Step-up converter fail

Ok, am baffled enough to ask for help. For those of you of you who know how little I like to ask for help, you’ll understand just how baffled I am.

Yet again, I’m failing to get a step-up converter to work. This one is a MAX608, aiming to give me 7 Volts and 500 mA from two NiMH cells (datasheet). I’ve rebuilt it three times, smaller each time, and it is doing something, but not what I want it to do. Symptoms are:
1) Not exploding, which is good.
2) Responding to the shutdown/enable pin.
3) Producing a 50 kHz waveform on the EXT pin that drives the MOSFET.
4) Producing some voltage, around 2.1 V. This is seems to be just the battery voltage, less the diode drop.
5) The output voltage is controlled by a feedback voltage divider. Shorting this to ground should give a set 5 V output. It doesn’t change the output at all.

Can I just say what awesome close-up pics my phone takes? That’s fives times magnification on my monitor. Nokia E90 – phone and microscope combined.

The square wave on the gate drive pin looks like what it should look like, only I think the high voltage should be VOut, i.e. 7 V. It’s 2.1V, the VIn from the bateries, minus the voltage drop over the diode. The MAX608 uses a PFM switching scheme, with a maximum on -time of 16 us. That’s what I’m getting. The maximum off-time is 2.3 us, this is giving me 3.4 us. So something’s happening. These times are controlled by the current limiting control scheme, so I’m guessing there’s a problem with the current? (Note: the important word here is “guessing”.) At least something’s going on here:

The MOSFET drain/inductor/diode voltage has spikes when switching is taking place, but I’m buggered if I know what this means:

So it appears to me that umm… dunno. Current hands in the air hypothesis is that there’s a problem with the inductor. The inductor isn’t inducting enough, or it is inducting too much. I don’t know what inductors do. Or rather, I can give you the standard A-level electronics answer – they store energy in a magnetic field, and the voltage they produce depends upon the rate of change of current. However, knowing what it is supposed to do is very different from knowing why it’s not doing what it is supposed to do. It appears to be behaving like a piece of wire.

In fact, bypassing it with a piece of wire does nothing except remove the spikes on the drain voltage. So the inductor is doing approximately bugger all.

The inductor I’ve got is whatever RF choke that Jaycar had. I’m using it coz one like it worked in a previous step-up. The ratings are DC Inductance: 22uH Rated Resistance: 0.84ohms DC Current Max: 410mA). I don’t know if any of these numbers are important, but the inductors suggested on the datasheet have full load currents of >1 Amp, DC resistance <0.16 Ohms. So this inductor may not be what's needed. My reading of the datasheet suggests that the inductor's saturation current should be greater than the output, which is should be about 500 mA. Farnell has inductors in a similar package with same inductance, 3 Amp rating, and ferrite core material, so I could try that? But there's lots of guessing here, how do I tell what inductor I need? Alternatively, have I completely misinterpreted the info I have about what this circuit is doing, or not doing, and is there an entirely different reason why I'm getting what I'm getting? Also, Farnell has a $10 minimum order and this inductor costs $3. Anyone want anything? Anyway, thanks to lots of help we have more earth floor:


45 thoughts on “Step-up converter fail”

  1. I’ll copy this to some of my coworkers.
    The first thing that occurs to me is the gate drive doesn’t look like it’s working anything like right — way too low a voltage. Until you can get that, the FET isn’t going to start switching and everything else is going to do nothing. (But on reading the datasheet, I see it wants a low-voltage/logic-level FET, so maybe that’s not as much of a problem.) Their one-shot comparator-based system, with no oscillator, makes it hard to test by feeding voltage into FB to make sure it’s turning on/off when it should. Hrmph. Lemme ask people who know more.

    1. Looking at some of the oscope pics in the datasheet, I think the gate drive should be at VOut, i.e. 7 Volts. I don’t understand how the chip is supposed to produce a high voltage on the gate drive. But cheers for the help.

      1. From looking at the schematic and rereading the datasheet I’m leaning towards the inductor choice being the problem — mostly because everything else looks perfect.
        But I still can’t answer how the chip gets enough voltage to start switching the first time.

        1. The FET should start switching somewhere between 2 and 4 Volts, which suggests it isn’t switching here. There’s a waveform at the gate but only to 2.1 V. So if the FET isn’t switching, then the inductor isn’t getting a chance to do its thing.

          The datasheet does say “Use logic-level or low-threshold N-FETs to ensure the external N-channel MOSFET (N-FET) is turned on completely and that start-up occurs” and I’m using the recommended FET. It also says to use a FET with “a minimum VTH of 0.5V below the minimum input
          voltage”. I’ll try running it off three cells tonight, to see if that makes a difference.

          Sadly, I can’t run this chip off a higher voltage from an external source, coz the power to the chip is also the current sense for the chip. A quick read around other suggested MOSFETs gives me the impression that most need more than 2.1 V to turn on, which you can’t reliably get from two cells, which clashes with what the datasheet says: “The MAX608 is ideal for two- and three-cell battery powered systems.”

          So yeah, really not understanding how this gets enough voltage to start switching.

          1. What frustrates me is that it looks to me like you’re doing their reference design — same fet they’re using, basically. Obviously it works for THEM, so that leaves you wondering.

          2. I know. Normally, I’ve had good luck with Maxim products. My usual approach is to build the reference design – I don’t know enough to buggerise around with the things, but I am good at following instructions.


            The simple versions of the boost circuit sort of hint the problem lies in the inductor. But it also suggests, reading the recommended inductor datasheet, the ones I have aren’t going to be useful.

            I am planning a bigish Farnell order very soon (at least $100) so I can easily chuck in a few inductors for you. But their footprint all looks like they’re most usefully flat wirewound things.

          4. Arse. Trying this with three cells, 4.1 V VIn doesn’t change the behaviour at all. The MOSTFET really should be conducting, it’s being fed 3.8 V.


          5. Inductance versus impedance

            Playing with an inductance/impedance calculator, that part looks to have way too little inductance for what you need.

            400 nH @ 10MHz (with trivial resistance) gives you the stated impedance of 25 ohms; 100nH @ 100MHz (with trivial resistance) gives you the stated impedance of 65 ohms. (The difference between these two suggests that part has non-trivial DC resistance, but it’s still a couple of orders of magnitude out from the suggested 22 uH. And the non-trivial DC resistance makes it contra-indicated by the MAX608 datasheet given the desire for a mostly inductance load.) I know your schematic lists 22uH inductor, but you might want to double check that the part you actually tried was both low resistance and the right order of magnitude inductance (use the calculator if the units you were quoted are impedance in ohms at a given frequency).

            That said, ‘s theory about the transistor choice also looks very plausible. Since in theory poor inductor choice (at least within the plausible range) should just make the circuit inefficient (either slow to start due to time to pump the inductor, or unstable due to too rapidly depleting the stored charge in the inductor — all these circuits effectively convert the DC to high frequency AC and use that to pump a higher voltage out, converted/smoothed back to DC; and it sounds like the pumping trigger is happening).


          6. Re: Inductance versus impedance

            Ah, that’s helpful, thanks.

            (I’ll be ordering a better inductor and a better FET. Won’t get to play with them till they arrive though, and have 300 kgs of tiles to carry up a ladder as well.)

  2. Photos

    The camera sensors in phones have a real advantage with close up photos, because the tiny sensor (a) is more easily filled with the image that you want to photograph and (b) the very small lens-sensor distance tends to mean a very high depth of field. Since “fill the sensor” and “keep everything in focus” are the main two things for macro shots, it’s a big win. (FWIW, your depth of field isn’t quite high enough in that first photo, unless you were going for artistic blur on the back half of the photo. But as you say, very impressive for a random “and we have a camera” feature on a phone.)


    1. Re: Photos

      The blur is entirely deliberate – I don’t want anyone able to read the markings on the diode at the back for fear of stealing my design and making millions of dollars (umm… off a design that doesn’t work).

      Normally, I’d take a shot like this square on, but this board has the world’s most excessive heat sink sticking up right in the middle

      1. Hmm… “Impedance:93ohm”? wut?

        If they mean 93 uH, then that’s a bit high, but wtf, I’ll give it a go, thanks. It’s not going to perform any worse than a straight piece of wire and it’ll give another data point to go on.

        1. Yeah that’s my thinking. They’re PTH and I have at least 8 or 9 left, so you can just grab a couple. 🙂

          Ping me uhh.. I can actually drive out somewhere and deliver them if you like, send me PM. 🙂 (Is excuse to go for a drive, since I am still n00b!)

        2. Impedence

          Impedance is to AC as resistance is to DC. Resistors, capacitors and inductors all have impedance. Because of the alternating in AC, there’s an extra factor which is the frequency of the AC — hence the specification giving multiple impedances at different frequencies. (The other approach is to give an impedance at a characteristic frequency. For completness, reactance is the non-resistance part of impedance.)

          Inductance is the thing measured by Henries (and more specifically in smaller sizes, uH). Impedance is measured in ohms (and often denoted “Z”). The relationship between inductance (H, uH) and impedance is more complicated than 1:1, and in particular depends on frequency: reactance (ohms) = 2 * pi * f (Hz) * L (H). (Amongst other things this means that inductors will pass DC, but have increasing impedance to higher frequencies; cf capacitors which won’t pass DC, but more easily pass higher frequencies.)

          Not sure if this really helps your problem. But I figured knowing the relationship between these units may help in future.


    1. Doh! I should have posted the schematic too. Here it is:

      (SHDN is controlled off the microcontroller, hence not drawn in this schematic.)

      I’ve tried with various resistive loads, from 220k to 15. Most testing has been at 900 Ohms, so a few milliAmps. The voltage decreases as the resistance goes down to 15 Ohms, but that’s just the batteries sagging.

      1. MAX608 shut down

        Comparing your circuit diagram with the example in the datasheet I see that the datasheet says ground SHDN for normal operation; you have left it floating, at least in your schematic. Given that SHDN is “shutdown input” (according to the datasheet), and is active high, allowing it to float is probably giving it enough bias from ground for the chip to shut down. So if you haven’t already in your implementation I’d suggest trying grounding SHDN and see if it springs into life. (From what I can see in your artistic photo, SHDN is floating in the implementation too.)

        As a general observation floating pins on an IC are a bad idea for stability (and heat, if not other reasons), and should always be grounded/pulled high (as appropriate for the pin) even if they’re not needed. This is especially true of input pins on a Quad-whatever chip where only some of the gates are used.


        1. Re: MAX608 shut down

          And then after actually posting the comment I see you say that SHDN is controlled from the microcontroller and hence presumably not really disconnected. Doh. Probably worth testing the isolated circuit, with that pin explicitly grounded, though.

          Also the other thing I was going to say is that 2 * NiMH is 2.4V max; 2 * Alkaline is 3.0-3.2V (when new), so if you’re concerned about input voltage then a quick test with a pair of alkalines would give you ample input voltage. The datasheet suggests it should work down to 1.8V though.


          1. Re: MAX608 shut down

            Yup, the SHDN line goes through a via to the back of the board, and then to one of the pins into the breadboard, and then to the microcontroller.

            Datasheet says it should work down to 1.8 V, but buggered if I know how. The MOSFET datasheet says the gate threshold may be as high as 4 V and the MAX608 datasheet says the chip should produce a gate drive voltage equal to VOut, which strongly implies that it should be able to raise the gate drive voltage above the battery input voltage, but umm…
            *scratches head*

            Will try with 3x NiMH tonight.

          2. Re: MAX608 shut down

            …and trying with 4.1 V coming in from three batteries reveals no change in behaviour. Still have a drive waveform on the FET gate, with an on Voltage of 3.8 Volts, so the FET should be switching. The drain is at VIn, with a trivial 0.1 V downards blip at the same frequency as the switching. Source is at GND, or as close as my DMM can get.

            I remain confussed.

          3. Re: MAX608 shut down

            Ok, I had a detailed email written and then loading the PDF datasheet for the FET crashed Firefox. Meh.

            I think that reading between the lines, if you use a MTP3055 FET, you can’t guarantee it to work with input voltage below 4.5v (Vth + 05v). There is no voltage step-up during bootstrap – it relies on the FET threshold being 0.5 below battery voltage. The alternate SI6426 device has a much lower Vth of 1.5v volts, so the circuit works worst case at 2v.

            Another thing to check (when you’ve applied 4.5v+ to it) is the resistance between CS and ground. Should be 50milliohms (how you measure that through contact resistance, I dunno, but basically it should show as zero).

            Given that the gate *is* being driven, the observed signal would indicate that one of those two has a problem.

            After that, I’d think about the inductor (serial resistance and saturation current are too high/low), but suspect that with the signals you’re seeing, it won’t work with any inductor value.

          4. Re: MAX608 shut down

            That’s all possibly true about the VIn, but the MAX608 datasheet specifically states that it is ideal for two cell-powered systems. It also has a minimum start-up voltage of 1.6 V. Now ok, that’s for the chip, not the FET, but you’d think they’d mention a glaring specification gap between the minimum voltage for the chip and a minimum for the FET?

            Anyway, you may well be right, I’ll get a FET with lower threshold to test the hypothesis. Farnell don’t have SI6426s, or their replacement, nor can you search on Vth, but I’ll have a hunt for something suitable tomorrow.

            (Current sensing resistor is within 3 mm of the CS pin, along a great big copper trace covered in solder. Other end is soldered straight to the ground plane. My DMM doesn’t go down to less than 0.1 Ohms and anyway, I’m just getting contact resistance, but I think the CS doesn’t have high resistance.)

          5. Re: MAX608 shut down

            FWIW, the MAX608 datasheet section on FET selection talks about choosing transistors with Vth less than 2V; but the MTP3055 FET has a minimum Vth of 2V (typical 3V; max 4V). They use a SI6426 in one of their examples because of its lower Vth (near 1V), after warning that if Vext getting too close to Vth will cause very inefficient operation.

            Also note that they’re switching the circuit at 300kHz (max) during normal operation, and up to 500 kHz during bootstrapping (same section of MAX608 datasheet). Which may help with choosing an inductor if the ones you’re looking at only quote impedances.


  3. Recap, now that it works

    Ok, after much scratching of heads, we came to the conclusion that the voltage threshold for the FET was too low. The VOut from the MAX608 wasn’t high enough to turn on the FET and get things switching. Solution – a FET with a lower threshold. I ordered an IR2703, chucked it in, and it works. Sort of.

    3.5 Volts in from 3 NiMH works and gives me the required 7 V output. Woo!

    However, still a couple of things baffling me. 2.1 Volts in from 2 flattish NiMH doesn’t. The FET should turn on at a minimum of 1 V, but the datasheet doesn’t specify what voltage will guarantee that it will turn on. Anyone know of a FET that will definitely be on by 1.5V or so? With 3.5 coming in, the waveform on VOut is a square wave up to 2.6 V, whereas the datasheet shows VOut switching at the output voltage. Thus it’s still doing things that I don’t quite understand, hopefully these things won’t come back and bite me at some point in the future.

    So that’s progress, but I was hoping to run this project of two cells, not three. Hmm… scratches head further, may be in danger of wearing through the skull and into brain.

    1. Re: Recap, now that it works

      *I* don’t know of any fets that start operating that low, and I don’t know that I’d trust them if they did because a low on-threshold usually involves a lot of compromises in Rds(on) and switching time. But I’ll email myself and look through our cabinets at work tomorrow and see if I can find anything good, and ask a couple apps engineers.

    2. Re: Recap, now that it works

      Well, I now have two apps engineers saying “… and HOW is this supposed to work?” so I’m going to try an ic design engineer tomorrow and see what he says.

      1. Re: Recap, now that it works

        Woohoo! My special power is asking smart people, and then finding questions that make them scratch their heads. Surprisingly, this special power pays quite well.

        Anyway, the start of this app note seems relevant:
        “IC1’s V+ terminal (pin 2) provides power as well as feedback to the chip. This “bootstrapped” operation (in which the chip is powered from its own output) enables start-up from input voltages as low as +1.8V, unless a heavy load prevents start-up altogether.

        Proper operation requires a gate-drive voltage sufficient to provide low on-resistance in the switching MOSFET, but at start-up this drive is limited to the battery voltage. The resulting high on-resistance in the MOSFET can prevent the converter output from rising to its specified level. On the other hand, connecting the output and load only after VOUT is within tolerance allows the MOSFET to turn on fully with minimum on-resistance.”

        (I’m knowledgeable enough to say this is relevant, but not knowledgeable to say what this means.)

Leave a Reply

Your email address will not be published. Required fields are marked *