Shop OBEX P1 Docs P2 Docs Learn Events
Hot regulator — Parallax Forums

Hot regulator

Erik FriesenErik Friesen Posts: 1,071
edited 2007-12-28 21:33 in Propeller 1
Any comments on why my 3.3v reg on my board is getting quite hot to the touch?· Is this quite normal for sot-223 to run hot?· The prop is cool to the touch.· I am posting my schematic for reference.· The 8.0v dpk also runs quite warm also. I estimate the temp to be 140F+
bmp
bmp
carcompv1.01.bmp
817K

Comments

  • Mike GreenMike Green Posts: 23,101
    edited 2007-12-22 16:50
    It all depends on how much load you have on it. I posted a description in the "Completed Projects" forum about a multichannel sound sequence that uses a uOLED-128 display, 12 reed relays, and 2 VMusic2 players on a Protoboard. At 9V input, the regulators run very hot. At 7.5V they run hot, but not excessively so. At 6V, they were just warm. I ended up epoxying a heat sink across them using a metal-filled epoxy and that helped a lot. I needed at least 7.5V to run the audio amplifiers off the unregulated input.

    I'm pretty sure the current drawn was at least 300-500ma. With a 9V input, that's 2W dissipation in the 5V regulator and maybe 1/2W in the 3.3V regulator. On the Protoboard, the regulators are cascaded. In your case, you've got a 4V drop across the 8V regulator, then a 3V drop across the 5V regulator and a 4.7V drop across the 3.3V regulator. All of that has to be dissipated as heat based on the current load on each regulator. You will either need to heatsink them or use a switching regulator instead.

    Post Edited (Mike Green) : 12/22/2007 4:56:44 PM GMT
  • deSilvadeSilva Posts: 2,967
    edited 2007-12-22 16:56
    I should think this is normal. SOT-223 and DPAKs (which I like better) are such tiny pieces. Guessing from your schematics you will draw 100mA+ @3V3, maybe the same @ 5V... So you disipate 0.4W + at the SOT and 1W + at the DPAK.

    Don't get confused from the large current values in the datasheet; they are ALWAYS meant with appropriate cooling and minimal drop-down smile.gif
  • RaymanRayman Posts: 14,801
    edited 2007-12-22 16:56
    I'm doing a similar thing: I go from 12V down to 9V down to 5V down to 3.3V... It's not very efficient, that's why the regulators get hot...

    Maybe you should put the 3.3V regulator downstream of the 5V so it doesn't have to drop so much voltage... At a minimum, the regulators must dissipate power equal to their voltage drop times output current...

    If you are really drawing a lot of current at 3.3V, you will have to attach a heat sink to the regulator (or switch to a switch mode DC-DC converter, which is more efficient).
  • RaymanRayman Posts: 14,801
    edited 2007-12-22 16:57
    I must be a slow typer... There were no replies when I started [noparse]:)[/noparse]
  • Erik FriesenErik Friesen Posts: 1,071
    edited 2007-12-22 17:06
    Well part of the problem is that I want to be able to run a gps module off of the 5v reg and that makes the 5v reg piping hot if I cascade it. Does the pcb act as a sufficient heatsink in these sm devices? If I do the math like this (8v-3.3=4.7V*.200(ma)=1watt*15c/w=15c temp rise. But somewhere I must be off.
  • Mike GreenMike Green Posts: 23,101
    edited 2007-12-22 17:16
    The heatsinking ability of a PCB depends on how much copper area is physically attached to the SM device and how thick it is. If it's a 4 layer board with power and ground planes, that can help spread out the heat, but you really can't run things very hot to begin with. A PCB is not really meant to be used as a power heatsink. Thin copper foil, glass and resin just don't conduct heat that well compared to a hunk of copper or aluminum.
  • Erik FriesenErik Friesen Posts: 1,071
    edited 2007-12-22 17:19
    What is a good way to heatsink a sot223 case? I haven't had much luck with finding a decent heatsink.
  • RaymanRayman Posts: 14,801
    edited 2007-12-22 17:41
    I've seen a lot of electronics where the heatsink regulator is bolted to a peice of metal (sometimes the case) with some thermally conductive paste inbetween. This is the reason the regulators have the tab with the bolt hole at the top...

    But, you have to worry about electrically insulating the tab too... I like the NJM regulators that have an epoxy covered tab. With those, you don't have to worry about electrical isolation...

    Post Edited (Rayman) : 12/22/2007 6:16:08 PM GMT
  • deSilvadeSilva Posts: 2,967
    edited 2007-12-22 17:54
    There are special SMP "clips" (Assmann e.g.), there are also special techniques to integrate an aluminium pad on the PCB, see here: www.andus.de/Archiv/archiv.htm#0706 (oct 2005)
  • AleAle Posts: 2,363
    edited 2007-12-22 18:25
    deSilva: That shop Andus hat really interesting stuff, a pity I only do electronics as a hobby, and chemistry as a profession smile.gif
  • kittmasterkittmaster Posts: 77
    edited 2007-12-22 18:33
    I agree, you have way to much of a voltage drop across the regulator. I would reroute the anode of the diode from the 8v regulator to the output of the 5 volt regulator ouput. This way the diode drop .7 of the voltage and the regulator will only have to deal with a 1v drop instead of 4, that will reduce the (loosely speaking) around 45% maybe more.
  • Sleazy - GSleazy - G Posts: 79
    edited 2007-12-23 21:56
    Actually , the heat sink efficiencies are greatly related to fin shape and length, finish (polish or matte) and color, solid angle viewfactors of adjacent equip and their emissivities/transmissivities/reflectivities/Specific heats/conductivities, contact resistances, thermal strain moduli, yada yada yada. You can get some good estimates for fin design·with·heat and mass transfer formulas, they are·quite·explicit for fins under free convection. They take into account·most of those variables.· If you get·into complex fin shapes and get ready to do·some finite·element analysis. like·FEHT.

    Think about this.· If you make two heatsinks out of thermocouple material,·connect them in series and push current thru them, you·should see the first heatsink in series have a net·heat flux out, and the second should see a net heat flux in, in order to·conserve power.··You are cooling one of the·2 thermocouples with current alone.·Negative thermal resistance , like how evaporative cooling sometimes exhibits, No coolant, no moving parts.· Theres alot you can do with this, not many people know that this exists.·

    Honestly tho if you arent making something hyper precise, the small bit of efficiency you get out of designing the heat sink is not really worth it. Go get a charge pump boost regulator from TI (TPS61081), will do up to 50 v @50 mA all the way to 3.3V, only one external inductor and super small.· Those linear regulators are terribly inefficient compared to the switching charge pumps, which can get to almost 93% efficient or so I believe.·
  • kevin101kevin101 Posts: 55
    edited 2007-12-24 03:18
    In your second paragraph, you describe current running through a heatsink and cooling another heatsink. Isn't this just a peltier junction-a.k.a. thermocoupler? Everything else is just an addition to my "don't get it" list.

    kevin
  • Erik FriesenErik Friesen Posts: 1,071
    edited 2007-12-25 04:13
    Ok but when is a heatsink really needed? Should I go to the bother to figure out how to use a switching regulator when power consumption is of little consequence? Does a switching regulator produce no heat? I suppose If I could use one regulator to drop it from automotive voltage level (14 volts)to 5 volts that would be some advantage if it could be done with little heat.
  • Mike GreenMike Green Posts: 23,101
    edited 2007-12-25 05:56
    "When is a heatsink really needed?"

    You'll have to answer that yourself based on your information about current drain, voltage drops, thermal resistance, etc. This is all Ohms law, the definition of power (Watts = Volts * Amps), and the equivalent of Ohms law for heat. Nothing really complicated.
  • Peter VerkaikPeter Verkaik Posts: 3,956
    edited 2007-12-25 07:17
    For my latest board I use a Recom R-7805-0.5 dc/dc converter that only
    dissipates 0.4W for dc input voltages up to 34V.
    http://www.recom-international.com/switching_regulator_R-78xx.html
    They are as easy to use as a 7805 regulator. Why bother with heatsinks at all?

    regards peter
  • deSilvadeSilva Posts: 2,967
    edited 2007-12-25 09:46
    This RECOM thingy is quite astonishing:
    http://www.recom-international.com/switching_regulator_R-78xx.html

    Some months ago you had only the choice between some additional caps and inductors (5 additional parts) or inductor-less tiny SMD devices providing between 50 and 100 mA.

    But I couldn't find a price tag for the R-78XX smile.gif
  • Nick MuellerNick Mueller Posts: 815
    edited 2007-12-25 10:20
    > But I couldn't find a price tag for the R-78XX smile.gif

    Should be something around 10.- .. 15.- EUR or such.
    A DC/DC 5V/5V converter costs 7.- EUR

    Nick

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Never use force, just go for a bigger hammer!

    The DIY Digital-Readout for mills, lathes etc.:
    YADRO
  • Peter VerkaikPeter Verkaik Posts: 3,956
    edited 2007-12-25 10:23
    I got them from Conrad. There not cheap but Conrad is a consumer supplier
    though I think they also have·industrial sales.
    The ordernumbers are
    154495 (3.3V 1A) SIP3
    154496 (5.0V 1A) SIP3
    154481 (3.3V 0.5A) SIP3
    154483 (5.0V 0.5A) SIP 3

    http://www1.int.conrad.com/scripts/wgate/zcop_in/~flN0YXRlPTExMzI2MzAyNTY=?~template=PCAT_AREA_S_BROWSE&glb_user_js=Y&shop=A_B2C_IN&p_init_ipc=X&~cookies=1

    Edit: here is the pricetag page from Recom
    http://www.recom-power.com/pages/highseller.php

    regards peter

    Post Edited (Peter Verkaik) : 12/25/2007 10:41:20 AM GMT
  • deSilvadeSilva Posts: 2,967
    edited 2007-12-25 10:41
    I have some suspicion where Erik got his SOTs and DPAKs from; if correct he paid 25 Cents each smile.gif

    Edit: RECOM (according to the links) sells for $10.82 (3.3V @ 500mA) and $9,75 (5.0 V @ 500mA; there is also a 1A type of unclear price..)) +VAT of course

    Edit2: CONRAD takes 9,28 € for the 500mA and 10,16 € for the 1A types (incl. VAT)

    Post Edited (deSilva) : 12/25/2007 11:03:44 AM GMT
  • Erik FriesenErik Friesen Posts: 1,071
    edited 2007-12-25 17:16
    Mouser it is. I have been building a number of these things and $10 a regulator is not an option.
  • kevin101kevin101 Posts: 55
    edited 2007-12-28 21:33
    what about this one? www.dimensionengineering.com/DE-SWADJ.htm Dimension engineering offers a bunch of switching regulators
    and they're almost the exact same size as a TO-220. For 15 bucks (plus 1.25 for shipping) You get 1 amp at any voltage between 1.25 and 10.
    NO extra parts needed.
Sign In or Register to comment.