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.

[FAQ] OPT8241: DC-DC converter choice for designing 3D ToF cameras

Part Number: OPT8241
Other Parts Discussed in Thread: TPS92515, LM3429, LM3409

What are some guidelines for choosing DC-DC converters for designing 3D ToF cameras?

  • MIX:

    This supply needs to have a very good transient response. The camera can operate in modes such that it has non-periodic quads. In such scenarios, the first quad and the last quad can have very different mix currents. Such mismatch in the sensor power supply across quads can cause the camera to have color error. If 2 objects at a given depth have different reflectivity, a camera prone to color error will report different depths for the 2 objects.

    The aim is to limit the average voltage mismatch across quads to within 12 bit precision which is approximately 0.025% . That is the extent to which the ADC can resolve. So the power converter specs obviously depend on the actual integration time per quad. For this calculation, one must add up the effect of all instances of a particular quad for a given frame. If Quad 0 occurs 16 times in a frame, we average across all the 16 Quad 0,1,2,3 individually. The other thing to take note of is that the first few microseconds within a given quad are used in resetting the sensor. It is ok to have more offset in the mix connection during this time. Duration of this period is 5us. So it is good to ensure that the effect of the DCDC transient response is limited to the first 5us. This greatly simplifies the spec for the rest of the converter design. If the voltage ripple is ±5%, the ripple time period 500ns, and the integration time 100.25us, the variation will be 0.05*250ns/100us = 0.0125%

    This is well within the limits and a safe operating zone. System level variations like capacitor derating, inductor saturation/derating, ageing, high temp degradation of inductor/capacitor, ESR can cause the actual ripple to exceed the designed limits. The ripple spec is achievable fairly easily and gives ample margin for any variations. If a converter of slower switching speed is used, the voltage ripple percentage variation should be reduced such that switching time period * voltage variation percentage = 2.5e-5

    TI has a series of converters called SWIFT converters which are known for their very good transient response characteristics. They are capable of operating with voltage ripple values ~75mV (5% of 1.5V) which works well for us. The older control schemes among the swift converters have a problem with DC load/line regulation. DCAP-3 control scheme alleviates this to some extent.

    The range of voltages used can be in between 1.4V to 2.0V. We typically use 1.5V in our designs.

    Illum:

    The illumination power supply is more involved as it is required to take a voltage input and give a constant current output. The illumination circuit draws bursts of power. During integration(exposure) time, the illumination circuit can potentially draw power in the order of 10s of Watts. During readout or dead time of the camera, the illumination circuit draws close to 0 power if leakage is not accounted for.

    It makes sense to use an IC which can be enabled or disabled within a switching cycle. A converter with such a feature can be enabled by the ILLUM_EN signal. It will not waste power when the illumination driver won't emit light.

    There are 3 such ICs which match all the expected features: LM3409, TPS92515, and LM3429. LM3409 and TPS92515 are buck converters. LM3429 is a boost converter, the other 2 are buck converters.

    These 3 converters are the recommended means of powering the illumination driver circuit.

    Other supplies:

    Among the remaining supplies, AVDD 3V3, AVDD 1V8 and PVDD are important. Any ripples in these supplies can potentially get introduced as noise into the signal chain. The recommendation is to use LDOs to supply these rails. The LDOs should have PSRR exceeding 40dB at frequencies up to the expected ripple frequency in the LDO's power supply.

    The PLL supplies should be a steady and not be affected by transients cause by varying loads in the system. For instance, if the 1.8V supply is used for MIX and for AVDD_PLL, the PLL could potentially go out of lock because of the power supply transient caused by the step in the MIX current.