Cable drop compensation — SY8286 + INA current feedback

Cable drop
compensation

SY8286A · INA current feedback · VFB = 0.6 V

A current-sense amplifier measures load current across a shunt and injects a proportional offset into the feedback divider, lifting the converter output by exactly the IR drop the cable will take away. The slope is fixed in hardware by R3, so it keeps climbing past the design point — the overload rows below show where that leads.

Circuit

Compensation network schematic The converter output V0 feeds a shunt, then the cable, then the load at V1. An instrumentation amplifier senses the shunt with IN plus on the load side, IN minus and REF on the converter side, and its output drives R3 into the feedback node where R1 from V0 and R2 to ground complete the divider. SY8286A Rsh 10 mΩ Rc cable 200 mΩ Load V0 V1 INA213 G = 50 IN+ IN− REF OUT V+ from SYS R1 55.6 kΩ R2 4.3 kΩ R3 76.8 kΩ C_f FB 0.6 V

Design equations

1 — upper arm at zero current
R13 = R2 × ( Vtarget / VFB − 1 )
R13 is R1 in parallel with R3. With REF tied to V0, the amplifier output equals V0 at zero current, so both upper branches sit at the same potential and act as a single arm.
2 — minimum shunt
Rshmin = Rc / ( G − 1 )
The raw signal G·Rsh·I is attenuated by R13/R3 < 1 on the way into the node, so it must start out larger than the drop it has to cancel. Below this limit R3 falls to R13 and R1 solves negative. Take two to three times the minimum.
3 — compensation branch
R3 = R13 × ( G · Rsh ) / ( Rc + Rsh )
A smaller R3 couples the amplifier more tightly to FB, giving it more authority over the output.
4 — local branch
R1 = R13 · R3 / ( R3 − R13 )
5 — converter output
V0(I) = VFB ( 1 + R13/R2 ) + I · ( R13 · G · Rsh / R3 )
Note what is missing: Rc, and any limit on I. Once R3 is soldered the slope is fixed and unbounded — the circuit measures current and applies an assumed cable resistance, with nothing to stop it climbing past the design point. This is feedforward, not feedback.
6 — voltage at the load
Vload(I) = V0(I) − I · ( Rcactual + Rsh )
The shunt sits in the current path, so its own drop is part of what gets compensated. Cancellation is exact only when Rcactual equals the design Rc; every 100 mΩ of mismatch costs 0.1 V per amp at the load.

Design current

5.0 A

The plot and table run to 1.5× this value so the overload region is visible. Nothing in the circuit clamps the compensation at the design point.

Parameters

Computed network

Topology

Actual cable resistance — three cases

Response

Voltage against load current, including overload The converter output rises linearly with current and keeps rising past the design point. Three load-side curves show the result for each actual cable resistance, against the 5.5 volt ceiling and the undervoltage thresholds.
V0 converter Load, case A Load, case B Load, case C Out of spec Undervoltage warning

Curves carry their cable resistance in milliohms at the right edge. All three leave the same intercept and fan out linearly, because the compensation slope is fixed in hardware while the real drop is not. The vertical marker is the design point; everything to its right is overload.

Table

Out of spec Below 4.63 V — undervoltage warning

5.5 V is the vSafe5V maximum from the USB Power Delivery specification, measured at the source receptacle. 4.63 V and 4.25 V are the approximate Raspberry Pi 5 undervoltage warning and PMIC shutdown points.

IN+ ties to V1 and IN− to V0 — reversed from ordinary high-side sensing. The amplifier output must fall as current rises, since V0 climbs by (R13/R3)(V0 − VINA). Wire it the conventional way and the compensation runs backwards.

Cff spans R1 only. It carries switching ripple from the real output node into FB, which the ripple-based modulator needs; the R3 branch carries the amplifier's DC output and has no ripple to give.

Standard resistor values will shift the result by tens of millivolts. Measure the built board.

打赏
  • © 2020-2026 X.
  • Powered by Hexo Theme Ayer
    • PV:
    • UV:

Buy me a cup of coffee~

支付宝
微信