LM3102 Output Regulation Issues

Hello,

What you are describing is common for output voltages above 5V. The control scheme used by the LM3102 is that of constant ON-time. The feedback loop consists of a comparator and reference and not the usual op-amp and reference. Your component values all appear to be correct. The issue occurs at light load because in a single switching cycle more energy is delivered than is being consumed by the load. 

For your application requiring wide load range and 10V output we suggest LM25005. Consider using www.ti.com/webench as a selector tool for other device options.

Thanks,

Anston

 Anston,

thanks for the feedback. The light load part that is troubling me is that switching still occurs when FB>0.8. The datasheet states a ON cycle is only started when FB<0.8. Once a burst of switching starts it only stops when the 'shut down ' FB=0.92V is reached. This makes me wonder if the  FB is experiencing a small glitch that pulls it below 0.8 is causing. Like maybe when the bottom side MOSFET is turned off. This then may explain when the behaviour changes with load as with light load the MOSFET will turn off in less time due to zero inductor current.

On your comments on the LM25005 - I has seen this part but didn't like the the part as it lacks the sychronous rectifier.

David

Hi Anston,

I am seeing something similar to DEP on an LM3102 with an output voltage above 5V.  I have two follow up questions:

1)  To confirm, were you also referring to the behavior of Vout dropping below the set point at light-medium loads and rising back up at heavier loads when saying that this is common behavior at Vout above 5V?  I'm observing the average output voltage falling further than approximately 1% below the set point before it rises back up.

2)  At light loads, I see a large sawtooth wave with a slightly jagged rising slope (due to the switching), which is what I think DEP is referring to.  However, I've observed it begin its long descent to FB = .8 well before it gets to FB=.92.  Would that descent start when the circuit finally reaches DCM (I'm assuming the "zero coil current detect" portion of the block diagram detects DCM)?  DCM is the only other thing besides thermal, overvoltage protection, and undervoltage lockouts that I can find in the datasheet that would cause the switch to turn off for an extended period of time.

Thanks

Hi John,

1)  This part creates an internal ripple signal which is added to the FB ripple and compared against the 0.8V reference. The added ripple allows the part to be stable with ceramic capacitors. Above 5V the DC value of this added ripple becomes higher and effectively shifts the FB voltage seen outside of the part at DCM.  This is the primary cause of the DC shift at light loads when the part enters DCM.

The version of the part in the 28–ball DSBGA, has fixed this issue by adjusting the reference based upon the internal ripple value.

2) If you are above 5V and you have are using ceramic capacitors, it might be that the out of phase capacitor ripple is large compared to the internal ripple.  This might cause the part to have what looks like groups of switching pulses as the internal ripple and external ripple fight for control of the system dynamics.  You might need to add a DCR network to avoid this behavior.  The DCR network adds ripple in phase with the switching action.

Regards,

Marc

Thanks, Marc.

Two quick follow-ups

1)  What is a DCR network?  I'm unfamiliar with that acronym in this context.

2)  Any plans to implemented the fix you mentioned in the TSSOP package?

Thanks,

John

Hi,

I just found app note AN-1481 discusses this DCR concept and provides calculation examples.

http://www.ti.com/lit/an/snva166a/snva166a.pdf

 

Thanks, Marc.  Another thing about LM3102 that I was reminded of recently.  Previously, I was told there was a 7V limit on Vout, but I haven't found anything in the datasheet that would support that.  Is there an absolute max and/or max operating condition for Vout?

Thanks

John,

The LM3102 can work with a max output of 7V and the min output will be dictated by the internal reference voltage, which in this case, is 0.8V.

Regards,
Akshay

Thanks, Akshay.  The part doesn't get damaged if Vout is above 7V, correct?  For example, if you set the nominal output voltage to 7V, it will likely go somewhere above 7V at light loads, and the part would not be damaged.  I assume it's just that the operation would not be guaranteed.

Thanks.

Yes, It's quite likely that.

Regards,
Akshay

Hi Akshay,

Thanks for the info. Pardon the delayed follow up, but I was wondering if this 7V is an absolute maximum (resulting in damage to the part), or if it's an operating condition (where proper functionality is simply not guaranteed)? I'm guessing it's an operating condition and wouldn't damage the part because more than 7V on the output wouldn't seem to present an overvoltage at any IC pin, but I may be wrong.

For example, what if you design the .8V reference to put your output somewhere below 7V, but for some reason the output floats up to correspond with the .92V overvoltage reference (such as the burst that David discusses earlier in the thread), which would correspond to a voltage above 7V?

Thanks.

Hello,
It is basically that for outputs above 7V, the operation cannot be guaranteed. The internal parameters are designed for optimal performance for outputs below 7V.

Regards,
Akshay