Troubleshooting the BAT54HT1: Symptoms, Causes, and Fixes for onsemi's SOD-323 Schottky

The symptom that brings most engineers to this part's datasheet is the same one: a high-impedance node protected by a Schottky clamp sits perfectly at room temperature, then drifts by hundreds of millivolts — or volts — once the enclosure warms up. The diode hasn't failed. It's behaving exactly as the BAT54HT1's leakage-versus-temperature curve says it will. That gap between the room-temperature electrical table and hot-box reality is the defining gotcha of this part, and it's where this post starts.

First, the part itself: the BAT54HT1 is onsemi's BAT54-class Schottky barrier diode in a miniature SOD-323 surface-mount package — a single two-terminal diode, not the three-lead SOT-23 of the classic BAT54. It is specified for high-speed switching, circuit protection, and voltage clamping, with a 5.0 ns reverse recovery time, a 30 V reverse-voltage rating, and an extremely low forward voltage (0.35 V typ at 10 mA) that the datasheet credits with reducing conduction loss. Its small package makes it explicitly suited to handheld and portable hardware where board space is limited. It is a small-signal and protection device: 200 mA DC forward current max, 200 mW dissipation on a minimum-pad FR-4 board at 25 °C. This post walks through diagnosing the failures people actually see with the BAT54HT1, how to integrate it correctly in the first place, how to verify it on the bench, and how to fix it when it misbehaves.

Symptom 1: The protected node drifts when the board gets hot (Why does my BAT54HT1 clamp leak at temperature?)

This is the big one, so it gets the most detail. The electrical table says reverse leakage is 2.0 µA max (0.5 µA typ) at V_R = 25 V — and that figure is a 25 °C number. The datasheet's own leakage curves (Figure 3, plotted at 25 °C, 85 °C, 125 °C, and 150 °C) tell a very different story at V_R = 30 V:

That is roughly a 100× increase in leakage at load temperature. If the diode shunts a high-impedance node, that leakage flows through the node's bias resistance and produces a voltage error of I_leak × R. At room temperature with the 2 µA worst case: 10 kΩ gives ~20 mV of error, ~200 kΩ gives ~400 mV, and 1 MΩ gives ~2 V. At 125 °C with ~150 µA, even a 10 kΩ bias resistance costs you 1.5 V.

The fix: keep the effective bias impedance on a clamped high-Z node well under ~100 kΩ, or accept and budget the hot-state leakage explicitly. If your offset only appears when the board warms up and disappears when it cools, this is not a defective part — it's the leakage curve doing what the datasheet says. Note that the datasheet plots 85 °C directly on the curve; only sub-85 °C intermediate points (like 70 °C, which interpolates to roughly ~12 µA between the 25 °C and 125 °C anchors) require estimation, and those interpolated values are approximations, not datasheet figures.

Symptom 2: No clamping, or rails pulling together (Why is my BAT54HT1 conducting in reverse?)

If the node clamps at the wrong voltage or the diode appears to conduct when it should block, check what reverse voltage you're actually asking it to hold. The absolute-maximum reverse voltage is 30 V, and reverse breakdown is specified as min 30 V at I_R = 10 µA. Push reverse bias to or past 30 V and the diode avalanches and conducts — which shows up in-circuit as "no clamping" or two rails bleeding into each other.

Two related points from the datasheet's ratings section matter here. First, the maximum ratings carry the explicit warning that they "are not valid simultaneously" — each is an independent stress limit, not a combined operating point. Second, there is no reverse-energy or avalanche rating anywhere in this datasheet, and no ESD rating either. The only surge-adjacent spec is the forward non-repetitive peak current of 600 mA (t < 1 s). So if your line can see inductive kickback or ESD-driven reverse spikes that might push past 30 V, this diode alone is not rated to absorb that energy — put a proper TVS in front of it rather than relying on unspecified avalanche capability.

Symptom 3: Fast edges got slower after adding the clamp (How much capacitance does the BAT54HT1 add to my signal?)

When the BAT54HT1 is used as a clamp across a fast signal line, its junction capacitance loads the node. The electrical table gives C_T = 7.6 pF typ / 10 pF max at V_R = 1 V, f = 1 MHz, and the capacitance curve is strongly voltage-dependent:

The good news: in steady clamp duty the diode sits reverse-biased near the rail, where capacitance is lowest. The bad news: near the conduction point — low reverse bias — it presents ~10–13 pF of load. On a slow control line that's negligible; on a fast or high-impedance node it's real loading that will slow edges. The fix: account for the ~10 pF worst case in your node-impedance budget, increase drive strength, or reposition the clamp. If edges only slowed when the clamp was added and the node sits near 0 V bias, this capacitance is your answer.

Symptom 4: The diode runs hot or dies (Is my BAT54HT1 overloaded?)

The SOD-323 package is small, and its thermal numbers are unforgiving. Total device dissipation is 200 mW on an FR-4/FR-5 board at T_A = 25 °C, derating at 1.57 mW/°C above 25 °C, with R_θJA = 635 °C/W and a junction range of −55 °C to 150 °C. Because 635 °C/W is high, the allowable dissipation collapses quickly:

In other words, at a 125 °C ambient you have about 43 mW to work with — roughly a fifth of the headline 200 mW. Compute T_J = T_A + (V_F × I) × 635 °C/W and check it against 150 °C. Also confirm which current limit applies: 200 mA DC, 300 mA repetitive peak, or 600 mA non-repetitive peak (t < 1 s) — these are independent limits and don't validate each other. If a fault path could exceed 200 mA continuous, add a fuse or series current-limit resistor ahead of the diode. The usual resolution for a hot BAT54HT1 is simply that the application belongs on a bigger diode; this is a signal/protection part, not a power-path rectifier.

Symptom 5: Wrong forward-drop reading (What V_F should I expect from the BAT54HT1?)

If your measured forward voltage looks wrong, first check that you're comparing against a specified test point. The datasheet only guarantees V_F at five currents:

The key worst case is 0.8 V max at 100 mA — don't budget a series-diode dropout to the typical value. Second, V_F shifts with temperature: the datasheet's Figure 2 family of curves (spanning −55 °C to 150 °C) documents that "forward voltage increases with decreasing temperature," so a cold or hot reading will differ from the 25 °C table even on a healthy part. The datasheet publishes no numeric mV/°C coefficient, so if you need a precise elevated-temperature value, characterize it on your own hardware.

Symptom 6: Switching behavior is off (Is 5 ns recovery guaranteed in my circuit?)

The headline tᵣᵣ = 5.0 ns max comes with an exact test condition: I_F = I_R = 10 mA, I_R(REC) = 1 mA, with the pulse width obeying t_P ≫ tᵣᵣ. The datasheet's own Figure 1 rig uses a +10 V supply, 820 Ω and 2 kΩ resistors, a 100 µH inductor, and 0.1 µF capacitors. If your switching currents or drive impedance are far from that 10 mA test point, recovery will differ — don't quote 5 ns as universal, and don't blame the diode for a switching fault until you've replicated something close to its specified condition.

Integrating the BAT54HT1 correctly the first time

Most of the symptoms above are preventable at design time. The part is a single two-terminal diode, so there's no supply pin, no decoupling, no pull-up guidance from the datasheet — the integration work is all about the components around it:

• Package and footprint: the "HT1" suffix means SOD-323 (Case 477-02), shipped 3000 per tape and reel; BAT54HT1G is the Pb-free version. The body is ~1.7 × 1.25 mm with ~2.5 mm end-to-end including leads. Mind polarity — a 180° mount on a two-lead diode behaves as a reverse-connected diode and can clamp or load a node incorrectly. The datasheet does not publish a land pattern; the thermal rating is referenced to an "FR-4 Minimum Pad" board, and soldering guidance lives in onsemi's SOLDERRM/D manual.
• Clamp rail margin: keep reverse bias comfortably below 30 V with real margin — never design a clamp point at the rating.
• Bias impedance: keep effective resistance on clamped high-Z nodes under ~100 kΩ so the 2 µA room-temperature leakage (and far worse when hot) stays a non-issue.
• High-speed loading: budget ~10 pF of junction capacitance on any fast node the diode touches.
• Current path: stay within 200 mA DC / 300 mA rep peak / 600 mA non-rep peak, and fuse anything that could fault past that.

Verifying the part on the bench

The datasheet's electrical table gives exact test conditions and pass criteria, so bench verification is straightforward with a source-measure unit and an LCR meter:

The single most important sentence in that table: every limit is only valid at its stated condition — the 2.0 µA leakage figure is specifically at 25 V and room temperature, and the 5.0 ns recovery is specifically at the 10 mA test point. A reading outside limits at some other condition isn't a failed part.

For in-system checks (general practice, not datasheet procedure): apply the expected transient and confirm the node clips near V_rail + V_F with no excursion past 30 V reverse; read the protected node while the diode is reverse-biased and confirm the leakage-induced offset fits your budget — including at hot ambient; and confirm T_J stays under 150 °C at worst-case steady load using the 635 °C/W figure.

A note on what the datasheet doesn't cover

Known limitation: The BAT54HT1 datasheet specifies no ESD susceptibility rating (no HBM/CDM/IEC figure) and no reverse-energy or avalanche rating. The only surge-type limit is the 600 mA non-repetitive forward current (t < 1 s). Treat reverse overvoltage beyond 30 V as an unsupported operating mode, and use dedicated TVS protection for lines exposed to ESD or large reverse transients.

The troubleshooting flow, condensed

When a BAT54HT1 circuit misbehaves, work through the symptoms in this order: (1) hot-only node drift → temperature-dependent leakage through your bias network, not a fault; (2) reverse conduction or no clamping → reverse bias hitting the 30 V ceiling; (3) slowed edges → the ~10 pF junction capacitance on a fast or high-Z node; (4) overheating → V_F × I against the sharply derated 200 mW budget; (5) odd V_F or switching readings → check you're at a specified test point before suspecting the die. And always rule out orientation first — a two-lead SOD-323 mounted backwards is the cheapest possible root cause and a surprisingly common one.

Part page: BAT54HT1.