I've created a multiple rail supply using LM2596Ts. I used a combination of the datasheet and internet searching to check viability and it seems that this should be possible.


The max draw of the 5V rail is 700mA and the +/-12V rails is 900mA.
I am finding that there is a noticeable whistle coming from the 5V area of the board, presumably the inductor. It's around 5.2kHz and with lots of harmonics.
The switching frequency should be 150kHz so miles from 5kHz. Max load is 3A so miles from 700mA.
I haven't been able to achieve the exact layout recommendations of the datasheet due to essential locations of other components (see the massive switch in the way!), but the components are close and use thick traces and planes.


The whistle is physically audible on the board and also audible from the audio output (the board also has audio path traces on it).
The component values were chosen via the various tables and graphs in the datasheet.
I tried switching the 68uH inductor for 100uH out of curiosity, but it still sounded the same to me.
Interestingly when using the the TI designer for the LM2596, it produces radically different results to the datasheet calculations.


I measured the nodes using Audacity and a picoscope. Interestingly there is some difference in the results, although this may be to do with the FFT settings.




I would appreciate any help or advice with diagnosing the problem.
many thanks.

 

I suspect that it is instability causing this. Can you put a scope on the 'SW' pin?

 

I suspect that it is instability causing this. Can you put a scope on the 'SW' pin?

Thanks for the info. Here's the scope on the SW pin. On x1 the input is outside max range, so had to use x10 probe.







And for comparison, here's the +12V





and -12V



Annoyingly the jpg export lost my timings, but they were all approx 6.5uS = 150kHz, so as expected.

I've also searched the datasheet and seen that the ESR value of Cout can cause instability. There's info on page 17 and 25. I'm using this cap with ESR of 340mOhms. I'm actually using a 25V cap as suggested, but I wonder if I should experiment with different ESR values/voltages.

 

Put the scope on a slower timebase, perhaps 1ms/div, so you can see any jitter of the waveform at around 5kHz.

 

At 1ms it just looks like a solid block on a digital scope. So I tried using persistence mode to check for jitter. 5V looks pretty stable:



as does +12V


But there's a lot of jitter on -12V



I also took a freq analysis, although not sure if I'm looking at subharmonics of the 150kHz or the 5kHz jitter!



So it looks like it is the -12V that is the culprit, not +5V. That has a 100uH inductor and 68uF cap (25V, think ESR is about 380mohms again).

 

my commercial APC 850 UPS also whistles ?? poor inductor build ???

 

So... still on this problem! I discovered page 14 of the datasheet which states:
"Because of differences in the operation of the inverting regulator, the standard design procedure is not used to select the inductor value. In the majority of designs, a 33-μH, 3.5-A inductor is the best choice. Capacitor selection can also be narrowed down to just a few values. "
So I switched the inductor to this spec (33uH) and the output cap to 220u (one of the options in fig 18 in datasheet).
The noise is much less, but is still there and probably still at an unacceptable level.
Where can I go from here? Can I lower the inductor value / increase cap value further? Would that help? i.e. am I trying to find the optimum combination of values for the inductor or cap?
I'm still a little confused how this PWM on the waveform relates to the 150khz switching frequency or why it's there.