General > General Technical Chat
Why did my LT3042 die?
teddybear:
I bought the same module, CJMCU-3042, from a Chinese seller very recently.
I've tested the unit under a resistive load (~100 mA) with a spectrum analyzer equipped with a low noise preamp and found that the modules were unstable. There is a noise bump in the power spectral density at around 100 kHz, which is about 50-100 times higher than those shown in the datasheet. The instability is caused by the poor design of the PCB. The noise bump was disappeared after the following MODs to reduce inductance of the PCB traces.
0) First of all, add an appropriate elecrolytic cap. in Vin.
1) Additional X7R cap. from the Vin pad of Cin to the GND pad of Cout. Use this point as GND of Vin/Vout leads.
2) Add wires between the GND pads. The wire from Cset to Cout/Cin is especially important.
24 Sep. 2021
Additional notes:
I found that the Cset cap. in the module was not a low noise part, at frequency especially below 100 Hz. After changing the Cset to Murata's X7R cap. all of my modified modules for 5V, 12V and 12V with PNP boost works as expected from the datasheet.
Gerhard_dk4xp:
I have used a lot of LT3042 but I have never seen signs of instability.
And nearly all of them had to live with ceramic SMD capacitors.
The noise peak was always close to the frequency and size given in the
data sheet.
Be sure that the exposed pad is soldered, positively.
You cannot expect 100 mA from that tiny chip with a drop of several volts.
Even if it has switch-off for over temperature, the thermal cycles will kill it.
Since someone mentioned feeding an OCXO: mine draw 1.5A
during warm-up. A clear NO for the 3042.
Exactly for this reason I have made a board with LT3042 + TO220-
transistor that has to take the heat. It is exactly the circuit of the
data sheet. In the noise plot, 0dB is 1 nV/rt(Hz). The noise peak is
just 3 nV/rt(Hz).
The steep rise on the left side goes on the much undersized input
capacitor of the preamplifier. Note that is is much worse than 1/f.
Power supply is an R&S NGT20, all open on the table with no
shielding whatsoever. The board is home-etched, in the mean time
there is a tidied-up version to make use of the leftover space in
different JLCPCB projects.
cheers, Gerhard
radar_macgyver:
Damping is often required when one uses a series inductor for EMI filtering, here is a reference:
https://training.ti.com/power-tips-damping-input-filter
This is in the context of a switch-mode converter which has a negative impedance, which can cause oscillation in conjunction with an undamped filter. In the AN 88 case, the inductance comes from the long leads.
To estimate the optimum damping (when filter inductance is known), I use LTSpice, and step through different values of damping resistor while looking at the AC response, and pick the one with the smallest peak.
Navigation
[0] Message Index
[*] Previous page
Go to full version