Contact Us

HX2580708L4

Brand: Hisiwell
Hisiwell HX2580708L4 is a discrete semiconductors component designed for DC-DC Converters & Power Supplies and related applications. In stock at Superb Automation — request a quote for competitive pricing and fast delivery.
quote now

Product Specifications

HX2580708L4 — Discrete & Power Semiconductors | Superb Automation

HX2580708L4

Hisiwell — Discrete & Power Semiconductors

Product Overview

The HX2580708L4 from Hisiwell is a discrete power semiconductor delivering efficient switching and power conversion. Hisiwell is a recognized semiconductor supplier providing reliable components for global electronics manufacturing.

Superb Automation partners with leading brands including Qorvo, TI, ADI, XILINX, NVIDIA, Intel, ST, MACOM, Renesas, and Coilcraft to provide complete BOM and supply-chain solutions.

Key Features

  • Low Rds(on) for minimal conduction losses in power switching
  • Fast switching speed with low gate charge for high-frequency operation
  • Avalanche-rated for robust operation under inductive load conditions
  • High junction temperature rating for demanding environments
  • Logic-level gate drive compatible for simplified design
  • RoHS and AEC-Q101 qualified options available

Key Specifications

Vds / VceV
Id / IcA
Rds(on) / Vce(sat)mΩ
QgnC
Technology—
Package—
Configuration—
Mounting—

Applications

Switch-Mode Power Supplies Motor Drives & Inverters Battery Management Systems DC-DC Converters Load Switching Automotive Power Distribution

Request Quote — In Stock, 100% New & Original, Full Warranty

HX2580708L4: a 70 W GaN transistor for 5 to 6 GHz on a 28 V rail

HX2580708L4 is an unmatched GaN power transistor from Hisiwell in the 5 GHz to 6 GHz band, printed in the vendor 28 V table with 70 W of output and continuous wave operation. Unmatched is the operative word: the input and output matching networks are not on the package, so the part is a die on a flange that a board tunes, unlike the internally matched HXN40xxx family which arrives at 50 ohm and is used behind connectors and microstrip. The vendor table for this part prints output power, band, rail voltage and mode, and leaves gain and efficiency blank, which means the numbers that decide a link budget have to come from the device document rather than from the table.

The band is the one that carries point to point links, radar front ends, satellite uplinks and the 5 GHz Wi-Fi band at the low end. In that region the hard problem is not usually gain, it is heat per watt of output and the loss of the combining network after the device. A 70 W part on 28 V sits in the middle of the ladder: small enough to combine four of them into a 250 W module, large enough that one device can drive a 20 W average signal with headroom for the peak to average ratio of a modern modulation.

Every figure on this page comes from the Hisiwell unmatched GaN 28 V category table, read on 4 November 2026, plus the arithmetic that follows from those figures. Where the table prints a blank cell the page says so instead of filling it with a guess. The device document is what governs a purchase, and the difference between the table and the document is stated wherever it matters.

What 70 W means once the amplifier is built

70 W is 48.45 dBm. A four way combiner takes four of those to 54.5 dBm before combining loss, and a typical 0.5 dB per path loss at 5.5 GHz brings the module to about 52.5 dBm of real output. That is the arithmetic that decides how many devices a module needs, and it is done in dBm rather than watts because the losses are in dB.

Heat is the other side. The table does not print drain efficiency for this device, so the DC input has to be taken from the document. As a worked example, at 50 percent efficiency a 70 W output draws 140 W from the 28 V rail, which is 5 A, and turns 70 W into heat on a flange that is a few millimetres square. At 40 percent efficiency the same output draws 175 W and turns 105 W into heat. The difference between those two cases is the difference between a heatsink that fits in a 1U chassis and one that does not, which is why the efficiency number is worth chasing before the board is frozen.

Continuous wave at this band means the thermal path is steady state, not pulsed. There is no duty cycle to divide the heat by, so the flange temperature the vendor specifies for the document is the number that has to be held, and the heatsink is sized from the steady state heat rather than from an average.

Where this part sits in the 5 to 6 GHz ladder

Part numberBandPrinted outputRailMode
HXN502145 GHz to 6 GHz50 W peak28 VCW
HXN503355 GHz to 6 GHz70 W peak28 VCW
HX2580708L45 GHz to 6 GHz70 W peak28 VCW
HXN502175 GHz to 6 GHz100 W peak28 VCW
HXN502225 GHz to 6 GHz140 W peak28 VPulse
HX5060140P25 GHz to 6 GHz140 W saturated, 11 dB, 50 percent28 VCW
HXN503344.4 GHz to 5.9 GHz25 W peak, 4 W average, 12 dB, 25 percent28 V5.2 to 5.9 GHz
HX446050F24.4 GHz to 6 GHz50 W saturated, 15 dB, 55 percent28 VCW
HX446030F24.4 GHz to 6 GHz30 W saturated, 14 dB, 55 percent28 VCW

Two readings matter here. First, the same 70 W appears twice, once as HXN50335 and once as HX2580708L4, with the same band, the same rail and the same mode. Two part numbers at the same point of a table usually means two die generations, two package options or two markets, and the way to tell them apart is the device document, not the category page. Second, the ladder is not linear in price: the 100 W part is a different die and a different matching problem, not a 70 W part pushed harder, so a link budget that needs 3 dB more should expect to change the board as well as the part number.

The figures the table gives, and the ones it does not

FunctionUnmatched GaN power transistor, 50 ohm interface at the flange
Frequency band5 GHz to 6 GHz
Output power as printed70 W in the peak column
Output in dBm48.45 dBm
Drain voltage28 V nominal
Operating modeContinuous wave
GainNot printed in the category table
Drain efficiencyNot printed in the category table
DC input at 50 percent efficiency140 W, which is 5 A from 28 V
Heat at 50 percent efficiency70 W steady state on the flange
DC input at 40 percent efficiency175 W, which is 6.25 A from 28 V
Heat at 40 percent efficiency105 W steady state on the flange
MatchingExternal, board side input and output networks
TechnologyGallium nitride on silicon carbide
PackageFlange mount, screw down
Thermal resistanceIn the device document, not in the category table
Load pull at band edgesIn the device document, not in the category table

The blank cells are the honest part of this page. A gain figure invented here would put a wrong stage gain into somebody's cascade, and an efficiency figure invented here would size a heatsink wrongly. What the table does give is enough to choose between the 50 W, 70 W and 100 W rows and to know which rail the part wants, and that is the decision the table was built for.

What an unmatched part asks of the board

An unmatched device moves the matching work onto the board. At 5 GHz to 6 GHz that work is microstrip or stripline on a low loss laminate, with the input network set for gain flatness and the output network set for power and for the second harmonic termination. The board material matters at this band: a standard FR-4 at 5.5 GHz has enough loss that the output match cannot recover it, which is why parts in this band normally sit on a PTFE or a hydrocarbon laminate with a controlled dielectric constant.

The bias feed is the second board decision. A quarter wave line at 5.5 GHz is about 13 mm in air and about 9 mm on a laminate with a dielectric constant of 3.5, so the bias network is short enough to be laid out as part of the output match rather than as a separate block. A bias line that is a quarter wave at the wrong frequency makes the pulse droop or the gain tilt, and on a continuous wave part it shows up as a slow thermal walk instead of an obvious fault.

The third is the flange. A screw down flange needs a flat surface, a thin layer of thermal compound or a phase change material, and screws torqued in the order the document gives. A flange that is not flat to the heatsink loses contact area at the corners, and at 70 W of steady state heat the corners are where the die is hottest.

Buying this part without a spec sheet surprise

Three things are worth settling before a purchase order. The first is which of the two 70 W numbers the requirement is written against, because the vendor table prints HX2580708L4 and HXN50335 at the same point and only the document separates them. The second is the gain and efficiency the link budget assumes, because those two cells are blank in the table and a quote cannot create them. The third is the flange and package variant, since a part that is correct electrically and wrong mechanically is a scrap part.

What we hold on a part like this is the commercial side: the exact orderable number, the date code, the packaging, and a stock position against a build plan. We also keep the alternative list for the case where the 70 W row is not available and the 50 W or 100 W row is, which is a conversation about the matching network and not a straight swap.

Related pages: RF power transistor cross reference for the alternative list at this band, RF procurement for the supply side, RF and microwave board support for the laminate and the assembly, and defense grade components for the traceability that a radar programme usually asks for.

Questions buyers ask about a 70 W GaN part at 5 to 6 GHz

Can one device cover both the 5.15 GHz to 5.85 GHz Wi-Fi band and a 5.4 GHz radar front end. The band is wide enough that one die can cover it, but the matching network cannot be optimised for both at once. A board tuned for the radar band will lose gain at the top of the Wi-Fi band, and a board tuned flat across the whole 1 GHz will give up output power. Pick the band edge that carries the requirement and tune to that.

Does the 28 V rail have to be regulated. It has to be clean and it has to be sequenced after the gate bias, because a GaN device with drain voltage and no gate bias is the fastest way to destroy it. The vendor document gives the sequence; the board has to enforce it, and the current limit on the drain supply is part of that protection.

How is the part tested before it ships. A flange device is tested at the die level by the vendor, and at our end the incoming check is a visual and dimensional check plus a sample test in a fixture. RF power measurement at this band needs a calibrated setup, so a programme that needs a data sheet of measured output per unit should plan for that test to be done in a laboratory rather than assumed at goods inwards.