
None of the other datasheets that Iāve recently looked at seems to have em.
Time for me to google āpass transistor tutorialā.
Make sure you google ābig ā ā ā ā heatsinkā while yer at it
Iām not sure where it came from, but the idea of semiconductors not being able to work in parallel is a myth. To test it, put three LEDs on the same model in parallel and hook up a battery. What happens? They all light. Measuring current through each current path with a meter will show while the current is not 100% the same (they have resistances and current will be proportional, and all that), but it will be very close. If it was true that all the current would flow through the one with the least voltage drop, only one would light and depending upon your input voltage, burn out. Even still if you look on sites, youāll see people saying you canāt run LEDs in parallel or you need to put a limiting resistor in with each LED, etc. Do the experiment yourself and see you donāt need to. That said, you probably should measure the average working resistance of your LEDs, calculate the equivalent resistance of the matrix and choose a round value for a limiting resistor for the array to keep the current under the LEDsā rated amount if you were actually going to use this circuit for anything real.
Iām guessing the myth came from when diodes had massive differences in voltage drop even between two from the same model. But even still, this type of failure would only happen if the deviation was extreme or you were pushing the components significantly beyond their ratings. Using a good margin of safety (as you always should) and diodes of the same model, or measured voltage drop at working current, there is no reason you canāt increase capacity by running many in parallel.
As far as using multiple regulators in parallel, if you look on the link to Roman Blackās website I mentioned earlier (heās an EE guru way beyond my abilites), he mentions using multiple transistors in parallel in his designs to get both better efficiency (due to the lower voltage drop) and higher current. Enough examples of such advice exist from experts both online and in text format such as in Horowitz and Hill: The Art of Electronics (the EE bible), that I feel confident in my recomendation to do so.
If you have real life experience with such arrays going bad, please share, Iād love to get to the root of the issue (if there is one) and improve my designs accordingly. However, I have used such designs in many real world projects and not had any trouble.
As far as the tantalums, I agree theyāre not need them here. Sorry I missed that.
Sorry to double post, but one final argument. How do larger current rated diodes actually work? Well, usually assuming the material used in the diode is the same, they just make the junction bigger. Thus there is a wider area for the electrons to tunnel through. Assuming large enough EMF, after the first hole has formed, another will form elsewhere on the junction. In fact, I wonāt be surprised if mulitple pathways existed in enough low current diodes. However, from here it is trivial to realise that the junction has a variance in the voltage drop as itās not perfect and then to logically divide a single large junction into multiple smaller ones. I fail to see how this differs from an array of discrete diodes with close enough voltage drop curves.
The one thing that seems to have been forgotten here in your myth busters dissertation is that diodes have an exponential voltage to current relationship, and temperature plays a significant part in the exponential coefficient. As your currents move up the knee of the V-I curve very small amounts of voltage change can represent very large amounts of current change. Since power is a straight linear relationship between current and voltage the power dissipated across a diode also follows this exponential relationship. Temperature dissipation is an exponential function in the opposite direction however so as you drive your diode power up it cools faster and thus there is a relationship between the two, based on the net thermal junction coefficient of the diode junction to the surrounding infinite heatsink know as ambient, that will determine the stable operating point of your diode. Now, stack 2, 3, 4, however many diodes you want in parallel all next to each other on a board and you change that thermal coefficient to ambient. Better yet the diodes on the ends of the stack have a BETTER coefficient than the ones in the middle so the ones in the middle get hotter and wind up with more of the current load. This means over time they will deterriorate (sp?) quicker than the other diodes. In a home brew project where the āproductā life expectancy is as long as ones attention span the problems presented by stacking a bunch of power diodes in parallel rather than using a single device designed for the load are probably easily ignored. As an EE designing a production product that is expected to function many years in lots of different environments you need to fully understand the implications of the āmythā so you can make intelligent design choices and not create an albatross that doesnāt rear its ugly head for a year or two after the product has been released.
LEDs? Meh. At 10-15mA in big T1-3/4 packages who the heck cares about series resistors. Superbright LEDs operating at 100mA+ in high density SMT packages with copper leadsā¦ IMO you are an idiot if you donāt put individual series resistors, and these days you can even get current controlled LED output drivers that will make your product live much, much longer. Again, home brew = who cares while real world = do it the right way, get paid for the work, get a repeat customer on their next design project. 
Okay, youāve sold me on the idea that diodes in parallel might not be a good idea in every situation and that the belief that you shouldnāt run diodes in parallel isnāt purely a myth like I thought it was. You actually seem a lot more reasonable about it that most people Iāve had this conversation with. I actually was involved in a week long argument with a guy who insisted I was going to melt my PC if I didnāt use a limiting resistor on each of my LEDs because of diode failure. They were 30 mA superbright T1s. As far as the 100mA SMD LEDs, where would one find those? Not that Iām doubting they exist, but because they would be perfect for a project of mine. Also, I would think that such a device would actually already be an array of smaller junctions, though I could be wrong about that.
However, for the vast majority of cases I still think its okay to run diodes in parallel. The heating argument makes some sense, but only in the case where the diodes are not properly cooled, and are running relatively continuously at current levels which cause the diodes to heat up enough to start effecting their voltage drop properties. In the case of the 100mA LEDs, there is no real way to cool them properly, i.e. you canāt cover them with a heatsink and the current is starting to get fairly large, so its one of the cases where it matters. Also if youāre designing for industrial applications, or a wide enough range of temperatures, I can see this definitely coming into play. However, there are cooling benefits to using a larger array. I wasnāt talking about say two or three diodes, but a few dozen. That way even at larger current each one is not carrying that much. I generally try to keep each one at under 1/4 of its rated current since generally that minimizes the voltage drop of the array. Even in less optimal situations, what about just putting a heat spreader or better yet, if the diodes are producing that much heat, a real sink? Yes the temperature of the array is not going to be constant, and the middle ones will still die faster, but the time it would take to do this seems immense. I mean diodes generally last a really long time even under suboptimal conditions. A few degrees above ambient and say a sub 1 degree difference between diodes doesnāt seem like its going to have much effect.
It kind of seems like we are moving off topic a bit in this thread (sorry) so Iām going to beg out a bit and just answer the LED question since that might prove useful to some folks.
Lumex makes lots of superbrights, this thing lumex.com/images/pdf/SML-LX1110SOC-ATR.pdf for instance at 25,000mcd is scary running at 350mA and 2V continuous. 
Here is a similar Kingbright part us.kingbright.com/images/cat ā¦ 1SEC28.pdf rated 40,000mcd typical.
Something note quite so crazy, catalog.osram-os.com/media/_en/G ā¦ 6459_0.pdf are all in the 5000mcd or so range at 50mA.
Also, if you are into LEDs, remember the human eye sort of averages things so you can multiplex many LEDs at their peak current rating and only pull as much power as having one of them on at a time. Just remember to keep you refresh rate up above 75Hz or so. (and oh yeah I know this is going to raise the why more than 20 hz or so question from several people but just try it under 60hz background lighting and youāll see. heh.)
Eddie is 100% correct. It is not a āmythā that paralleling certain devices is a bad idea.
To re-state something he said that I think is a key point here: No professional would be caught dead designing with paralleled diodes, or paralleled bipolar transistors (in power circuits), or paralleled LEDs.
You CAN parallel power MOSFETs and vacuum tubes, because they work on different principles than the other devices mentioned.
To be complete in the discussion, I can think of one case where paralleling diodes would make sense: A special type of diode called a varactor is used in RF circuits because it has a junction capacitance that is related to the reverse voltage across the diode. Paralleling varactors will increase the effective capacitance, and is mostly fine, although itās still easier to just use a different (single) varactorā¦
Your description of putting many diodes in parallel can be made to work in some cases, but as Eddie said you canāt go to manufacturing with that type of design. Suppose you have just one diode in the lot that is ābadā, and is significantly out of spec? That diode will suck a lot of extra current, and thus heat up, and thus suck even more, and go poof. Thereās a good chance that it will short (the common failure mode for diodes), and then do damage to other parts of your circuit.
Why not just use the proper diode?
If you need lower V-drop, use a Schottky.
If you need essentially-zero V-drop, there is an IC that you connect to a MOSFET that behaves like an āideal diodeā. With just 2 tiny SMT chips you can make a āperfect diodeā that handles 6 or 8 amps. I can point you to the exact chips, if you want.
Regarding putting bipolar power transistors in parallel (or linear regulator chips): Just putting them directly in parallel is not done, because of the temperature-coefficient problem. One guy will heat up faster than the others, go into thermal runaway, etc.
The solution which has been used for nearly 50 years is you put a low-value resistor in the emitter circuit of each transistor. By ālow valueā I mean (typically) a fraction of an ohm for a multi-amp power supply. The resistors produce a āload balancingā effect as follows:
One transistor wants to suck more current than its neighbors.
But that current must go through the resistor.
More current through the R means more V across the R.
Now the V-drop across the transistor-resistor combo is higher, which leads to less current.
Meanwhile, the other parallel transistor-resistors are responding the same way.
Therefore, no single transistor is allowed to suck a lot more current than itās neighbors (unless it is way out of spec, in which case all hope is lostā¦).
Nick: If you really want to put 2 smaller regulator in parallel to get more current, one way to do it ācorrectlyā is to put a small-value resistor (say 0.1 ohms) in series with output of each regulator. Itās not really a good solution though, since the V-drop across the 0.1 ohms will vary with the current - when each regulator is doing 1.5 A, the net output will drop by 0.15 V. A pro would not do it, but it is āsafe and effectiveā, if you understand the limitations. Of course, you may not have a 0.1 ohm resistor lying aroundā¦
Another similar solution can be done that involves 2 regulators and 3 diodes. I can describe that with a picture if you like.
Pete
Thanks, I would really love to have such a design handy. It would totally reduce any need for a didode array. Could this circuit could be put in parallel for more current also? There is one instance where I would want to put around 30 Amps through such a device.
I agree that weāve effectively hijacked this thread. Maybe we should move to a new one on the General Electronics forum? Apologies moonbuggie.
Also, Iāve never had any diodes fail forward, when mine blow they generally become highly resistive. I guess this isnāt the normal case? Are diodes really that fragile? I was under the assumption that they were fairly hearty devices.
Ahā¦ The regulators I always use are switching (I assume they have a FET(s) in there somewhere) which would explain why I havenāt had any problems. The Black regulator and many other examples of semis in parallel were probably also FETs. However, Iād love to have the linear circuit you offered Nick, even if he doesnāt need it. Always nice to learn something. 
Iāve had experience with thermal run away in bipolar amplifier circuits, so I can easily imagine that it would be a problem here.
Thanks for the links.
I fully bow to you guys superior EE knowledge.
I wouldn;t say hijacked- lets call it borrowed 
I am an information sponge and as I asked the original question - all these responses add value and depth to understanding what I need to achieve.
now if you guys could post some schematicsā¦ 8)
All the more reason to keep those batteries together?
- My power specs -
2 batteries ea @ 7.2 V
qty part description
[1] Micro - 5V supply (250 mA draw)
[4] Lynxmotion Rover motors ( how many mA if all in motion/speed?)
[1] SSC-32 - 5V
[1[ Sabretooth motor controller @ 7.2 V
[5] servos from the Lynx 6 arm
[2] cheap servos for my sonars
[2] Max botix sonars 5 V
Near Future power needs
[1] additional Java based micro
[1] RF Data transmitter
[1] RF Video transmitter
What I need is a power header circuit that will allow me to tie in all these devices.
How would you guys approach this problem?
Neil
If you guys donāt mind horribly, Iāll make a final post to answer this:
Check out this image, and note the LTC4412 device in the top-middle (E5):
geocities.com/saipan59/robots/power.jpg
The LTC4412 is made by Linear Technology. They call it a ālow-loss controllerā. There are likely other gadgets that do the same thing.
The way it works is that E5 āsensesā any attempt for current to go ābackwardsā through the FET (Q7), and turns off Q7 to prevent it.
I would guess that it would work with a bigger MOSFET for more current, but youād have to study the datasheet. The tradeoff is probably that itās response-time becomes longer, but for the type of stuff we do it probably doesnāt matter.
The image shows the power system on my 'bot, although the stuff related to U2 has never been implemented.
Pete
now thats a hijack but there is a schematic!
Neil
Iām not sure whatcha mean about that header board.
The SSC-32 has itās own 5V supply, and if your micro is a BASIC clone, you can get the Mini ABB to supply and breakout it.
Both the SSC-32 have seperate VL and VS inputs.
VL input is intended for 9V batteries.
The battery connected to VL passes through the 5V regulator which suplies all of the ālogicā portion of the board.
The VS header is just a straight-through connection to the battery connected to VS.
This is intended to be a 6V or higher, large capacity battery to power motors and servos.
So, you should be fine, right?
Thanks for the idea diode circuit! I actually will get a lot of use out of that.
As far as the header thing goes, I think moonbuggie means a board to wire up to the packs and then run the appropriate voltages (regulated or not) to the appropriate boards. This may include some FET switches like in saipanās power supply circuit so he could turn off some devices to conserve power. I can imagine not needing servo control or servos if the arm is stowed properly and the rover is moving around, etc.
My only suggestion is that while the two packs should be more than enough for the motors, with all that equipment, you might need another one, or two. Iām not sure how you want to handle the power board. I donāt think perf. board is really going to cut it, due to the high currents that could involved. I guess it would be possible to drill bigger holes for things like the screw terminals, and use multiple wires to connect components carrying more current, but it would be a pain and not come out looking very nice. If you happen to have access to Eagle (not sure if the trial will cut it for what you want to do) you can design a small PCB and then you can get it fabbed with something like BatchPCB for relatively cheap. Just make sure to make your power traces nice and thick.
As far as regulation, Iām fairly sure that as long as youāre using switching (i.e. FET based) regulators you can put them in parallel. However, the only thing you really need to regulate is your micro board and maybe your motor controller board. The servos can handle a wider voltage range and so can the motors. So you probably wonāt even need to run switching regulators in parallel. This is assuming your charger doesnāt overvolt too much. Basically unless you want FET switches, your power board could just be a few rows of screw terminals, the regulator, and capacitors. Also, assuming you didnāt add more packs, you might also seriously want to consider my diode and cap circuit for the electronics.
I came across the SWADJ simple 3 pin variable switching regulator. It gets nice efficiency too, around 85-90% at 500-1000 mA. The spec sheet is here espritmodel.com/documents/de_parkbec/DE-SWADJ.pdf, this might be of interest to everyone. The downside is that it is very expensive.
I donāt know that I would jump into using this with an R/C receiver without some solid proof of other people using it successfully. It is basically a switch mode UBEC and the long road of battles in getting switch mode regulators NOT to interfere with R/C receivers is a well worn path. Notice in their list of applications they do not list r/c equipment. Iām not saying it wonāt or canāt work, what I am advising is caution in plunking down a wad of cash for this type of item unless there are documented examples of people (other than the people selling them) using them successfully in r/c applications.
edit: I have just confused this project topic with the guy making a 4wd w/pan+tilt and an arm being controlled via r/c where he is already dealing with learning about power routing, ground control, and separation. If you are looking at this for autonomous or hard wired control applications it is probably just fine to use. My concern was in regard to using it in the r/c application I just mentioned. sorry for any confusion I may have introduced.
No problem, your post actually got me thinking I might need to be more careful about using switching regulators with radio controlled projects. I havenāt encountered any problems (I mainly use bluetooth and WiFi as my radio links) but I also am not at the stage where I need any range as all the pieces are either on my workbench or across the room. Any tips on how to plan things so they donāt interfere? I can certainly see how it would be nice if I wasnāt working against myself when I finally did want to give the bot a good long range WiFi link.
actually bluetooth and wlan are packet based with error detection and correction capabilities. noise just reduces their throughput (sp?). typical r/c radios are analog and you only get 50 updates a second so scrambling even 1 frame can cause a perceptible glitch. most wlan access points, certainly your computer and laptop with a wlan or bluetooth card, and non-battery powered bluetooth peripherals are using switchmode power supplies. What you need to do is keep the power clean. It is not uncommon to isolate rf or analog sections power from the main digital power with small r-c or sometimes pi filters. filtering and common mode chokes at the point the power enters the board also help keep the switching crap out of the board to begin with. it is very common in multi-layer pcb designs to carve up and control how the power and ground planes allow current to flow between portions of a design. in 2-layer design you are always trying to balance ladder power distribution against star distribution layout to control how noise moves between circuits. very generally speaking just consider every conductor both an emitter and an antenna, remember that signal strength falls of as the square of the distance, and itās always easier to keep noise out of a system than to remove or deal with it once it is in there.
Thereās the LM2574 .5A simple switcher family that I know of.
The 12V one has been used in many of the OSMC designs, so Iām pretty sure that itās safe to use when dealing with large loads as well as with or close to radios.
Then again, Iām basing that simply upon them repeatedly using it.
Maybe thereās been interference and no oneās noticed.