Why is there such a large difference in the current capacity of these vregs?

Btw: 400 mA drop-out compared to 250 mA of drop-out means 60% higher power dissipation at a minimum.

That's what I would have thought myself, but none of the power dissipation equations I have seen in the datasheets or online appear to include drop out in the power dissipation calculations. And since the Iq for both parts is the same, the amount of power that needs to be dissipated for both is equal.

I think I found the difference between the two regulators though.

It's not that the two packages are mounted differently. The landing pad sizes are virtually identical.

The difference can be seen on page 5 note 3 in the first datasheet, and in the Absolute Maximum ratings on page 4 of the second datasheet. The junction-to-ambient thermal resistance: ?j-a

For the first part, ?j-a = 220°C/W
For the second, ?j-a = 206°C/W

And the equation for calculating the amount of power that can be dissipated at a specific ambient temperature is:

p(max) = ( Tj(max) - Ta ) / ?j-a

So, the power which can be dissipated is the maximum junction temperature, minus the ambient temperature, divided by the junction-to-ambient thermal resistance.

This power calculation isn't new to me, but yesterday when I was tired I could not for the life of me find those thermal impedance ratings and I saw something in one of the datasheets indicating what a typical value would be on FR4 pcb, and I thought that it must be the same for both parts because the two packages are the same size and neither has any kind of heat sinking.

This leaves me with a little problem though. A lower thermal impedance is better. And doing the calculations for am ambient temperature of 25 degrees, I get:

(125-25) / 220 = .455
(150 - 25) / 206 = .607

Which means the first part, the one which is 3x as expensive and has a higher current rating, can actually dissipate less power than the second part.

But that's for the same package. The first part comes in a second type of package with a heat sink on the bottom. And that package has a thermal resistance of 282 C/W on a 2 layer board with no heat sink (worse than the first two) and 62 C/W on a 4-layer board with 6 thermal vias.

Plugging that into my equation I get:
(125-25) / 62 = 1.61

Which is 2.65x better than the 50 cent regulator. So by all rights it ought to be rated for 400mA if the cheaper one is 150mA, but I don't know what criteria they used to decide to rate it... ie what amient temperature, what current, what input and output voltages. So that could explain the difference.

Anyway after all this, assuming I don't go with the third option, the part with the heat sink on the bottom:

It would seem that the 50 cent part would be the better of the two choices.

Now I just have to figure out how much current it can really carry by doing the rest of those power calculatons.

Just did some calculations for the 50 cent 3.3v regulator:

Assuming a max PD of .607:

A = .607 / (Vin - Vout)

Vout = 3.3v, so:

.607 / (5 - 3.3) = .357 (Piggybacking off my 5A 5v regulator)
.607 / (8.4 - 3.3) = .119 (7.2v NiMh -> 8.4v at full charge)
.606 / (12 - 3.3) = .070 (12v battery - probably more like 13v+ but whatever)
.606 / (16 - 3.3) = .048 (16v worst case)

So based on this information, I think I should be okay using the same 50 cent 3.3v regulator I used for my SD card. Using a 7.2v rechargable or 6v alkaline will be what I reccomend, and if someone absolutely must run at 12v, they can plug the expansion board with the regulator into one of my servo ports which supply a regulated 5v instead of Vcc, which will put some more stress on the 5v 5A regulator but take a lot off the 3v.

Decided to do the calcs for my 5v reg as well:
http://www.micrel.com/_PDF/mic29150.pdf

Junction temp max - Tj(max) = 125°C
Ambient temp - Ta = 25°C
Thermal resistance - ?j-a = 2°C/W

Pd(max) = (Tj(max) - Ta) / ?j-a
Pd(max) = (125°C - Ta) / 2°C/W
Pd(max) = 25W @ 75°C
Pd(max) = 50W @ 25°C

I'm not sure which ambient temp to use, the regulator docs mention 75°C but 167°F seems excessive to me. Still, I'll go with 75 for my calculations.

This regulator gives a slightly different calculation for power dissipation which takes into account the ground current:
Pd = Iout (1.01Vin - Vout)

So if Vout = 5v, and Pd = 25W per my above calculation, and I rearrange the equation like so:

Iout = Pd / (1.01Vin - Vout)

Then, @ 75°C ambient and:
6v in, Iout(max) = 23.5A
8.4v in, Iout = 7.2A
12V in, Iout = 3.5A
16V in, Iout = 2.24A

A = .607 / (Vin - Vout)

Vout = 3.3v, so:

.607 / (5 - 3.3) = .357 (Piggybacking off my 5A 5v regulator)
.607 / (8.4 - 3.3) = .119 (7.2v NiMh -> 8.4v at full charge)
.606 / (12 - 3.3) = .070 (12v battery - probably more like 13v+ but whatever)
.606 / (16 - 3.3) = .048 (16v worst case)

Let me toss in a few ideas here. These last calculations are the most pertinent. They show
you are quite limited in how much current the v.regs can "realistically" supply, without melting
your pcb. Assuming the NiMH cells are the most likely to be used, you've got ~120 mA max.

HOWEVER, and it's a BIG however, you're allowing 150degC temp on the v.reg. How hot is that?
100degC is the temp of boiling water, and how long could you keep your hand in boiling water
for? 0.5 sec, that's it. IOW, you're really pushing it here.

PD = IOUT * (VIN - VOUT) + VIN*IQ
where VIN = input voltage; VOUT = output voltage; IOUT = output current, A; and IQ = quiescent current, A

I assume quiescent current is also known as ground current as that seems to match up with the minumum quiescent current listed in the absolute maximum ratings and there's no other quiescent current listed. In which case for both vregs, it's 850uA typical at 150mA output. Which means as far as the power dissipation equation goes, the two regulators are identical.

Secondly, you're initial ideas here were offbase. The PD eqn is simply how you calculate PD from
Ohm's Law and circuit theory, and has nothing to do with the characteristics of the v.regs. The 1st
term is the important one, and has nothing to do with what's "inside" the box, only with the
input-output characteristics. The 2nd term is basically insignificant, since IQ is > 150X smaller than
IOUT.

Thirdly, for my part [and I've designed a lot of pcbs], I wouldn't even think about using such a tiny
smt part at all, unless I was sure PD would always be maybe < 50% of your 0.6W value. Then the v.reg
would not overheat so much, and it would still be a 75degC = 167degF on the part. Still pretty
darn hot.

Fourthly, I don't even like the SOT223 parts [160degC/W], let alone the tiny ones you're using. I
go for the NCP1117 DPAKs [67 degC/W].

Fifthly, I think the answer to your initial question ["difference in current capacity"] mainly has to
do with the physical characteristics of the PNP output transistor in the parts, and not with the
size of the package, etc. Eg, the smaller the collector area of the transistor, the lower the
max current.

Thanks for the info Dan. :slight_smile:

Let me toss in a few ideas here. These last calculations are the most pertinent. They show
you are quite limited in how much current the v.regs can "realistically" supply, without melting
your pcb. Assuming the NiMH cells are the most likely to be used, you've got ~120 mA max.

HOWEVER, and it's a BIG however, you're allowing 150degC temp on the v.reg. How hot is that?
100degC is the temp of boiling water, and how long could you keep your hand in boiling water
for? 0.5 sec, that's it. IOW, you're really pushing it here.

Well, I was calculating absolute maximums, not necessarily the average current the regs would be putting out.

In my particular setup for example, I would like to power a 3v vibration motor or a few lasers. A massive 10G vibration motor or a green laser might need 200mA. A smaller vibration motor or three red lasers might need 150mA. And a small vibration motor or a single laser might need onl 60mA. But in any case, they would only be on for brief periods of time. A few seconds at most, usually.

Thirdly, for my part [and I've designed a lot of pcbs], I wouldn't even think about using such a tiny
smt part at all, unless I was sure PD would always be maybe < 50% of your 0.6W value. Then the v.reg
would not overheat so much, and it would still be a 75degC = 167degF on the part. Still pretty
darn hot.

Well, my plan as of this moment is to forget trying to run the 3.3v off the battery because even a NiMh doesn't leave me with much amperage to work with before the 3.3v reg would overheat. So the plan now is to run both 3.3v regs on my board off the 5v regulator. That will allow them to put put almost 350mA, which is 150mA more than what I ever expect the reg will need to put out, and even if does get to 200mA, it will only be for brief periods as I mentioned earlier.

Oh, and in case you're wondering why I have two 3.3v regs instead of just one bigger one, it has to do with the fact that the board is modular. The 3v3 reg that would drive the motor/lasers would be on its own board, and the other one drives an SD card. It would be nice if I could just use one big 3v3 reg on the main board, but I don't really have the space for it, and it would make the way everything connects to the main board a lot more complicated for people. Trying to keep things simple. The board is already studded with pinouts:

Fourthly, I don't even like the SOT223 parts [160degC/W], let alone the tiny ones you're using. I
go for the NCP1117 DPAKs [67 degC/W].

That looks like a nice part, but the problem is, that regulator doesn't have an enable pin, and the plan is to switch the motor and laser using the enable pin on the regulator. Keeps things simple. I've calculated the regulator should be able to switch at the PWM frequency of the Arduino (around 500hz) but if not, it's no big deal, I can do software PWM more slowly or not at all. It's just a vibration motor. Can't do PWM on the laser anyway; the driver boards usually don't like being switched faster than 10-100hz from what I've read.

Now usually, someone would use a mosfet for switching, but like I said, I don't really need to switch fast, I think the reg can handle PWM speeds, it saves me a component, and I don't have to worry about finding a mosfet which can handle the same current the regulator can. I had that problem with the first 5A reg I chose, because it didn't have an enable pin, and I need to be able to turn the power off, and I didn't want to have to use a 5A switch, but a mosfet that could handle 5A was the same size as the regulator. In the end I decided to just choose a larger regulator with an enable pin, and I probably got more current capacity as a result to boot.

I'd be interested in your thoughts on this plan though.

On another note, I'm looking for a suitable flyback diode for the aforementioned vibration motors.

This motor here is probably the worst case scenario:
https://catalog.precisionmicrodrives.com/order-parts/product/320-100-20mm-vibration-motor-25mm-type

So that would be run at 3.3v around 170mA, but I'd like to assume 200mA just to be on the safe side.

I found this diode on Digikey:

And I found this thread on flyback diodes:

Retrolefty there states the diode should be 4x the voltage rating of the coil, and 40v is well over 12v. And it's got a 5.5A surge current rating which is over 25x as much as I think I need.

the cheaper one's maybe just will make "pop" one day if you work them too much towards the margin.

It happened to me, using very tiny SMD diodes for a dc/dc converter.
Fine at 1.2 volts, at 5 volts it only survived for some minutes.

Then I changed to a big, way overdimensionated diode (2A rating, max. current maybe 100 mA).

I would not bother too much about it.

Well, I was calculating absolute maximums, not necessarily the average current the regs would be putting out.

In my particular setup for example, I would like to power a 3v vibration motor or a few lasers. A massive 10G vibration motor or a green laser might need 200mA. A smaller vibration motor or three red lasers might need 150mA. And a small vibration motor or a single laser might need onl 60mA. But in any case, they would only be on for brief periods of time. A few seconds at most, usually.

One thing about those teensy-weensy v.regs is they might work ok in a dedicated app, like a cellphone/whatever where
you know what the load will always be, but when you have a more general-purpose board, you never know what might
be connected to it.

Just think, 150degC = ~300degF. That's almost the temperature most things are cooked at in the oven. If nothing else,
you run the possible risk of someone burning the heck out of their fingers. To me, even the 175degF I mentioned is
more than I want to reckon with.

The switch on/off bit is a problem, but I still might look for a more robust part with enable.

Well, my plan as of this moment is to forget trying to run the 3.3v off the battery because even a NiMh doesn't leave me with much amperage to work with before the 3.3v reg would overheat. So the plan now is to run both 3.3v regs on my board off the 5v regulator. That will allow them to put put almost 350mA, which is 150mA more than what I ever expect the reg will need to put out, and even if does get to 200mA, it will only be for brief periods as I mentioned earlier.

Even running off 5V is a problem. (5V-3.3V) * 0.2A = 0.34W * 206degC/W = 75degC temp rise. That still gives 212degF
on the part. For me, that's too hot, boiling water hot, but it's your board.

... but I don't really have the space for it, and it would make the way everything connects to the main board a lot more complicated for people. Trying to keep things simple. The board is already studded with pinouts:

http://shawnswift.com/arduino/layout2.png

Space is always a problem on pcbs. I can never get them as small as the original goal. Having burned up boards is
worse. In the end, you just have to make your own choice of tradeoffs.

Fourthly, I don't even like the SOT223 parts [160degC/W], let alone the tiny ones you're using. I
go for the NCP1117 DPAKs [67 degC/W].

That looks like a nice part, but the problem is, that regulator doesn't have an enable pin, and the plan is to switch the motor and laser using the enable pin on the regulator.

Ok re on/off. I would still look around for another part, but that's just me/LOL. I'm sure they're there.

A lot of Arduino bds use the little SOT223 parts, and a few use the DPAK parts. Most don't even have a 3.3V reg. On
my own Arduino form factor pcb, I ended up with a TO220 5V reg and a DPAK 3.3V reg, both 1 Amp. That's my very
conservative idea of "general-purpose", where anyone using it will be less likely to burn the thing up.

At least you have some other considerations to think about. Good luck.

Last point, I might try buying 10 proto pcbs for $100 first, and try out the circuit, before jumping in and
buying 1000 bds.

Dan:

I understand your concerns, and even though I still don't think heat will be an issue because the regulator will be off 95% of the time, (I can't control what people hook to the board and even a 5A reg will get hot if they short it, but it's not an Arduino so most shouldn't be connecting anything too crazy) I'll take another look and see what I can find as far as a 3v3 regulator goes. I'm not against less heat and more power.

As far as size goes, the board's already at the absolute maximum. It has to fit inside small props, and I've done a lot of research to determine exactly how wide and tall it can be and still fit inside most, and it really can't be any larger than it is now. So at this point, my goal is simply to maximize the power output at that size, without driving the cost up too much.

And as far as prototypes go, while I don't plan to have 1000 of these made, (I wish! ) with the first batch likely being around 50-100 boards, I will of course have a prototype assembled first to make sure everything is in order. Doing my best to ensure that the board works the first time though; that's why I'm asking questions to verify if my assumptions are correct or not. :slight_smile:

I always use the 10 protos for $100 service first. My last 3 pcbs each had exactly "1" nontrivial
layout error, which makes me ever leary of getting more than 10 the first go. No matter how many
checks I do, some little error always creeps in. But then they're a little larger than yours, 3"x3"
to 3"x4" and 500-600 pads or so.

All the PCBs I have made so far had "errors", but all manageable, for instance no MCLR resistor.

I saw the photo from the board layout, having so many row headers is a lot of effort to solder, why don't you just only install a few row headers, and use flat cable connectors for the rest of the pins?

both the connectors and cable assemblies are available from a distributor I know, for good prices, I can add a link if you are interested.

Board size could become reduced considerably as well the efforts to solder all these pins one by one (flat cable connectors could be reflowed or row soldered).

I've had pretty good luck with my PCB's so far. :slight_smile:

I had one problem with the second circuit I designed where I miscalculated the current capability of the Arduino Pro Mini's onboard regulator (it was piggybacked on my board and driving an array of 64 leds) so I had to work up a solution by cutting a trace and wiring in an offboard regulator.

And with my third circuit, which was far more complex, (and similar to, but less complex than this design, and which used through hole parts) the only issue was with the pad sizes for a library part I downloaded for a 2mm connector. The holes were too small. Ended up having to wire that end directly into the board and put the connector on the board it attached to instead.

Not saying I won't have any issues with this board, but I'm cautiously optimistic. Double checking all those pad hole sizes for sure.

Takao:

I saw the photo from the board layout, having so many row headers is a lot of effort to solder, why don't you just only install a few row headers, and use flat cable connectors for the rest of the pins?

Several reasons:

  1. I won't be soldering these. They'll be assembled by the PCB manufacturer. I don't have any experience soldering SMD components and my last board design which was half as complex as this took me 10 hours to solder each board together.

  2. Flat cable connectors take up a lot of space.

Using them for the long LED pin headers would require me to leave .100" of room all the way around like the ISP header. Except unlike the ISP header, which is just a pin strip, and thus leaves room for me to sneak parts under the socket, the shrouded headers I assume you're suggesting for the LEDs would require that space to be entirely bare. I'd either have to make the board larger or put components on both sides to accommodate that.

  1. Flat cable and IDC sockets are less flexible and harder to use.

When I say less flexible, I don't mean the cable can bend less easily than individual wires (though that is true, and a potential concern) but rather that you can't stick a six pin socket in one location, and an eight pin socket somewhere else. It's all or nothing. This board is designed to be modular, and I'd like someone to be able to attach a 10 segment bargraph to a 10 pin socket (plus 1 for a common anode), and another group of 3 leds to a 6 pin socket, and plug those in wherever they wish, separately.

It's also more difficult to separate and strip ribbon cable. And with this setup you don't even need to strip anything. You can just buy some sockets and wires from Pololu with crimped terminals on both ends, snip the leads on the LED down to 0.25", plug the LED into one socket, and plug the other end into the board. I've used this setup myself on many occasions, and it works great. Also, as an alternative, if you really want the LED to be permanently affixed, you can leave the terminals unsocketed, slip them onto the LED, and solder them in place. Either of these is easier for a beginner than soldering bre wire leads on and crimping their own connectors.

It's more expensive, sure, but these boards are geared towards people who might not even own a soldering iron, so it's the perfect solution for them. And for me, it's worth spending a few bucks on connectors and crimped wires if it saves me hours of stripping wires, soldering, crimping, and heat shrinking.

Also in regards to flexibility, the pins along the top need to allow for servo connectors.

Hm... I just found this regulator:

Can supply 1A, is $1.30, and when mounted on 1 sq in of copper, the SOT-223 model has a thermal resistance of 70C/W, compared to 206C/W for the 150mA $0.50 SOT-23 reg.

So for 80 cents more I get a much more capable part, which may survive if someone does something crazy like hook three green lasers to it.

But to be sure let's do the power calculations:

?j-a = 70°C/W
Tj(max) = 150°C

Pd(max) = (Tj(max) - Ta) / ?j-a
Pd(max) = (150°C - Ta) / 70°C/W
Pd(max) = 1.78W @ Ta = 25°C
Pd(max) = 1.07W @ Ta = 75°C

Then:

A = Pd(max) / (Vin - Vout)
A = 1.78 / (5 - 3.3) = 1.04 @ 25°C (Piggybacking off my 5v regulator)
A = 1.07 / (5 - 3.3) = .629 @ 75°C

So based on this, the part would likely survive if someone hooked three green lasers (~200mA each) up to it. But I don't expect anyone will do that because green lasers are really expensive.

scswift:
It's also more difficult to separate and strip ribbon cable.

I assume, when I talk about flat cable, you think of these used for ATA hard drives, and legacy floppy drives?

I mean the type used for modern LC displays, and generally more and more found inside consumer electronics gadgets.

http://radionics.rs-online.com/web/p/ffc-fpc/5351915/
1.25mm pitch, through hole, not true SMD

http://radionics.rs-online.com/web/p/ffc-fpc/7388940/
0.5mm pitch, SMD

and the cable assemblies: http://radionics.rs-online.com/web/c/connectors/general-purpose/cable/

these assemblies are never cut, they are purchased ready-made.

The idea is to save time (the connectors can be soldered at once), space (they are much smaller than row headers), and in some cases, it's also cheaper to use a ready-made flat cable.

My idea was to have a number of conventional row headers, and if these run out, flat cables are used.

The technology is around since the 1980s but I have seen a lot of increase in recent years.
1.25mm is easy to use, still through hole.

From datasheet:
The SOT-223 package has a ?J-A of approximately 125°C/W when soldered down to a minimum sized pattern (less than 0.1 square inch) and approximately 70°C/W when soldered to a copper area of one square inch.

I think you're still being a little optimistic in choosing the 70 value versus the 125 degree value, since 1"
square of copper means just that, but all in all, probably a better thrmal choice than the others.

I think you're still being a little optimistic in choosing the 70 value versus the 125 degree value, since 1" square of copper means just that

Well, it's true that the expansion board this will be placed on is a bit less than a full square inch, but I'll have thermal vias going to the ground plane like on my 5v regulator, so it should function as a reasonable heatsink. Also, there's a mounting hole right next to it which is also a via connected to the ground plane and I have thermals disabled on vias, so the user could attach the board to their chassis to help with the heat dissipation if it does become an issue.

I've done the best I can with the limitations I've set for size and cost. It's gonna cost me over twice as much for these regulators, but you're probably right that I needed the extra current capacity to be on the safe side.

I assume, when I talk about flat cable, you think of these used for ATA hard drives, and legacy floppy drives?
I mean the type used for modern LC displays, and generally more and more found inside consumer electronics gadgets.

Ah, flat flexible cable. Yes, I'm familiar with that. I tried to use it on my last board but had to go with much more expensive 2mm headers because it couldn't bend the way I needed it to.

The idea is to save time (the connectors can be soldered at once), space (they are much smaller than row headers), and in some cases, it's also cheaper to use a ready-made flat cable.
My idea was to have a number of conventional row headers, and if these run out, flat cables are used.

Yeah, I'd love to be able to use those, but they're just not suitable for this design. Those don't work well when you don't know where someone will need to mount a part or how far from the board it will have to be. This board could be mounted in the handle of a gun, and the leads to the leds at the tip might need to be 2' long. Same thing goes for the expansion boards. Main board might fit in the handle, but the expansion boards might need to go in the body, and FFC doesn't bend around corners sideways.

Well, it's true that the expansion board this will be placed on is a bit less than a full square inch, but I'll have thermal vias going to the ground plane like on my 5v regulator, so it should function as a reasonable heatsink.

Full square inch of copper isn't the same as saying 1" square pcb. That's what the other number means:

"approximately 125°C/W when soldered down to a minimum sized pattern (less than 0.1 square inch)".

@scswift

I understand your considerations. As well the end user may not always have access to a distributor carrying these cables.

Myself I try to move away from any kind of parallel bus or connector where possible.

I figured out, when I made PCBs for PICs, and included many headers/jumpers/extra parts, most of the time indeed just a blank PCB was needed.

However I do not know the depth of the design or the considerations for instance using it inside guns.

I even changed away from headers for the ICSP programming, just pushing a connector inside the holes, I also see this practice on many commercial PCBs.

What I do not fully understand is why many people actually are afraid of switching ICs, there are many different types, these big (I think DPAK) LM25xx running at low frequency, or also some smaller 500mA SOIC chips @ 2 MHz. I'd understand it for a DIL 2.54mm only design, but if you use TQFP/SMD anyway? Commercially, 7805 has been used in the early 1980s, there is also 3 Amp version, and somehow this regulator is prone to failure after a decade or so, due to heat stress on other components, even failing solder joints. There has been one in the SEGA Master System as well SEGA Megadrive.

I used these regulators too when I was younger, as a teenager.

Well somehow I don't get it the PCB design looks expensive I'd guess at least $10 for each PCB, but you are concerned about this voltage regulator.

I'd never start calculations for a voltage regulator, anyway. I get some ideas from commercial applicances which I examine, and they do fail after some years, parts of the PCBs show heat stress, and all this. But, I think it is because they are cost optimized too much in some cases. One could think they do not always calculate sometimes just replace with cheaper parts and well, total lifetime becomes shortened.

Old capacitors from the 1980s still work fine but only if they did not experience heat stress...

Full square inch of copper isn't the same as saying 1" square pcb. That's what the other number means:
"approximately 125°C/W when soldered down to a minimum sized pattern (less than 0.1 square inch)".

I don't follow.

The way I interpret that, it's 125°C/W if I just solder it to pads.

But if I have a ground plane, and attach it to that via thermals, that's going to be a big sheet of copper, and the thermal resistance should be much less. (I guess the ground plane will be broken up a bit by traces on the main board, but that will be minimal on the expansion board we're talking about here.)

I guess if you feel thermals will be inadequate to transfer the heat the ground plane, I can place a polygon under the regulator to provide some additional copper as I have with my LED drivers per TI's specs.