5V regulator overheating with only a backlit parallax 2x16 LCD
eagletalontim
Posts: 1,399
I have a circuit I am testing out and am trying to figure out this extremely hot regulator issue. The regulator I am using is an LM2940S (Surface mount LDO 5V regulator) that is connected to a 12V battery for testing. I have filter caps after the regulator and before the regulator along with a TVS diode, inductor, and a 1N4004 diode to prevent reverse connection issues. The ONLY time this heat issue happens is when I connect the backlit display and turn on the back light. The temp of the regulator gets up to 146F degrees in 3 minutes. It has not reached the auto cut off limit yet, but I don't want to push it that far. Since this is a surface mount regulator, I added 8 vias on the pad to help transfer heat to the large ground plane on the back of the circuit board. This does not appear to be enough. I would like to run the backlight so I can see the screen at night, but with this heat issue, I am afraid I am going to pop something. Any ideas on how to reduce the heat massively without losing the back light?
Comments
Quick solution is either to add resistors in the 12V feed to drop some of the voltage but they will get hot too or you could PWM the backlight or reduce the current so that the average power is much less than it is now. Alternatively the backlight could be run from 12V through a suitable resistor or in combination with PWM which is a method I have used before and only needed a small SOT-23 transistor.
The answer is to reduce the input voltage. If you have to use a 12V input, I'd add a switching pre-regulator with a 6V output. That's enough headroom for an LDO regulator. You wouldn't need the diode or inductor since the input to the 5V regulator is regulated to 6V. Dimension Engineering makes an adjustable 1A and 3A switching regulator that's very compact. You can also roll your own.
Well, I'm sure I've said it many times before that I switch-mode regulate to 5V and linear regulate to 3.3V so "switching noise" is not even a consideration then. However a proper ADC will have a filtered precision reference voltage anyway and be less affected by supply noise.
BTW, 8-bit is hardly anything fancy and even if you had noise it would probably only amount to a bit or two.
I've had trouble with radio interference using switching regulators but if Peter doesn't think the switching regulator would cause a problem from your ADC chip, I'd believe him.
I think if anyone has had problems with switching noise then it may have had more to do with poor layout or even using breadboard (and poor layout). Just remember that any switching device, be it logic, regulator, driver etc can cause problems with analog circuits if it is implemented with a poor understanding of electronics which is manifest in the circuit design and layout. For instance, a simple rule, don't connect your power load (motors) ground to the logic ground which is then connected to the regulator ground, run these separately back to the power source ground. The same goes for analog grounds, don't run them "through" the digital ground but connect them separately back to the source. Same goes for the supply, I would use another regulator just for the analog in some cases. BTW, you have heard of Prop's being so severely affected that they are destroyed because power signals have been run through logic ground and even sometimes through the Prop's multiple ground pins itself (always connect every pin to the same ground). The Prop's not to blame, the load is not to blame, the power supply is not to blame, but some id 10 T is!
I've read (and think I follow) the circuit design rules you just listed. It's my understanding radio noise is a different beast than a noisy supply voltage. Don't many switching regulators produce radio interference no matter how well laid out the circuit is? I understand (from Mike Green) there are ways to block this noise but I thought this required some sort of metal box to house the offending circuit.
I realize the OP was concerned with electronic noise not radio noise but you quoted my statement about radio interference so I wanted to make sure I understood your reply.
(I'm really glad I didn't ask what an "id 10 t" was (I wont say how close I came to doing so). I would have felt like an idiot.)
Just a low cost alternative.
FF
You can't go past these little 7805 switching regulator replacements, the R-78E5 which are under $3 in one off.They will take up to 28V in and deliver 5V @500ma. Being a 3 pin module they are very easy to use.
There is a minimum load current spec of 5mA. So add a 1K load resistor to be safe.
And the 50mV ripple thing.
They really look nice though at $2.84us.
Duane J
Nice find Peter. That can make PCB design a lot easier. Put in one pad, and later the choice can be made regarding which regulator to use.
re
it can be a tricky compromise. You put in a small linear regulator but then need a larger PCB footprint or a vertical mounted heatsink to dissipate the heat. Peter's module may well be the smallest and cheapest solution.
There is a 3.3V R-78E3.3 out there too. Searching around for the best prices...
Quick and easy solution is to PWM the backlight through an NPN from the Prop etc. Usually you don't need your backlight running at full brightness as they are normally transreflective and so you only really need the backlight when it's dark and then you don;t need a bright backlight. You will find that even a 50% duty cycle is bright enough and the linear regulator will thank you and not lose it's cool (too much). You don't really need PWM as you can cheat with the counter duty mode, the 40MHz square wave will "smooth out"
Is there anything that I would need to change in the circuit besides the regulator?
I've run the backlight LED on those 1602 displays w/ PWM. It measured almost 100 mA with PWM at $FF, but it can be driven at 1/4 level ($40) with very decent brightness with only 30mA drawn. The Arduino Shield version has the backlight control pin on D10 of the arduino connector already buffered, it runs into the base of an NPN transistor. I don't know how it's handled in the Parallax serial version, it's the same display, and there could be something similar on the controller board, or it could be added. Or just use fixed resistors, it's very legible at 1/4 voltage.
A string of regular 1N400x will work fine.
Databook says 1 volt drop per diode for 1N4001.
But there are other approaches too.
The main idea is to reduce to Vin so that there is less voltage dropped across the regulator.
From what you have described, THAT"S the problem here.
The backlight on that display only pulls 80 ma!
A good tutorial here. Junction diodes start on page 3
http://www.electronics-tutorials.ws/diode/diode_1.html
You might have to scrap your PSU then and do a new one especially since your system can't cope with 12V then the 14 or 15V typical that you will find when the engine is running will surely cook your goose. I don't think you mentioned a red LED display before but that could draw a bit. 5V linear regulators are from another time when power supplies were constructed with good solid transformers and big heatsinks and the typical input voltage for the 5V side was around 8V so that the regulator would not get too hot. Many people look at the current rating of a linear regulator and think that this means it can handle the current, which it can, but what it can't handle is the heat if the input/output differential * current is too high. They find uses these days for lower input voltages or lower currents, the rest of the time it's done by SMPS.
Of course with switching regulators they perform a power conversion with a small amount of loss so you can expect your 12V side to draw around about half the current that the 5V side is drawing.
http://www.reuk.co.uk/Zener-Diode-Voltage-Regulator.htm
You just have to crunch the Ohm's Law numbers in consideration of your worst case loads, so your resistors and zeners can handle the dissipated heat, yet be operating in their optimal region to regulate even at minimum and maximum load. That way the 5v and 3.3v regulators don't need a monster heatsink. It's a fun exercise. You have to use some electrolytic filter caps anyway to stiffen up response to changing load and to keep any residual ripple acceptable, and add a few bypass ceramic caps to constrain any noise arising in whichever part of the circuit, so it doesn't propagate. If you drop the voltage gradually, rather than all at once, these dirt cheap and tiny components can work wonders in a very small space. Simple and cheap.
Of course thermodynamics can't be bypassed, about the same total amount of heat WILL be dissipated, and once you get into it, you'll see that zeners are not the ideal regulators, one issue being that you always waste some energy keeping them in the knee region, but by spreading the heat around instead of focusing it all on the 5v 3 terminal reg, you can manage it better. If that's not good enough, then you can go with an external wall wort supply say at 8-9 volts, or go all the way and use switching regulators. But IMHO first try to reduce the current used and see if you can spread the remaining heat well enough to KISS.
Have you considered the temperature it will be used in? If you have to cover a wide range of added modules, and handle the Arabian desert and Antarctica, not to speak of storms, lightning, sea spray corroding the connectors, you're just getting started. And can you tolerate failure? At what statistical rate? Is operator error a valid excuse? Might it kill people if it fails, sparks and ignites fuel? Must it meet "explosion-proof' specs, like for ammunition-handling machines, where the failure of ANY two components in the circuit at once must never EVER possibly lead to a temperature high enough to detonate gasoline, or other "stuff".
Having worked in such environments, I can tell you it's all about money and compromises. Ain't nothing plug and play that will cover all the contingencies. So you may as well start out with some cost guidelines, and unless you have no monetary limits, decide how to balance it for the cases that matter. Linear Regulators run hot, that's what they do. They have thermal protection and shut-down built-in. You put a heat sink on 'em, drop as much voltage up front as is practical, and move on. If the switchers work for you, cheap enough, and you know how to keep the noise down and keep them stable, that's fine too.
Your design has a clean look and includes almost everything that is needed to efficiently do its job but, since this is the first time i'm seeing your power supply in the other thread and you are asking for help at both.
I have a few sugestions to improve it but i have no access to a CAD package, so I will try to describe my thoughts into words.
- Since you have a large ground plane at the bottom layer, do use it. Connect each SMT ground solder pad to it thru as many vias as you could add in the vicinity of the solder joint. Short rectangular traces could be used to work the geometric transition between the solder pad and the "stitching" vias. They will also help in diminish unwanted heat dissipation during the soldering process.
- Connectors are known as being a nightmare source for peeling copper in the underneath of its termoplastics housing or pin holding tabs, even when slight bending forces are applied to their pins. Avoid running traces that can be blocked from being visualy inspected by a termoplastics tab or housing. Use both layers and vias to "stitch" their solder pads irrespective to the fact they are from the trough-hole or smt kind. During debugging, the more the vias "stitches", the less your eyeballs and head itches.
- At any cost, avoid using thin traces to connect something. Capacitors in special will have their expected action derated by any intervening trace impedance. Placing them as close as possible to the regulators input and output pins will improve the overall efficiency of the design. Since C2 and C3 are there to "bridge" any transient noise, place them as "bridges" right between regulator pins, if possible. Surely provided it will not make them too hard to solder without slanting/shorting/removing the other one.
- Extending the former a bit, always balance component dimensions looking for the best soldering/reworking conditions. Unless you have a lot of experience, and even there, an hyper clean minimalist look was almost never reached at the first try. Experimentation and observation will always drive your designs to perfection.
- If your mounting requirements do not call for that exact layout, please relocate the output connector to the right side of the board and place it vertically. This will shorten output leads and provide an extra gain in lowering the overall output resistance.
- Always remember: copper will be removed from the layers during board fabrication, so the larger the remaining traces (in power supply design), the better. Thin traces will degrade both the regulation and the transient response.
- If, for any reason, you choose to replicate at one of the layers a connection that is already done at the other one, make sure you place as many vias between the layers as you can when running the traces. Vias will certainly add to the board cost, but also adds a lot in current and heat dissipation.
- Almost any voltage regulator design will be improved by a minimum current comsumption to ensure optimum performance, so provide it in the form of a fixed resistor right at the power supply. If you have enough room for it and its current needs do fill the specs, a resistor/led couple will be a nice (and noticeable) adition.
- TVS's will do its job shorting voltage transients, so run its leads to the power and ground layers or pins using vias as stated above. If there is a suitable fuse at the power input lead, check its specs against the choosen TVS to ensure the fuse will be blown by the TVS and not the reverse. If direct connection to the power source is intended, provide the fuse at the power supply. If the power source includes a battery, the fuse is almost mandatory. The same applies to the choosen input/output connector regarding its power handling specs. More expenses, less damages.
- If you must use thin wires to run connections to the power supply, replicate them at the cost of replicating pins at the connector, as impedance matters when dealing with currents and noise specs. If you have two circuits to power from the same source and one of them will use a lot more of the available power than the other, consider having two connectors to power them independently, even if you need to parallel them at the power suply. If the less consuming one will be controlling the power hungry one, be careful when you design the control interconnections to avoid creating ground loops and possibly degrading the overall performance. Also avoid transiting high currents at the same wires you use to power noise sensitive circuitry, for the above stated reasons.
- Remember: There are ocasions when only local filtering/regulation can ensure intended circuit operation. Being open-minded and allowing for a bit of experimentantion will always result in a better final product.
I hope it will help you a bit,
Yanomani
The LCD backlighting is a power hog and in some cases requires a non-standard voltage. Plus, a separate power source can allow you to dim the back-lighting or turn it off.
One has to be careful to read the literature about surface mount versions of devices, they often cannot dump as much heat as the free-standing alternatives. I was recently reading about LM117 adjustable regulators and the packaging often is the ruling factor in power capacity. The TO220 device is rated to deliver 1.5 amps and the surface mount devices fall all the way down to .250 amps.
I have at times inserted a 7809 regulator between a 12 volt supply and a 7805 just to cool things down. You have to realize that automotive 12 volts can run much higher when the alternator is charging the battery... 14.2 volts is not uncommon. You could even use a LM117 and have it deliver 7.5 volts while taking most of the heat away from the board. It would also block any automotive transisents or junk. You might find something that is even better in a LDO that includes automotive protective features. The LM78xx series and the LM117 are not the most durable for automotive... there are newer and better.
Just remember Loopy that inserting a 7809 is just a kludge or stop-gap measure, you could just as easily insert a large resistor or diodes but all of these still dissipate power as heat. The 1.5A current rating on the TO220 will probably be with a big heatsink and with a supply voltage of 8V otherwise 1.5A x 7V (or more) = 10.5W. On a TO220 the thermal resistance without heatsink is 50'C/W so that's a complete meltdown or a very large heatsink (think brick) to spread the heat and hopefully dissipate it somewhere (think fan and ventilation).
The surprising thing about SMD versions of the same chip is that it is possible to handle more power without a traditional heatsink as you can use the copper and the PCB to spread the heat. So you would use thermal vias and as the regulator is soldered to the "heatsink" it has very low thermal resistance. Saying this I have also soldered the tabs of TO220s (as current loop regulators) straight to the pcb in SMD fashion to avoid nuts and bolts and heatsinks and grease etc. The copper is exposed without solder mask and I do have thermal vias to more copper on the other side as well as making sure the tab is absolutely flush with the pcb (clamped) when it is soldered (needs a big iron).