Rectifier Diode Selection: Schottky, Fast Recovery and Standard Compared

A rectifier diode is an easy part to choose from habit. The 1N4007 is in the parts bin, it is rated 1000 V and 1 A, and it goes into the design. Put that same diode in the freewheel position of a 100 kHz buck converter and the converter runs hot and radiates. Forward drop is the number everyone quotes, and it is rarely the one that decides.

The three families, and what each one is for

Standard recovery diodes are ordinary silicon PN junctions: cheap, high blocking voltage, slow. A Vishay 1N4007-E3/73 blocks 1000 V at 1 A with a 1.1 V forward drop and a reverse recovery time near 2 microseconds. Fine at 50 Hz, useless above a few kilohertz.

Fast and ultrafast recovery diodes apply lifetime control to the minority carriers in that same junction. They keep the blocking voltage and cut recovery by two orders of magnitude. A Yageo ES1J blocks 600 V at 1 A with 35 ns recovery and pays 1.7 V for it. An STMicroelectronics STTH2R02UY does 200 V and 2 A at 30 ns with a 1 V drop.

Schottky diodes replace the junction with a metal to semiconductor barrier. No minority carrier storage means no reverse recovery to speak of, only junction capacitance, and the forward drop is low: a Taiwan Semiconductor SS34 gives 3 A at 40 V with a 500 mV drop in SMC. The cost is limited blocking voltage and much higher leakage.

Reverse recovery is the parameter that decides

When a PN diode has been conducting forward current and the circuit reverses the voltage across it, the stored charge in the junction has to be swept out before the diode blocks. Until it does, the diode conducts backwards. The time this takes is trr, and the charge involved is Qrr.

At 100 kHz the switching period is 10 microseconds, so a standard recovery diode with a 2 microsecond trr spends a fifth of every cycle conducting in the wrong direction and the high-side switch is briefly shorting the input rail through it. The current spike shows up as switching loss in the MOSFET, as a hot diode, and as broadband noise on an EMC scan. A 35 ns ultrafast part shrinks the same event to 0.35% of the period. Family choice therefore comes before forward drop: below a few kHz take the cheap standard recovery part, and above it the decision is between fast recovery and Schottky, where voltage decides.

Forward drop and conduction loss, worked

Take a non-synchronous buck: 12 V in, 5 V out, 3 A load, 100 kHz. Duty cycle is 5/12, so the freewheeling diode conducts for 58% of each cycle. With an SS34 at roughly 500 mV, conduction loss is 0.5 x 3 x 0.583 = 0.87 W. With a Vishay MURS360, a 600 V 4 A ultrafast part with a 1.28 V maximum drop, the same position dissipates 2.24 W. That is 1.4 W of extra heat and a heatsink decision that did not need to exist, because a 12 V rail never needed a 600 V diode.

Read the drop at your actual current, not at the datasheet headline. The 500 mV figure for an SS34 is quoted at its full 3 A, while a small signal Schottky quoting 480 mV may be doing so at 100 mA.

Where Schottky diodes stop being the easy answer

Reverse leakage (Ir) is the Schottky's weak point. It runs orders of magnitude above a PN junction and roughly doubles for every 10 degrees of junction temperature rise. Leakage times blocking voltage is real dissipation, that dissipation raises the junction temperature, and the higher temperature raises the leakage again, a loop that can run away in a hot enclosure. Read the Ir curve at your worst case junction temperature, not the 25 degree number.

Repetitive peak reverse voltage (Vrrm) is the other limit. Schottky parts thin out above 100 V while ultrafast PN parts reach 600 V without difficulty, so above roughly 200 V the choice is made for you. Give the rating headroom on a hard switching node as well: a 12 V buck with a leaky layout can ring the switch node past 24 V, so 40 V is the floor and 60 V the safe answer. The STPS3L60SY is the same 3 A in the same SMC package at 60 V, at 940 mV instead of 500 mV.

Non-repetitive surge current (IFSM) matters at mains input rather than in a converter. A 1N5408 is rated 3 A continuous but 125 A for a single half cycle, and that second number is what keeps a bridge alive at plug-in, not the continuous rating.

Comparison table

Bridge rectifiers and mains input

For a 230 VAC input the peak is 325 V, so a 400 V bridge leaves nothing for line transients and 600 V or 800 V is the normal choice. A Diodes Inc KBP series part at 800 V and 2 A covers a small offline supply. Size average current for the DC output rather than the RMS line current, then check the surge rating against the bulk capacitance you are charging.

Selection checklist

• Fix the switching frequency first. Below a few kHz standard recovery is fine, above it is not.
• Set Vrrm from the measured peak on the node including ringing, never from the nominal rail.
• Check Vf at your actual current and duty cycle, then compute conduction loss in watts.
• Read Schottky leakage at the maximum junction temperature you expect, not at 25 °C.
• Confirm IFSM against inrush wherever a capacitor is charged from cold, and match the package to the dissipation. SMA and SMC parts with the same die differ in thermal resistance.

FAQ

Q: Can I use a Schottky diode everywhere to save the forward drop?

Only where blocking voltage and leakage allow it. Standard planar silicon Schottky rectifiers thin out above 100 V (with trench TMBS parts reaching 150 V to 200 V), and reverse leakage climbs steeply with junction temperature. Above 200 V, standard silicon Schottkys cannot compete with ultrafast PN diodes. However, for high-voltage DC links (400 V to 1200 V in PFC stages and inverters), Silicon Carbide (SiC) Schottky diodes provide near-zero reverse recovery with high voltage ratings, bridging that exact gap. On low-voltage 5 V or 12 V rails, silicon Schottky remains the dominant, low-cost choice.

Q: What is the difference between fast recovery and ultrafast recovery?

There is no standardised boundary, so the labels are closer to marketing than to specification. Fast recovery roughly covers a few hundred nanoseconds and ultrafast covers tens, but the only number that means anything is the trr on the datasheet and the conditions it was measured under. Compare those directly and ignore the family name.

Q: Why does my Schottky rectifier get hotter than the loss calculation predicts?

Usually one of two things. Reverse leakage dissipates during the blocking interval, which the conduction calculation ignores entirely and which grows sharply with temperature. Or the forward drop came from the typical curve at 25 °C rather than the maximum at operating temperature. Recompute with maximum Vf at the hot junction and add the leakage term.

Q: Do I still need a rectifier diode if I use a synchronous converter?

The synchronous MOSFET replaces the freewheeling diode for conduction, but many designs still fit a small Schottky across the low-side FET. It carries current during the dead time before the FET turns on, keeping the FET body diode out of conduction. Body diodes are slow PN junctions with poor recovery, so shunting them removes a source of switching loss and ringing.

Filter by voltage, current, forward drop and package across Novapart's Schottky diodes, fast and ultrafast recovery, standard recovery and bridge rectifier ranges, or go straight to a brand you already qualify such as Vishay Schottky rectifiers. The MOSFET selection guide and the switching regulator comparison cover the other half of the loss budget, and the thermal guide turns those watts into a junction temperature. Sizing a rail or an offline front end? Upload the BOM for consolidated sourcing with cross-references and lead times.