As part of my Smooth Power project, this feedback loop works independent of the microcontroller. In this section I build and test the feedback loop that mitigates the load on the battery by variably turning on a source power from the bus voltage in attempt to reduce battery discharge current to zero.
The concept "in real time" can mean a variety of things depending on the context. An electrochemical stack, ie a battery, presumably has a deficit in its ability to provide very short and very intense dischare pulses.
Plate and electrode geometry, surface area of electrochemical sites will have a direct impact on cell resistance, and resilience in the face of pulse discharges
Abstracted portion of the feedback under investigation
Using an opamp with extremely low quiescent draw because I'm running on supercapacitor energy, I want to detect a load via a shunt resistor measuring battery current, and provide power to offset that current.
The Optoisolator pulls this PMOS transistor's (Q4's) gate towards ground which turns it on and begins reducing/offsetting the current that turned it on to begin with - a classic negative-feedback scheme
To test, I connected a circuit that would draw current from the battery in a pulse fashion (upper trace) and observed the net battery current (lower trace)
At greatly-expanded time-scale, each pulse of the load can be seen to be causing the opamp amplifier circuit to call for several pulses of offsetting current
Unless you KNOW you want multiple pulses, this WILL cause problems
To understand the parameters better, I increased the length of time the load was drawing power
This divulges that the problem IS a sluggish response to a load, an overcorrection, a saturation, a re-overcorrection to the saturation....etc.
Placing a capacitor in parallel with a resistor increases the mathematical 'differentiation' component to the feedback. If the pulsations were caused by a delayed response, this should help. Turn on the power in response to 'a change in load'
Not Solved. Now the batt looks to be being CHARGED at every pulse of the load.
We want that lower trace to have only a glimpse of a load, but almost flat
While this waveform could be fine in the context of a gate-drive response to a pulsed load, we see from the previous slide it resulted in net battery charging - that could be dangerous if it overrode other battery voltage-limiting circuitry
With the response being so forceful and proactive, even affecting a net CHARGING in the face of pulsed discharging, it's likely an Integrating component is needed to smooth the initial hyperactive response
This looks pretty good. Batt discharge is a small fraction of total load. Let's look at the gate drive.
A DC offset? Is it really staying on over the whole period? It makes sense in terms of an integrator that's not getting reset, but it's harder to understand how this could result in the desired regulation.
If it results in the proper cancelling of the current, we wouldn't care THAT the gate drive has a DC offset, but we would want to go back and understand WHY it occurs
I widened the load interval and it flattened out between the 2 transients, so that seems good for absence of oscillation. Integrals above and below the reference line on the batt current (lower) trace looks good, although battery DC voltage should be measured. Area below can be slightly greater than area above, because that represents battery discharge. We DON'T want this area to be positive because that would charge the battery uncontrolled
Although there is still the unexplained DC offset, this looks perfect. A DC offset could be explained by the threshold voltage of the Optoisolator diode. The slight pulse downward suggests it's enforcing the prevention of net charge which we want. The fact that there's the single pulse per period, and no additional ringing or oscillation is a good sign this can move forward to higher order testing.
Smooth Power Group
Senior Design Project
Demo Setup