This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

SN74AUP1G125: Minimizing dynamic / shootthrough current

Part Number: SN74AUP1G125
Other Parts Discussed in Thread: TLV3691

I have an application where parasitic currents are crucial.

Static consumption of AUP is fine, but when buffering signals with an AUP2G14 I found that dynamic current is far too high for me.

I assume main contributor is shootthrough, not output capacitance. I assume that hysteresis circuitry does not consume current, but am not sure about that.

Now I intend to use two AUP1G125 as a totem-pole and enable one for low and one for high with "break before make" in between (like a half-bridge).

I wonder if the OE-input also has a totem-pole structure inside (i.e. is buffered) which causes shootthrough again.

  • Why do you want to use Schmitt-trigger inputs? Do you have slow edges, or signals that do not go to the rails? If yes, you will get shoot-through currents (see [FAQ] How does a slow or floating input affect a CMOS device?). The hysteresis of a Schmitt-trigger does not result in higher or lower currents.

    All logic inputs have shoot-through currents, and AUP already is the lowest-power family. The only way to avoid shoot-through currents is with a low-power comparator like the TLV3691.

  • Alfred,

    The shootthrough current will happen on schmitt trigger and non schmitt trigger devices. If you have slow inputs then you will get shoot through, but the AUP devices will be the best to limit the peak currents. 

    I agree with Clemens that if you want to get rid of these kinds of currents to use a comparator to tolerate the slow edge and set the switch point.

    Best,
    Michael

  • Indeed, I do not need a schmitt-Trigger input, I used it for an RC-oscillator and then used the same component for buffering, but slew rate is as fast as AUP gets.

    The output stage is essentially a full-bridge driving an isolated current sensor (i.e. powering, not signaling; believe it or not).

    I now am looking at the 74AUP2GU04 (only from Nexperia, it seems; U ought to have lower shootthrough) and the 74AUP1G74 (also gives me complementary outputs).

    Looking again at the 125 datasheet it seems that there must be a buffering stage at the OE-input, because not driving the signal to the rail results in a worse delta-Icc current than at the A-input.

    So assembling a "break-before-make" stage using them apparently is not an option.

    My dynamic current consumption right now is around 500µA and I need to get it down by a factor of 100. Oops.

    As far as I can tell AUP logic is still the best and maybe only choice because of the static leakage current.

    Something like the CD4007 would have been nice, allows crazy tricks.

    TLV3691 is amazing, but pretty slow, not sure if the speed is sufficient.

    I have no clear picture about dynamic power consumption, say at 30kHz, but the output stage is said to be "no shootthrough".

    Aside from that it derails my cost structure ...

    What I need is a no shootthrough complementary output stage.

  • Alfred, 

    So you are using the AUP device to power the sensor, is that the idea, and is the sensor consuming this current? Maybe im misunderstanding.

    When you say dynamic current consumption, do you mean the peak current during switching, or you mean the RMS current consumption?

    Are you switching at a high frequency? 

    If you have fast edges at the input, I wouldn't expect there to be much current consumption overall? Unless you are switching at high frequencies.

    Any input to the device will have shootthrough including the OE pin.

    Michael

  • At 30 kHz, shootthrough currents do not matter.

    How did you measure that 500 µA? Any CMOS device should be below 1 µA.

    If that 500 µA includes the current consumption of your sensor, then no amount of fiddling around with the driver/switch will be able to noticeable reduce it.

  • I am currently switching at 70 kHz.

    The 500µA are not all dynamic by the AUP2G14, but approximately 450µA can be traced to it (not anything around such as the sensor).

    This current is RMS, it increases non-linearly with supply voltage i.e. it is not capacitance discharge.

    Static leakage current is as expected.

    Then what is left? 

    There is a page where Nexperia characterizes the current consumption of AUP depending on input voltage.

    The peak values are so high that shootthrough is an absolutely plausible explanation for what I see.

  • Are both inputs driven with a valid logic voltage? Leaving an input open could explain this.

  • No, the complete design works precisely as intended, except for the dynamic power consumption.

  • Every µA matters, almost all CMOS are above 1 µA, check the datasheets (not least at higher temperature).

  • Then I'd have to conclude that there is some capacitance in the sensor.

  • Wrong conclusion, the current is even there when i disconnect the sensor. Been there.

  • Please show the schematic of your circuit, and a photograph of your board.

    Is it possible that there are leakage currents due to soldering errors or flux residue?

  • For the record: I now compared AUPU04 and AUP17 to AUP14.

    There is a measurable difference:

    AUP2GU04: 484µA

    AUP2G17: 527µA

    AUP2G14: 541µA

    So there is a plausible delta between components, but the major part still appears to be elsewhere.

    Forging on....

    In order to compare the measurements exactly I repeated them with an external pulse generator as driver (70kHz).

    And that got me eh pretty interesting results:

    AUP2GU04: 48.6µA

    AUP2G17: 51µA

    AUP2G14: 48.9µA

    AUP2G14: 46.2µA

    AUP2G14: 47.8µA

    AUP2G14 are 3 different boards.

    The conclusion now is that components do NOT matter and somehow my RC-oscillator sends current where it should not go.

    The current ist proportional to frequency with a static base of 31µA at 1 kHz, which mostly comes from an OPAMP which is also on this rail.

    That leaves me at ca. 18µA of dynamic current, which is still far above of what I want to have.

    So AUP probably must be replaced by something else. 

    Final note:

    6.6µA are going through my measurement shunt (the rest is elsewhere) and 60% of that is static consumption or some diode leakage. 

    So 2.5µA can be attributes to shootthrough and parasitic capacitance discharging (part of it may again be some clamp diode parasitic, replacement is intended). 

    I guess this is now within everyones estimates.

  • Hello Alfred,

    Glad you were able to debug this, but sorry if the AUP devices cannot address your low current needs.

    A few things to keep in mind for future reference:

    1. A slow ramp on the input of a CMOS device is going to cause the static current consumption to be higher (RC oscillator), as the device will be in the transition/shootthrough zone longer, so the peak current consumption that happens around 1/2*Vcc during an input transition will last longer and therefore have higher RMS current consumptionThis higher current happens on both schmitt trigger and non-schmitt trigger devices.

    2. Overall current consumption of a logic device can be calculated based on the following FAQ

    https://e2e.ti.com/support/logic/f/151/t/875721 

    Please let us know if we can help further here.

    Best,

    Michael

  • Yes, it was finally a measurement glitch that showed so much deviation.

    Though the analysis that it comes from a shootthrough was not wrong, but it was in the RC-oscillator, where actually it does not matter.

    In the end I'll keep the AUP-device at the essential buffer stage, because there is no apparent cost-effective transistor-level alternative (such as CD4007 was), and do some other small tricks to mitigate the impact of its own shootthrough and dynamic current consumption.

    The initial question was about the internal logic at the OE-input of the 125 tri-state buffer and I guess that my interpretation that there would be shootthrough too is correct, unfortunately.

    Case closed.