What to look for in a low dropout regulator (LDO)?

Discussion in 'General Electronics Chat' started by osx-addict, Jul 14, 2012.

  1. osx-addict

    Thread Starter Member

    Feb 9, 2012
    122
    9
    Hi all.. I'd like to use a pair of LDO's (not sure why they're called LDO's -- why not LDR's??) to go from 12VDC down to 5VDC with one LDO and then the 2nd LDO would step the 5VDC down to 3.3VDC.. However, in picking the LDO's I'm not sure what qualities to look for -- specifically the low dropout voltage and load regulation.. I'm assuming the load regulation is how close the output is to the desired target voltage -- correct? However, what about the other one.. In doing a search on Mouser I see a range from 0.03V@100uA to 1400mv.. I'm lost!! Please advise! :confused::confused:

    P.S. This is NOT for a car or anything automotive
     
  2. upand_at_them

    Active Member

    May 15, 2010
    246
    29
    The whole point of the Low Drop-Out regulator is to have an input voltage not much larger then the output. If you're going from 12V to 5V...Why do you want an LDO?
     
  3. crutschow

    Expert

    Mar 14, 2008
    13,052
    3,244
    It's an LDO (Low Drop Out) Regulator not an LDO. LDR might more sense but the convention is LDO Regulator.

    As upand_at_them noted you don't need an LDO regulator for goring from 12V to 5V. The LDO means that the minimum voltage between the regulator input and output for proper regulation is typically a volt or less compared to the 2-3V for a standard regulator requires. It has nothing to do with the accuracy or stability of the output regulation.
     
  4. #12

    Expert

    Nov 30, 2010
    16,343
    6,828
    What to look for in a low dropout regulator?
    Low dropout voltage at the current required.
    Load regulation is the amount the output voltage changes when the load changes from minimum to maximum current.
    The required maximum dropout voltage depends on the circuit you are building. I am not surprised that the dropout voltage specifications don't make sense when building a 12 volt to 5 volt regulator circuit because any regulator chip will do that.
    To get from 5V to 3.3 volts requires a dropout voltage of not more than 1.7 at the desired current. However, you should try for 1.5 volts or less in case the 5 volt regulator is a biit off its voltage specification.
     
  5. osx-addict

    Thread Starter Member

    Feb 9, 2012
    122
    9
    Thanks guys.. I was looking at another schematic (for the Chipkit Uno32) and it uses two of these (the first is an NCP1117ZDT50T to reduce down to 5VDC and then a 2nd LDO to further reduce down to 3.3v) -- all from an input supply of between 7-15VDC.. However, that first LDO above claims to have an input as high as 20VDC hence the interest to use it regardless... However, I'll look at a more traditional regulator which may be more practical for this case but will still look into using one for the step-down from 5V to 3.3 as it sounds like a reasonable use-case. Thanks as usual!

    P.S. Since I see that the above Chipkit Uno is designed to operate off of a 7-15vdc supply, I'll probably just follow suit and go down the same path -- the parts are nice & cheap and I'm not going out of spec for anything in terms of what they're capable of doing. I might consider adding a diode on the input to care for reverse polarity issues...
     
    Last edited: Jul 14, 2012
  6. hairball45

    New Member

    Jul 13, 2012
    6
    0
    5v to 3.3 volt? a diode.. 1.7 volt drop?
     
  7. bountyhunter

    Well-Known Member

    Sep 7, 2009
    2,498
    507
    Low Drop Out


    With 7V of headroom, you don't need an LDO. Use a 7805.

    Load current. ???
     
  8. ErnieM

    AAC Fanatic!

    Apr 24, 2011
    7,394
    1,606
    For 2 regulators such as this the power used is the same if you have 12V->5V-3.3V or separate 12V->5V and 12V->3.3V. The main difference is the first way there are fewer constraints on the 3.3V regulator, as it has a narrow and predictable input of 5V.

    The 5V regulator does not need to be an LDO unless your 12V can drop to the 7 or 8 V range where a 7805 would start to loose regulation.

    "Modern" LDO's (compared to the 7805 which was designed 30-40 years back) are very good devices in that they have very low quiescent currents (the current to run themselves) and very often have nice features such as an output inhibit pin. Besides just regulating power they are also quite useful to drive higher current outputs as the enable pin can come directly from a micro to switch a large current. I did that once to drive some LED backlights as an LDO was actually the same cost but a few less components then using a simple transistor.
     
  9. steveb

    Senior Member

    Jul 3, 2008
    2,433
    469
    Something to keep in mind is the specification of power supply noise rejection. This is less of an issue now than it was 20 years ago, but LDO regulators can have poor power supply noise rejection.

    Most of the first LDO reg. designs to come out used a PNP (or P-channel) ouptut transistor driven in a voltage mode. This allowed low dropout since only the saturation voltage of the transistor kept it from hitting the supply rail. However, the driving in a voltage mode allowed direct feedthrough of the higher frequency power supply noise into the output.

    Only the feedback regulation (which is limited by feedback bandwidth) suppressed the noise at low frequency. In the early days, I tried one device that had noise rejection up to 1 kHz only. This was unacceptable and caused A/D converters to be noisy. Output caps helped to a certain extent, but having too much output capacitance is often not good. I ended up making my own LDO reg. design with an opamp and a current drive on an output PNP transistor. This allowed noise rejection based on feedback, which worked to about 1 MHz. Small output caps can take care of anything above this.

    Modern day ICs that are based on good design priciples should get to 10 MHz (EDIT: see correction below in post # 13) in feedback-based noise rejection. Hopefully, the old crap designs are discontinued, but I mention this issue just in case a few lemons are still out there. The power supply noise rejection specs should be given for any good device. If not, I'd look for another device.
     
    Last edited: Jul 16, 2012
  10. osx-addict

    Thread Starter Member

    Feb 9, 2012
    122
    9
    Thanks guys.. I realize I can use the age-old 7805 or equiv and be done with the 12v-->5v direction. I guess I was thinking, based on some playing around I did a LONG time ago, that these parts run warm -- and may need a heatsink.. I found a thread elsewhere indicating that an LM7805 going from 15V down to 5V may run in the range of 111C -- assuming that's true, it seems like I might need a heatsink which would be an additional cost on top of needing additional board space. However, IF I go with the NCP part mentioned above (spec sheet here), their material indicates that I can use the PCB copper as the heat shield but I'm still not sure about the expected temperature based on my input/output voltages. I did see that it lists thermal resistance at 15C/watt (assuming I'm interpreting the material correctly which is a big 'if').. Unfortunately I'm not sure what my current draw will be yet.. I don't believe it will be even close to 500mw but it's a bit early to tell yet.

    Steveb -- if I want to look for feedback based noise rejection, what are some good search terms for the above PDF (or others)? I see references to ripple rejection and RMS output noise -- I'm thinking that the RMS output noise might be what you're talking about.. Can you confirm?
     
  11. steveb

    Senior Member

    Jul 3, 2008
    2,433
    469
    The power supply ripple rejection relates to the feedback based noise suppression for power supply noise getting into the output. Figures 10 and 11 give the details for this specification. In particular Figure 11 shows the frequency dependence. The rejection is reasonably good up to 10 kHz, and it is adequate up to 100 kHz. Beyond that, it is not going to be great. Where this matters is if you are using a switching power supply with high frequency switching. That part will clearly be effective at reduced 50 or 60 Hz line frequency ripple, but if you have a 100 kHz to 1 MHz switching supply, it might not be a great part to use.

    The RMS output noise is also important, depending on the details of your application.
     
  12. takao21203

    Distinguished Member

    Apr 28, 2012
    3,577
    463
    Easy to cancel with a reactor choke. If you have some around just try and use the smallest/cheapest that will do...
     
  13. bountyhunter

    Well-Known Member

    Sep 7, 2009
    2,498
    507
    That's a bit optimistic. The widest bandwidth LDO that I have seen had a unity-gain bandwidth of about 3 MHz, which means the loop gain is zero at that frequency. The more typical LDO's have unity gain bandwidth of about 100 - 200 kHz. BTW, bandwidth on an LDO varies with load current. At lighter load (higher load resistance) the bandwidth goes down.
     
    steveb likes this.
  14. steveb

    Senior Member

    Jul 3, 2008
    2,433
    469
    If you can completely cancel it easily, then I want to get some lessons from you. :)

    Based on my experience, I would say it's easy to supress to first order. Still, I've found it can be challenging to completely eliminate the effects of switching noise, and some systems (e.g. high resolution A/D) are so sensitive that good rejection from the regulator is a good additional feature to include.

    Also, even if one succeeds in elimnating negative effects of the switcher noise itself, there is risk that a sensitive system might be impacted from injected noise sources from the environment in the application.

    There are also PCB layout issues to consider. So I'm not trying to address all aspects of the noise issues, but only trying to help identify important regulator specs to look at when chosing an appropriate device.
     
  15. steveb

    Senior Member

    Jul 3, 2008
    2,433
    469
    OK, thanks. I trust you. I'm going from memory, and I haven't looked at these specs in over 10 years.
     
  16. bountyhunter

    Well-Known Member

    Sep 7, 2009
    2,498
    507
    Where was that magic choke all the times I was in the EMI test room ripping my hair out trying to get the stupid switcher supply to meet the emission profile.:mad:
     
  17. bountyhunter

    Well-Known Member

    Sep 7, 2009
    2,498
    507
    The reason I remember ripple rejection so clearly is that it is the single most misunderstood spec in all of linear power supply design. While many LDO's have "bandwidth" out to 100kHz or even a bit more, the loop's ability to reject noise is very low near the crossover frequency.... but the worst mistake is that ripple rejection only applies to sine waves at the specified frequency.

    Switcher output voltage ripple is typically similar to a square wave in shape: much of it's frequency content is many Mega Hertz.

    I wish I had a dollar for every irate customer who called up demanding to know whay the LDO, which had 20 dB of ripple rejection at 50 kHz, couldn't "filter out" the 50 kHz switcher ripple shooting right through it.....
     
    steveb likes this.
  18. osx-addict

    Thread Starter Member

    Feb 9, 2012
    122
    9
    Is it safe to assume (or not?) that IF the power being fed into the LDO (12v in this case) was direct from a battery and NOT any sort of power supply that the noise would be non-existant? I'm assuming the noise would be fed in by some 'up-stream' power supply whether it be a wall-wart or computer power supply,etc..? Is this a bad assumption?
     
  19. steveb

    Senior Member

    Jul 3, 2008
    2,433
    469
    Generally, assuming noise is non-existant is a risky assumption. Sometimes you can get away with it, but remember that noise sources (of various types) are always present.

    In this case, a battery is not going to have 50-60 Hz ripple, or switching noise, but noise can still be picked up on the connection wires. More troublesome, time dependent loading on the battery will effectively create ripple on the voltage, at your LDO regulator input. Remember, the connection to the battery has resistance and inductance. Also, the battery has internal resistance. Hence, any loading variation (either due to the LDO reg load or other system loads) will induce supply ripple at the input of the regulator.

    Of course, placing capacitors on the input voltage to the regulator, right next to the regulator terminal will help filter the input ripple.

    That is difficult to answer. The application is the application. How will your system be used? That is what you need to find out in order to properly design your system for the application. Still, it is a good idea to make the system robust for various input supply types, whether wall-wart, switcher, battery etc. So, I would not say it's a bad assumption at all, but I would say assumptions are often risky.
     
    osx-addict likes this.
  20. crutschow

    Expert

    Mar 14, 2008
    13,052
    3,244
    The magic sometimes is adding common-mode chokes on both input and output power and ground since conducted EMI can be ground noise currents caused by capacitive coupling from the switcher switching transistors to the grounded heat sink.
     
    osx-addict likes this.
Loading...