• State Resolved
  • Locked Locked
  • Replies 6 replies
  • Subscribers 62 subscribers
  • Views 172 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

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.

TPS2553: Current limiting behavior differences between PCBs

Roderick Hinman
Prodigy 110 points
Part Number: TPS2553
Other Parts Discussed in Thread: TPS2065C, TPS25221, TPS2051C

We are using a TPS2553 to limit current output from a DC/DC converter with 5 V output (not a USB application, though).  The circuit is as shown below, set for a nominal limit of approximately 1.1 A with R103.  R106 is not populated, as it was intended to allow trimming the current limit if necessary.  

We are noticing differing behavior between boards in the batch of 75 that we built.  As load is increased at the output, some boards hold the output at a fairly constant voltage until they reach the overcurrent limit, while some have a much greater amount of droop.  Two examples are in the table below.  These boards were loaded with a set of resistors, starting with approximately 4.9 ohms connected continuously, and then adding more resistance in parallel.  We measured the output current and voltage with two DMMs, as well as viewing the voltage trace on an oscilloscope.  

The table below shows the results for two boards as load is increased (not necessarily exactly the same in each row).  The resistance values are calculated from Vout and Iout for reference, not from the parallel combinations of the various resistors used.

SN59 R103 = 23620 ohm *** SN53 R103 = 23750 ohm
Iout (A) Vout (V) Equiv.R Comment Iout (A) Vout (V) Equiv. R Comment
0.953 4.63 4.86 0.967 4.73 4.89 Note that this is already higher voltage than SN59.
0.961 4.60 4.79 0.989 4.72 4.77
0.980 4.55 4.64 1.003 4.71 4.70
0.999 4.43 4.43 1.008 4.71 4.67 0.34 V higher than SN59 at similar current.
1.006 4.37 4.34 1.062 4.69 4.42
1.010 4.28 4.24 1.099 4.64 4.22
1.012 4.21 4.16 1.103 4.56 4.13 Current limit about 1.1 A.
1.016 4.06 4.00 cycling Added 220 ohms more to above step, or ~20 mA additional if it weren't limiting/cycling.
0.978 3.76 3.84 Current drops from previous row?
0.987 3.58 3.63

I measured the value of R103, in circuit at least (I know there are potential pitfalls, but I'm not ready to disturb the boards yet).  The values are also in the table above, and based on the equations in the datasheet, I would expect the following range of current limits:

  • SN53: 1.010 - 1.170 A
  • SN59: 1.015 - 1.176 A, which is slightly higher than we would expect from SN53, but in fact it is quite a bit lower.

Any insight as to what could be happening here?  From what I've been able to see and also measure with the DMM when the circuit is off, differences are minimal.  Yet SN59 has a noticeably higher voltage drop and doesn't ever seem to go into shutdown.  Furthermore, its current limit is right at the edge of what one might expect from the resistor value...that seems a bit unlikely.  And as I increased load, the current limit decreased a little in the bottom two rows.

The shutdown behavior of SN53 is more common amongst the boards (though I tested them all with an electronic load for speed).  It is also more like what I would expect from the "-1" version of the part (and in retrospect, the better behavior for this application).  But this whole batch came off the same piece of cut tape, so it seems unlikely that there are mixed types here, and they likely even came from the same wafer.  

I welcome your thoughts.

Kind regards,

Rod

  • 0
    TI__Expert 4540 points

    Hi Rod,

    Thanks for the detailed explanation - your assumptions/calculations all appear correct.  Based on my understanding, SN59 is performing as you expect, but SN53 is not.  I'm not certain what may be happening but I have a few ideas/questions to debug the situation:

    • Are the results the same if you drive IN from a standalone DC power supply?  Have you verified VIN when running these tests?
    • How long is the path from ILIM, through R105+R103, and back to the IC GND?  We want this to be as short as possible.
    • Can you take a scope shot of V_IN, V_OUT, V_FLT, and I_OUT?

    Regards,
    Eric

  • 0 in reply to Eric Hartigan
    TI__Expert 4540 points

    Additionally, putting aside the exact current limit value, it looks like SN53 is entering constant current operation whereas SN59 is cycling in/out of thermal shutdown.  These are both normal behaviors of TPS2553.  The device will attempt to drive the MOSFET in the linear mode to maintain constant IOUT.  This requires an increase in VIN-VOUT dependent on the load, and this eventually leads to thermal shutdown.  If SN53 has a lower VIN than SN59, this may be reduce the voltage drop required to maintain constant current, thus allowing SN53 to stay out of thermal shutdown.

  • 0 in reply to Eric Hartigan
    Prodigy 110 points

    Thank you for the quick and helpful response.  The TL;DR is that I think it is partially one device going into thermal shutdown and the other current limiting, as you say, although I do not understand why they behave differently.

    To answer your questions:

    I have not yet driven the circuit from a benchtop power supply.  I have verified Vin, and it is shown in the scope shots below.  The average value is about the same in all cases and maintains its value near 5 V, though it appears that SN59 has more ripple when loaded at ~1 A. compared to SN53.  (That is something else for me to look into.)

    The path from ILIM, through R105+R103 and back to the IC GND is shown in the attached screenshot from the CAD.  The total length of the blue line, including in the ground plane, is 21 mm.  The via depth does not add much as the first two copper layers are separated by only 0.1 mm (39 mils) dielectric.   Admittedly, this isn’t the absolute shortest path possible; I could have put a via nearer the resistors to hit the ground plane closer. 

    Some scope shots are also included.  In all cases:

    1. Ch1 (yellow) is Vin at 2V/div, except for the first image which is at 1V/div and has a different position on the screen, to show the ripple better.
    2. Ch2 (cyan) is Vout at 2V/div, and it has the same reference point on the screen as Ch1.
    3. Ch3 (magenta) is Iout at 500 mA/div.
    4. Ch4 (green) is Vflt at 2V/div. Its reference value is at -4 divisions, which gets obscured by the information line showing channel scales.

    The plots show:

    SN59_RR_5_VinRipple shows SN59 with a real resistor load of about 3.7 ohms in its current limiting mode.

    SN59_RR_7_thermalShutdown is slightly more loaded at about 3.5 ohms and shows occasional thermal shutdown.  When the output is on, it is current limiting to just under 1.0 A.

    SN53_RR_4 and SN53_RR5 show the other PCB loaded by about 3.7 ohms at different time scales.  In each case, I’m turning on the input voltage with the load already attached.  When the output is on, it is current limiting to about 1.1 A.  What we see is a number of 8-10 msec pulses, and then the pulsing stops and the output stays constant at about 4.05 V. 

    SN53_RR_6 reduces the resistor to about 3.4 ohms.  In this case, there is a long string of 10 msec pulses, a few fitful starts at constant output, and then finally the thermal shutdown cycling shown in SN53_RR_8.  This is the behavior that I was curious about in my initial question, though before I did not wait long enough for it to shift from the brief low duty-cycle pulses to the long ones. 

    SN59 does not appear to do this (the overcurrent shutdown within 10 msec), at least not with a purely resistive load.  I have only seen it do the long cycles with fairly high duty cycle that look like thermal.  However, it will do the overcurrent pulses when loaded a programmable electronic load in constant current mode.  (I know, all bets are off for the dynamic response there; we have loads from two different manufacturers and in other situations we have observed different dynamics.)

  • 0 in reply to Roderick Hinman
    TI__Expert 4540 points

    Hi Rod,

    I don't have a complete answer for you, but can help explain some of these graphs.  In my initial assessment, I had forgotten you are connecting FLT and EN pins, so what I thought was thermal shutdown is actually the FLT pin asserting low, which disables the device until it recharges beyond the EN high threshold.

    • SN59_RR_7_thermalShutdown:  Device recognizes IOUT reaches IOS, FAULT asserts after FAULT deglitch = 5-10 ms.  Device is disabled because EN is logic low.  FAULT/EN recharges and device re-enables once a logic high is recognized.
      • I am uncertain why IOS isn't immediately recognized.  That is, I don't see a load current change prior to FAULT asserting low.
    • SN53_RR_4, SN53_RR5, and SN53_RR_6 are what I expect to observe.  At least until the pulsing stops.  The sequence is: EN charges to logic high -> VOUT goes high and immediately reaches current limit/IOS -> FAULT deglitch is 5-10 ms -> FAULT asserts, forcing EN logic low -> repeat.
    • SN53_RR_8 appears to be similar to the first plot, but on a wider scale.

    For SN59, have you tried an even lower resistance load?  I imagine there is a threshold where it will start behaving more like SN53.  I still recommend testing with a standalone supply when possible to eliminate variables from the setup.

    Regards,
    Eric

  • 0 in reply to Eric Hartigan
    TI__Expert 4540 points

    Hi Rod,

    I was able to speak with our design team and we have an idea of what may be happening.  The input voltage ripple may be causing the internal, pre-deglitch FAULT signal to toggle.  This keeps resetting the FAULT deglitch timer.  So although SN59 is current limiting, the internal FAULT signal never passes the deglitch filter, thus we never see it assert on the oscilloscope.

    We can test this theory by using an external supply with lower ripple.

    One solution is to increase the input bypass capacitance to reduce VIN ripple.  There are also parts with a different deglitch mechanism to address this issue.  TPS25221 is pin-to-pin compatible but would require a new R_ILIM (~50kOhm). It may also be worth checking TPS2051C and TPS2065C.  These have fixed current limits and are not pin-to-pin compatible.

    Regards,
    Eric

  • 0 in reply to Eric Hartigan
    Prodigy 110 points

    Hi Eric,

    Thanks for the additional information and suggestions for alternate parts.  To answer your questions,

    1. It is sort of possible to add more load to SN59 and get it to fault like SN53, but it is inconsistent.  See this scope plot of startup, where there is a mix of longer turn on and the 5-10 msec FAULT pulses.  About a minute later, it transitioned to the longer pulses again.  

    2. Powering the circuit directly from a bench supply does seem to fix the issue. The image below is the behavior with the bench supply when loaded with enough to trigger the fault, and it is consistent.

    Both of these results seem to support the designers' theory that ripple on Vin is affecting the FAULT determination.  This seems to resolve the issue, at least with this part.  I need to work on the DC/DC converter, apparently.  And I'll take a closer look at those alternates.  

    Thank you very much for the help.

    Rod