• State TI Thinks Resolved
  • Locked Locked
  • Replies 5 replies
  • Answers 1 answer
  • Subscribers 63 subscribers
  • Views 337 views
  • Users 0 members are here
Support feedback
Options
Options
Related

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.

TPS7H4010-SEP: [TPS7H4010-SEP/LM73606] Inverting Buck-Boost (IBB) output voltage dropping

RoCo
Prodigy 10 points
Part Number: TPS7H4010-SEP
Other Parts Discussed in Thread: LM73606

Tool/software:

Hi TI,

I've been working on an inverting buck-boost converter utilising the TPS7H4010-SEP based on the application note SNVAA76 - Radiation Tolerant Inverting Buck-Boost Converter with
TPS7H4010-SEP. For my prototype boards I'm using the LM73606.

The configuration (seen below) is based on the 12V/-12V configuration from the application note, modifying it to operate as 15V/-15V.

Note: R5 is not populated, as well as the part list exclude parts.

Note that the UVLO from this design does not seem to work, I've rectified this issue already and R9 (the 10k) resistor is removed for now, pulling up the EN pin. 

The Issues
In general I'd say there are currently two issues:

1. The switching frequency does not seem to correspond to the RT values. It appears to be variable depending on the load, rather than consistent.

2. The output voltage "drops" (because it's negative it actually increases, but you get the point) for increasing load.

Output Voltage vs Output Current

The first verification step was to simply plot output voltage for different load scenario's. I'm loading the circuit with a programmable load in constant resistance mode.

In experimentation, trying to figure out what's going on with the switching frequency I've noticed that if RT is left unconnected, the device appears to operate better, however there doesn't seem to be any significant change in actual switching frequency observed on the SW node.

Switching Node Measurements

In order to assess the issue I've performed measurements of the switching node SW (pin 1 of L1). It should be noted, that for these measurements the output voltage itself is actually quite stable (except the last one). Just not at the right value. I've looked at the plots of the 12V/-12V design in the application note and the behavior of this IBB design seems to not be the same. Debugging has proven difficult since a lot of the statements from the regular datasheet do not appear to be valid for this IBB configuration.

No Load - Output = 15.2 V.
See two measurements below of the same no load condition. The timing between pulses appears to fluctuate. I believe this makes sense since the device is operating in the auto mode, only switching when required. Note that purely looking at the square wave part of the switch, it appears that the "Switching frequency" is ~ 2Mhz here, whilst based on the RT setting (39k) it should be 1 Mhz.

1k Load - Output = 15.17 V
I take this as my baseline, this is a "light load" and behaviour seems to be what I'd expect from the auto mode.

50R Load - Output = 15.01 V
With a 50R load, the output voltage has already dropped to 15.01 V. The current here is ~ 300 mA. Looking at the switching node below I notice that even-though the output voltage has already dropped by 200 mV, the device still appears to idle for about 50% of the time. (my assumption is that that's what happens when those sine oscillations appear, perhaps their presence is my whole issue, but I haven't been able to find the reason for their existence) Furthermore, the switching frequency (looking purely at the square wave) appears to be ~ 1 Mhz here.

10R Load - Output = 14.4 V

Current here is ~ 1.4A. Here approximately 600 mV is dropped and the switching frequency appears to be 500 kHz, with still some "off time". Another observed effect is the short peak in the SW node whilst the node is at it's lowest point. Finally, there still seems to be some "non switching" time, eventhough the output voltage is nowhere near reached.

5R Load - Output = 9.22 V
This is where the device seems to completely lose it. This output voltage is no longer stable and the SW node seems to be a bit all over the place. I'm still intruiged by those "peaks" I noted earlier, my first thought is that there might be some internal overcurrent protection kicking in, but the hysteresis of that is in the order of 46 ms, which is far too long.

Additional Information
Some additional invo that may help:

1. The Inductor has a saturation current of 5.6A, which I believe may be part of the issue and I'll be trying to find a higher rated inductor.
2. I've tried to connect the RT resistor to GND instead of the output voltage, as well as using a potentiometer in it's place. With the RT to GND, the switching frequency seems completely unaffected, regardless of value. With the RT to the output voltage (N15V) at least it appears to do "something", but not what one would expect.

  • 0
    TI__Expert 6891 points

    Hello Rogier,

    I believe what you are seeing is quite severe instability. After the output current rises enough to move the device out of PFM mode, while it may look like the switching frequency is changing with load, it is actually ~1MHz the whole time.

    As seen in the above image, each red arrow points to a new switch cycle. The SW node goes high for almost two whole cycles, then spends the next three cycles at min-on-time (essentially 0% duty cycle for this purpose) and it can be seen that the inductor current goes all the way to zero before the switch is high for more than the min-on-time again.

    Inductor current reaching zero is what those "sine oscillations" are that appear in many of the scope shots. When the inductor current becomes low enough that it can no longer produce a voltage greater in magnitude than VOUT, it can no longer provide energy to the output and so begins oscillating with the LC formed by the inductor and COUT. This does not happen in FPWM mode, as the device will simply leave the low-side switch closed and allow negative current to flow through the inductor instead of letting it settle at zero.

    The TPS7H4010-SEP does not have external compensation of the control loop, so compensating the device is quite limited. In general, adding COUT will decrease the control loop bandwidth, and reducing COUT will increase it. The feed-forward capacitor can also be adjusted to change the frequency response of the converter, but it may not be required at all with a top feedback resistor less than or equal to 100kΩ.

    The datasheet recommends selecting COUT so that the crossover frequency is between 1/10 to 1/8 of the switching frequency, and also provides equation (18) on page 24 to estimate control loop crossover frequency based on VOUT and COUT. Equations (19) (20) and (21) are also provided for guidance on Cff selection if needed.

    Like you had mentioned in your post, the inductor may also be coming into play here. The high-side FET is staying on for ~1.75µs, which with 15V across a 3.3µH inductor will put the peak current at ~8A. This is far higher than your inductor saturation and is also well within the range of the high-side switch current-limit too, which could mean that the switch is only turning off when it is, because the device is hitting it's current limit.

    I recommend taking a look at table 8-1 on page 29 of the device datasheet. It provides component selection guidelines that takes into account the control loop for various conditions. For example, the 12V 1MHz condition is fairly close to your use-case, and recommends a 6.8µH inductor to keep the inductor ripple between ~10% and 30%, and only 22uF of output capacitance to ensure the control loop bandwidth is around 100kHz to 125kHz. 

    Thanks,

    Andy

  • 0 in reply to Andy Fondaw
    Prodigy 10 points

    Dear Andy,

    Thanks you so much for the supply, I have some follow-up questions and remarks that I hope will aid in tracing down the issue.

    One immediate item of importance, we discovered that the capacitors placed for the input and output capacitors were 4.7 uF instead of the 47 uF stated in the schematic above. This means that my real output capacitance is actually closer to the 22 uF you suggest. I'm not sure if this means your initial hypothesis of my Cout being too high is now invalid?

    1. I just want to verify you're aware I'm using this converter in "not" it's regular mode of operation, but instead as an inverting buck-boost converter. I've found through SNVA856B and SNVAA76 that this has further implications.

    2. The design I've started with is based on the design presented in Table 2-1 of SNVAA76. Here it states that for a -12V output, a 3.3 uH inductor in combination with 141 uF of output capacitance is used. This is what I based the design that you see above on. But if I'm understanding you correctly, 141 uF would be far too much for this device to work? Looking at Figure 3-4 it seems to operate well. 

    Combining the fact that my measurements were (erroneously) performed with ~14,2 uF Cout and the fact that in this TI aplication note it seems to work correctly with 141 uF. Do you still stand by the conclusion that reducing Cout would be required? Would it need increasing?

    Future Steps

    My first steps are to replace the inductor with one that has a current rating that is sufficient. I'll make sure to get both a 6.8 uH and 3.3 uH value to experiment. But I'm unsure whether I should increase or decrease the output capacitance here.

    Furthermore I'm going to measure the input to see if my input capacitance may be insufficient currently.

  • 0 in reply to RoCo
    TI__Expert 6891 points

    Hello Rogier,

    Yes, the 47uF cap being 4.7uF is a big change. Do you actually have all 5 caps mounted to the board? If so, it would indeed be very close to 22uF.

    I am aware that you are using this device as an IBB, and although that does have implications, many things can also be generally applied to an IBB config from a buck config. Therefore I would still recommend using table 8-1 in the device datasheet to guide component selection.

    Since it turns out that you have been running the device with a fairly quick control loop (low COUT), and the application note does show functionality with a slower loop (greater COUT) I would recommend trying to add additional output capacitance to slow the loop. Decreasing the control loop bandwidth can increase gain and phase margin, and give greater stability. I would also recommend either re-calculating your feed-forward capacitor for your conditions if you have not already done so. 

    For inductor selection, please also keep in mind that since this is an IBB configuration, the inductor current ripple will be larger than in a buck. Choosing a larger inductor will help to keep current ripple lower. You had already mentioned SNVA856B, and this application note contains the equations needed to calculate inductor currents in an IBB.

    If your input caps are also all 4.7uF, then I agree that measuring VIN would be good. Depending on the parasitics between this converter, and whatever is powering it, and considering the increased current stresses of an IBB, some extra bulk input capacitance may be required. It may also be a good idea to place some amount of input capacitance not only between VIN and GND, but also between VIN and VOUT, as these are two separate current loops that both benefit from decoupling.

    Thanks,
    Andy 

  • 0 in reply to Andy Fondaw
    Prodigy 10 points

    Dear Andy,

    I'm a lot of debugging and board changes further, but unfortunately it appears that ~1.8-1.9A is still some magical limit where the device becomes unstable. To not be a "trim pot wizard" I've decided to go back to the drawing board and see if there was some critical error in my design. Looking at SNVA856B "Working with Inverting Buck-Boost Converters" once more and using the equations there combined with the equations from the datasheet I'm not entirely sure what would be an appropriate design here.

    Using the following inputs:

    Inputs
    Vin 15 V
    Vout -15 V
    Vout,maxripple 0.1 V
    Current Draw 3 A
    Fsw 1.00E+06 Hz
      1 MHz
    Inductor Ripple 0.3 -
    Inductor Ripple Current 0.9 A
    Duty Cycle 0.5 -
    Efficiency 0.85 -


    Inductor
    Eq. 8 from SNVA856B yields a 9 uH inductor. I've currently selected an 8.2 uH one.

    Output Capacitance
    Using Equation 28 from the datasheet, a value of 66 uF is the minimum output capacitance. The board I currently have has approx. 90 uF placed.

    Feedback Resistance
    I'm still assuming the 100k/7.2k feedback network.

    Compensation Cff

    And I believe after this point is where the design becomes difficult. The datasheet, through eq. 18 specifies that the crossover frequency fx which is dependant on Vout and Cout alone, must be "1/10th to 1/9th" of fsw. Using the Cout and Vout the crossover frequency is currently 17.8 kHz, not the ~100 kHz that's recommended. However the two parameters used here are kind of "fixed". Cout has a minimum value and if I'd reduce it too much the output ripple becomes unmanageable.

    Going with the calculated value of fx (17.8 kHz) this means the compensation capacitor should be 344 pF, which seems rather large. 

    In order to rule out issues with Cff, I've reduced the resistive divider to 18k/1k3 instead, which should (per page 22 of the datasheet) means Cff "may not be required". However even with that configuration I'm not getting much more than 1.8-1.9 A out of the supply. The efficiency also seems rather unaffected by most of the changes I've made.

  • 0 in reply to RoCo
    TI__Expert 6891 points

    Hello Rogier,

    Looking at your design parameters, I believe your component selection looks reasonable. Having a slower control loop due to needing additional output capacitance should be acceptable in this design too, although that is almost certainly the primary reason why the calculation for the feed-forward capacitor is returning such a large value.

    While I do not think it is the cause of the instability you are seeing, reaching 3A may require a larger inductor, as with a 3A current draw and a 8.2µH inductor, the peak switch current is above the Min high-side current limit threshold (using equation 6 in snva856b, and assuming 75% efficiency). Using a larger inductor would be getting close to 10% current ripple (10% of 6A) which is the minimum ripple recommended for this device in the datasheet.

    With behavior similar to what you are observing, the typical recommendations are to adjust the COMP network, use a smaller inductor for larger current ripple, or increase slope compensation. None of those options are available to you however because the TPS7H4010-SEP is internally compensated, the inductor is already too small for 3A of IOUT, and the TPS7H4010-SEP does not have an externally adjustable slope compensation.

    Regarding your efficiency target and comment, I would expect the device to be approximately 75-80% efficient. IBBs will have a lower efficiency than their buck counterparts, by nature of larger current stresses through the device producing greater losses.

    As for potential next things to try:

    Although you already have a decent amount, adding additional decoupling capacitance to PVIN can sometimes reduce internal noise and help clean-up falling edge instability.

    Running at 500kHz may improve stability, although if very low ripple is needed then this may require additional COUT.

    Using a slightly different input voltage may help by allowing the device to run at a different duty cycle. Although if you try this, please be aware that 17V on VIN will give the device's maximum voltage stress of 32V across the device.

    Thanks,

    Andy