- Mon Apr 18, 2016 12:23 am
#45669
Update: Had a bad batch of PCB due to a layout bug; Corrected that, and got a new batch; Delayed a couple of weeks doing another project; Finally got time today, assembled and tested.
Again, it is a mixture of success and failure.
First off, the circuit had to be "kick-started". My guess is that:
1. Reducing the directly attached output capacitor (to reduce energy wastage entering shutdown mode) causes low efficiency of the boost circuit at start-up -- i.e. the output voltages raises much slower
2. MC33464 kind of tie !RST pin with Vin before turn-on (below 0.5v)
3. TPCP8406's N-channel has very "good" response at low gate voltage
These three factors combined causes the shutdown signal to be asserted prematurely (i.e. when output reaches ~0.5v, instead of 3.3v)
At the very beginning I did not discover this weird behavior, and thus was misled to think I some how fried the TPS630252 (I got the YFF package, which is a *&#$^*#& pain to solder correctly). After tossing three pieces (~$4 per-piece, my heart was bleeding...), I finally start to think out-of-the-box...
I found as long as the secondary output capacitor has as low as 0.1v, the circuit starts up correctly. No kick-start was needed even if I remove input power with a minute or two. So even without fixes, it wouldn't pose too much of a problem.
After figuring this caveat out, everything else seems matching what I expected:
With no load, the power supply circuit starts up and boosts the output capacitor to 3.3v, then it quickly goes into shutdown mode; every 60~90 seconds, when the output capacitor voltage drops to 2.9v, the power supply restarts for about 50~100 ms, charging the output capacitor to 3.3v and then go back to shutdown. During shutdown all components together consume <10uA (limit of my meter).
But I noticed a problem when I attach a small load (LED, ~2mA) to the circuit: the output ripple is so large, I can visually see the flickering.
The problem roots in the design which attempts to enlarge the voltage hysteresis of the supervisory circuit via an RC circuit. By delaying the charge of voltage monitoring reservoir, my intention was to give enough time so that output capacitor can be charged to a higher voltage (3.3v) so that the circuit can enjoy a longer shutdown time. But when there is a good amount of load on the output, the RC circuit also causes delay for the supervisory circuit to realize the voltage drop.
I tried a quick patch by putting a diode across the timing resistor, so that the outward current flow is faster than inward. This helps, the restart cycle approximately reduced by 50%, but the flicker is still quite visible.
I am currently thought-experimenting a revision plan. The idea is to take output load into consideration -- only when the load is light, e.g. <1mA, the extra power saving makes sense; when the load is high just bypass the supervisory circuit and eliminate chances of large voltage ripple.
This means that I have to work with OpAmps, another piece I was trying to avoid touching, just like I tried to keep myself away from anything to do with inductors . I guess I have no choice now...
MCP6441 with 450nA power consumption, seems a good candidate, at least on paper...
Again, it is a mixture of success and failure.
First off, the circuit had to be "kick-started". My guess is that:
1. Reducing the directly attached output capacitor (to reduce energy wastage entering shutdown mode) causes low efficiency of the boost circuit at start-up -- i.e. the output voltages raises much slower
2. MC33464 kind of tie !RST pin with Vin before turn-on (below 0.5v)
3. TPCP8406's N-channel has very "good" response at low gate voltage
These three factors combined causes the shutdown signal to be asserted prematurely (i.e. when output reaches ~0.5v, instead of 3.3v)
At the very beginning I did not discover this weird behavior, and thus was misled to think I some how fried the TPS630252 (I got the YFF package, which is a *&#$^*#& pain to solder correctly). After tossing three pieces (~$4 per-piece, my heart was bleeding...), I finally start to think out-of-the-box...
I found as long as the secondary output capacitor has as low as 0.1v, the circuit starts up correctly. No kick-start was needed even if I remove input power with a minute or two. So even without fixes, it wouldn't pose too much of a problem.
After figuring this caveat out, everything else seems matching what I expected:
With no load, the power supply circuit starts up and boosts the output capacitor to 3.3v, then it quickly goes into shutdown mode; every 60~90 seconds, when the output capacitor voltage drops to 2.9v, the power supply restarts for about 50~100 ms, charging the output capacitor to 3.3v and then go back to shutdown. During shutdown all components together consume <10uA (limit of my meter).
But I noticed a problem when I attach a small load (LED, ~2mA) to the circuit: the output ripple is so large, I can visually see the flickering.
The problem roots in the design which attempts to enlarge the voltage hysteresis of the supervisory circuit via an RC circuit. By delaying the charge of voltage monitoring reservoir, my intention was to give enough time so that output capacitor can be charged to a higher voltage (3.3v) so that the circuit can enjoy a longer shutdown time. But when there is a good amount of load on the output, the RC circuit also causes delay for the supervisory circuit to realize the voltage drop.
I tried a quick patch by putting a diode across the timing resistor, so that the outward current flow is faster than inward. This helps, the restart cycle approximately reduced by 50%, but the flicker is still quite visible.
I am currently thought-experimenting a revision plan. The idea is to take output load into consideration -- only when the load is light, e.g. <1mA, the extra power saving makes sense; when the load is high just bypass the supervisory circuit and eliminate chances of large voltage ripple.
This means that I have to work with OpAmps, another piece I was trying to avoid touching, just like I tried to keep myself away from anything to do with inductors . I guess I have no choice now...
MCP6441 with 450nA power consumption, seems a good candidate, at least on paper...
