Dear Fellows,
My name is Guilherme Müller and I'm an Product Engineer at Altus Sistemas de Informática, a Brazilian LPC manufacturer.
A few weeks ago I've found out that the DCR010505U begins to resonate when connected with two other DCR010505U in parallel. This product is used on our redundant Profibus module, which is connected though a back-plane to CPUs, Power Sources and other peripherals.
Also, the problem stops when the fourth module is connected, but when I insert the sixth module, it starts again.
Knowing this is an intrinsic characteristic of the product, I've started to change some system parameters upon the load of the DC/DC converter. I've found out that the converter do not react well to a large capacitive load.
Here is the schematic of one of the DC/DC Converter
C22: 10uF 0805
C141: 470nF low ESR 0805
C10: 100nF low ESR 0805
R202: 10k 0402
TZ1: Tranzorb,5V,600W,BIDIRECIONAL,SMD
Connecting the sync pin did not result in any difference on the resonance.
Well, each Profibus Module consumes about 400mA from the power source. Connecting two modules results in 800mA load, but when the third module is connected, it reaches 2200mA and it is possible to hear a low volume frequency from the PCBs. Measuring the resonance, I've found about 10-15kHz frequencies.
Originaly our engineer used 10uF capacitors on C141 and C10, and my first step was to change them to the values specified above. I've changed C22 also to a 2.2uF low ESR capacitor, with no big difference.
But the biggest step was when I was trying to measure the current over the DC/DC: It simply stopped resonating.
Knowing that the amperimeter have a 10 ohms series shunt resistor, I've decided to add a 10 ohms series resistor right after C22 and the problem was solved, even with all the wrong capacitors.
So, my guess is that the DC/DC consumes a very large amount of current at the input even with good decoupling capacitors, and the series resistors kill those frequencies, representing no difference on the output voltage. Of course, the load after the isolation is about 30mA, which corresponds about 53mA at the input and a 0.5V dropout and this would imply the converter to work on it's limits.
Obviously, we will not use a series resistor on this DC/DC converter, considering the high dropout voltage, even knowing that it is inside it's limits. Changing the capacitors as mentioned above results in a very slow system, that gets stable at about 10-15 seconds, consuming 2200mA until it drops to 1200mA specified current.
Talking about systems, adding the resistor, the differential gain gets lower and the system does not oscillate. Reducing C22, which would decrease the differential gain does not corresponds to a big change on the problem. I did not tested with a inductor, for integral error compensation because I do not know the current frequency and amplitude the DC/DC oscillates.
We will keep the capacitors as specified above, because when connected to our power source, which has large capacitors, it does not resonate as before.
But maybe increasing the CI internal inductance would solve the problem for everyone (here is me putting my finger where I should not, hehe).
My kindly regards to all Texas Instruments fellows, and, if you read this post all over here, congratulations ;)
