LM46002 - precision/accuracy of the adjusted switching frequency

Hi everyone!

We are designing a new power supply with LM46002. Our devices use 3 LM46002 power supplies on one 2-layer PCB (approx. 330mm X 330mm). The pcb is mounted on the aluminum heatsink. The power supply specification is following:

VIN = 22V - 52V

VOUT = 3V - 5.3V

IOUTmax = 2A

fsw = 420 kHz (Rt = 95k3 1%)

The problem is in the adjusted switching frequency Fadj. In the data sheet it is stated that precision of the Fadj is +-10% with 1% resistor. We produced 8 prototype samples of our device with LM46002 power supplies and measured switching frequencies. There is total 24 power supplies on 8 boards. The results are following:

It can be seen that 2 power supplies (of 24 total) are not inside specs from data sheet (+-10%). Measurements are made with Lecroy WaveSurfer 44MXs-B oscilloscope and 9.5pF probes. VIN = 52V Iout = 2A and Vout 3V. If we change VIN and Vout frequency stays stable! Even if we change the LM46002 position (desolder that IC and put it on a postion of another LM46002 - no impact of the surrounding components). If we remove RT resistor switching frequency is 500kHz! We even try diffrent values of RT - frequency linearly increases or decreases but not with expression given in data sheet . Please note that switching pin signal waveform regular (no ringing)!

All devices (including ones that have Fadj outside specs) passed full verification at -40°C and 80°C(only DC and AC value of output voltage), 25°C (DC and AC value of output voltage + waveforms of tests like transient load response, turn on/off under different load, switch pin and output ripple under different loads).

Our concern is that frequency will shift even more (on other ICs) and cause instability of LM46002 power supply (on other devices). Is it normal that Fadj is not in +-10% range?

Thank you in advance!

Schematic

PCB

-Top layer

-Bottom layer