Hello all,
Have you ever encountered the following issue?
We have a board with two LTC3546 (QFN package) circuits. Both are powered by Vin = 5.20V (nominal). One provides 1.0V and 3.3V and the other one provides 1.2V and 1.8V. These circuits were designed for 1-2Watt output power per output. Forced continuous mode is used. The design was thoroughly tested for transient response at various loads and was found perfectly stable. Feedback resistors have 1% tolerance. Boards were assembled using lead-free reflow process.
On the section with 1.000V (nominal) output, we noticed the following discrepancies form the datasheet (data from 4 sample boards, Ta = 25oC, test load = 100mA approx):
- Output accuracy: At Vin = 5.2V, three of the four samples have Vo = 1.043V-1.045V (4.3-4.5% tolerance). One has Vo = 1.024V. Thus, on three samples, the tolerance is larger than the expected 3.5% (2% = 20mVdue to Vref tolerance at full temp. range plus 1.5% max. due to the combined worst-case resistor tolerance error). Actually, the expected error should be much smaller than 3.5%, since the tests were performed at room temp and the actual resistor tolerance error will be statistically much smaller. The feedback resistors are 10K, 15K so the bias current error is negligible.
- Line regulation: At Vin = 4.2V, all these outputs drop by 11-20mV (i.e. far more than the expected 0.2%/V = 2mV !!!)
- Temperature coefficient of output voltage: A rough temperature test shows a very strong (and unexpectedly fast) positive temperature dependence but a more formal environmental test is following.
We tried replacing one of the faulty (?) ICs but the situation didn't improve a lot (the respective 1.0V output is still out of spec on both accuracy and line regulation). Since, more or less, the 1.2V, 1.8V, 3.3V outputs behave as expected (wrt output voltage accuracy, line regulation, tempco), I wonder whether this device is not meant for Vo < 1.2V. Ok, this sounds silly but I run out of ideas.
Any thoughts?
