Other Parts Discussed in Thread: TPS61071
The TPS61072 variant of the TPS6107x family doesn't seem to be included in the more detailed documentation of the datasheet or app notes. I was asked to check out TPS61072 for conversion efficiency with ~1V input over a lower power range ~<100mW.
The first issue was inductor choice. The TPS61072 runs at half the conversion frequency of its counterparts, doubling ripple current.
From an efficiency point of view, a lower peak to average switch current ratio will generally work better. When average input current is not expected to go much over 100mA, the typical 400mAppk produced by recommended parts didn't seem to make much sense, nor did they produce very efficient results.
Which brings up the second issue larger inductance effects on chip behavior.
I could meet or exceed published efficiency behavior (for companion TPS61071) if the choke value was in the region of 33 to 120uH. However the larger choke values cause the output voltage regulation discontinuity (Fig 7 and 8 of the family data sheet) at much lower current/power levels. The voltage regulation discontinuity that is illustrated to occur above 100mA (>330mW) in the spec sheet figures is present in a 33uH circuit at 15mA (36mW).
There was also a stability issue - if you could call it that.
At loads near zero, with the 120uH choke (the most efficient, despite double turns and half copper x-section), the chip starts to, not so much oscillate, as go chaotic.
Large (400mA 100uSec) reverse current surges develop, at a rep rate dependent on average load. The circuit doesn't snap out of this condition until the (2V4) load is increased above a 40mA/100mW level.
The larger choke also shifts the DC output set-point voltage from the nominal 2V4 produced with the 33uH choke, to a voltage that's ~45mV higher.
This voltage regulation discontinuity isn't explained or justified in the app notes. How and under what load-time constraints is the DC regulation loop altered internally? It must make transient loading a really curious sight, depending both on load transient amplitude and duration. There is also no suggestion that choke values or ripple current levels will modify DC voltage set-points. Is this part of the same internal circuitry?
There is some mention of feedforward capacitance that may be necessary if choke values are increased, to improve stability and transient response. Is this the chaotic condition that is being referred to? Won't the shifting DC set-point act as a chaotic 'attractor'?
Rob Legg
legg@magma.ca