Bootstrap fundamentals¶
Single-sentence summary. Bootstrap gate drive — the standard cheap way to drive a high-side N-channel MOSFET — works flawlessly in ground-referenced bridges and fails fundamentally in cascaded floating bridges. This project does not use bootstrap for the high-side gate; it uses optical isolation (TLP250) with an isolated 15 V rail per bridge (B0515S).
The classical bootstrap arrangement¶
A half-bridge with both MOSFETs being N-channel needs the high-side gate driven 10–15 V above the high-side source (VS) — not above ground. The bootstrap arrangement provides that voltage without an isolated supply:
flowchart LR
Vcc[V_CC = 15 V] --> BD[Boot diode]
BD --> BC((Boot cap))
BC --> VB[V_B
high-side
driver supply]
VB --> HSDrv[High-side
gate driver]
HSDrv --> HSGate[HS MOSFET
gate]
HSDrv --> Vs[V_S
connects to
HS source]
BC --> Vs
Vs -.->|when LS is ON| GND[(GND)]
When the low-side MOSFET is conducting: - VS is pulled to ≈ GND through the LS MOSFET. - The bootstrap diode is forward-biased and the bootstrap cap charges from VCC to ≈ VCC − Vdiode.
When the high-side MOSFET is turned on: - VS rises toward VDC (the bridge bus). - The bootstrap cap rides up with VS — the cap maintains VB ≈ VS + (VCC − Vdiode). - The high-side driver now has its 10–15 V supply referenced to the (now-floating) VS.
The bootstrap diode prevents the cap from discharging back into VCC when VS rises. Total parts count: one diode, one cap. Cheap, well-understood, and the reason the IR2110 has been the workhorse gate driver for ground-referenced bridges for 30+ years.
Where it breaks¶
The arrangement above has one hard requirement that's easy to miss: VS must return to ≈ GND every PWM period to recharge the bootstrap cap. Three independent things can violate that requirement:
| Violation | Mechanism | Failure |
|---|---|---|
| Duty cycle too high | The HS leg conducts > 95 % of the period; the LS-on window is too short for the bootstrap cap to recharge. | Bootstrap cap voltage sags over consecutive cycles; HS gate drive eventually drops below threshold; HS leg can't turn on; output collapses. |
| Bridge floats | In a CHB, the upper bridge's VS never returns to true ground — it stays at the cascade voltage minus its own bridge voltage. | Bootstrap diode reverse-biased the whole time; cap never charges; HS gate never sees ≥ Vth. |
| Bootstrap diode too slow | Cap charging current spikes hard at the LS turn-on edge; a slow recovery diode (e.g. 1N4007) can't refresh fast enough. | Cap charges to less than VCC; HS gate voltage marginal. |
The first failure mode (duty > 95 %) is handled in firmware. The third is handled by part choice (UF4007 fast recovery, 75 ns trr). The second is fundamental to CHB topology and is the reason this project does not use bootstrap for the high-side gates.
How the firmware respects the duty constraint¶
Core/Src/pwm_modulator.c clamps the high-side duty:
| Modulator | LOW clamp | HIGH clamp |
|---|---|---|
STAIR |
0.01 | 0.95 |
STAIR_ALT |
0.01 | 0.95 |
PSC |
0.05 | 0.95 |
The 95 % HIGH clamp guarantees a minimum 5 % LS-on window per cycle — enough for the 10 µF UF4007-fed bootstrap cap to refresh.
The firmware also runs a 6 ms bootstrap precharge sequence in the PRECHARGE state before transitioning to RUN: all low-sides are driven ON for 3 PWM periods at 500 Hz, so the bootstrap caps are fully primed before any HS switching happens. See Build Guide v4.0 §6 Gate drive.
Why this project uses isolated supply, not bootstrap¶
The bootstrap arrangement, even with the firmware clamps and the precharge sequence, would still fail for the upper bridge: its VS never returns to ground because the bridge is floating at the cascade voltage. The team simulated the IR2110 + bootstrap path in Simulink and confirmed the failure mode — see docs/design-notes/chb-isolation.md and the ELE 401 interim report §4.4.
The as-built design uses TLP250 optical isolation per gate with an isolated 15 V rail per bridge (B0515S 5V→15V isolated DC-DC). No bootstrap cap, no bootstrap diode, no duty constraint, no floating-bridge constraint. The TLP250 has 2.5 kV galvanic isolation and the same effective drive regardless of where VS sits.
This adds parts and cost (eight TLP250 + two B0515S vs. two IR2110), but eliminates the bootstrap failure mode entirely. For a CHB, the trade is not optional — it is the only correct choice.
What to remember¶
- Bootstrap is for ground-referenced bridges only.
- The duty constraint (LS-on > some minimum each cycle) is real and is enforced by the firmware's 95 % clamp.
- The bootstrap precharge sequence is what makes the 6 ms PRECHARGE state non-negotiable before any RUN command.
- If anyone proposes adding bootstrap drive to a cascade stage above the ground-referenced one — point them at this page.
Related: CHB isolation, IGBT vs. MOSFET, PSC vs. LSPWM, and firmware → modulators for the clamp values in code.