AMD

XCKU040-L1SFVA784I - Kintex UltraScale FPGA, 784-FCBGA | AMD

Référence : XCKU040-L1SFVA784I ✓ Actif
0.95 V or 0.90 V Vdss 784-BFBGA (SFVA784) FCCSP Package -1L (low-voltage, reduced static power) Speed
MOQ : 1 |

XCKU040-L1SFVA784I Présentation

The AMD (formerly Xilinx) XCKU040-L1SFVA784I is a Kintex UltraScale field-programmable gate array built on a 20 nm process node, packaged in a 784-pin FCBGA (SFVA784) surface-mount package. The device delivers approximately 424,200 logic cells with 468 user I/O, delivering high logic density at low power for mid-range signal processing, industrial, and communications designs.

The -1L speed grade is the key variant feature: these devices can operate at either a 0.95 V or 0.90 V VCCINT core supply and are screened for lower maximum static power than standard -1 grade parts, which directly reduces thermal design burden in dense designs where logic utilization is high.

UltraScale devices provide 16.3 Mb of block RAM, 192 DSP48E2 slices, and 32.5 Mb of distributed RAM, together with 12.5 Gb/s Transceiver Wizard GTX line rates, integrated memory interface controllers, and dedicated PCIe Gen3 hard blocks. These resources allow a single Kintex UltraScale device to replace multi-chip solutions in protocol bridging, video and image processing, industrial control, and test instrumentation.

The UltraScale architecture improves power efficiency through selective bias control and programable dynamic power management, while the AXI4 interconnect fabric and clocking architecture scale to thousands of taps. A 784-ball FCBGA package exposes 468 I/O pins, which is the highest I/O count in the Kintex UltraScale offering and is ideal for parallel sensor arrays, high-parallelism data acquisition, and multi-channel telecom interfaces.

Typical applications include 5G and broadband baseband equipment, radar and electronic warfare signal chains, high-speed industrial data acquisition, machine vision, video infrastructure, and instrumentation requiring large parallel I/O and DSP throughput. The industrial temperature grade and extended product life cycle make the device suitable for long-lifetime industrial programs.

Design consideration: this package has 784 balls on a 31 mm x 31 mm footprint, so PCB design must be done on a multilayer board with buried vias and controlled differential impedance. Differential pairs for Transceivers, clocks, and DDR memory must follow the AMD UltraScale design guidelines, and a complete power-up sequence across VCCINT, VCCAUX, VCCBRAM, and the VCCO rails is mandatory before configuration.

Alternatives de remplacement direct pour XCKU040-L1SFVA784I — même boîtier, compatible broche à broche. Les références à boîtier différent nécessitant une retouche du PCB sont exclues.

Variantes de cette série

Modèles de la même série directement compatibles avec XCKU040-L1SFVA784I (même facteur de forme et même empreinte) — différant par Device Type, Package, Speed Grade, Industrial Junction Temperature, Logic Cells.

AMD
Device Type: Kintex UltraScale Field Programmable Gate Array
Package: 784-terminal FCBGA
Speed Grade: -1
Comparer avec XCKU040-L1SFVA784I →
AMD
Device Type: Kintex UltraScale FPGA
Package: 784-BFBGA, FCCSP
Speed Grade: -1
Comparer avec XCKU040-L1SFVA784I →
AMD
Device Type: Kintex UltraScale field-programmable gate array
Package: 784-FCBGA, 784-BFBGA, FCCSP
Speed Grade: -1
Comparer avec XCKU040-L1SFVA784I →
AMD
Device Type: Kintex UltraScale FPGA
Package: 784-BFBGA, FCCSP
Industrial Junction Temperature: -40 °C to 100 °C
Comparer avec XCKU040-L1SFVA784I →

XCKU040-1SFVA784I

✅ Remplacement direct
AMD
📦 784-BFBGA (SFVA784) FCCSP
Kintex UltraScale field-programmable gate array · 530,250 logic elements · 30,300 LABs/CLBs · 21,606,000 bits · 468 user I/O pins · 784-FCBGA, 784-BFBGA, FCCSP · Flip-chip ball grid array · 784 balls

Contactez-nous pour le prix

Voir la fiche technique →

XCKU040-2SFVA784I

✅ Remplacement direct
AMD
📦 784-BFBGA (SFVA784) FCCSP
Kintex UltraScale FPGA · Kintex UltraScale · AMD · 20 nm · -2 · -3, -2, -1, and -1L · 468 I/O

Contactez-nous pour le prix

Voir la fiche technique →

XCKU040-3SFVA784I

✅ Remplacement direct ⚠️ Spécifications non vérifiées
📦 784-BFBGA (SFVA784) FCCSP
same SFVA784 package, -3 speed grade is the fastest timing class in the family while -1L is the slowest, so power and thermal budget differ

📋 Alternative de référence (hors catalogue)

XCKU040-1SFVA784C

✅ Remplacement direct ⚠️ Spécifications non vérifiées
AMD
📦 784-BFBGA (SFVA784) FCCSP
Kintex UltraScale FPGA · 530,250 logic cells · 468 I/O · 784-BFBGA, FCCSP · Square · 784 terminals · Ball · Kintex UltraScale

Contactez-nous pour le prix

Voir la fiche technique →

XCKU035-1SFVA784C

✅ Remplacement direct ⚠️ Spécifications non vérifiées
AMD
📦 784-BFBGA (SFVA784) FCCSP
Kintex UltraScale Field Programmable Gate Array · 444,343 logic cells · 25,391 CLBs · 468 I/O · 0.95 V · 784-terminal FCBGA · B784 · 0.8 mm

Contactez-nous pour le prix

Voir la fiche technique →

XCKU040-L1SFVA784I Valeurs maximales et caractéristiques électriques

Device Type Kintex UltraScale FPGA (logic cell array)
Logic Cells 424,200 cells
Process Technology 20 nm
Speed Grade -1L (low-voltage, reduced static power)
VCCINT Options 0.95 V or 0.90 V
User I/O 468 I/O
Boîtier 784-BFBGA (SFVA784) FCCSP
Package Dimensions 31 mm x 31 mm FCBGA
Type de montage Surface Mount
Block RAM 16.3 Mb
DSP Slices (DSP48E2) 192
Distributed RAM 32.5 Mb
GTX Transceiver Line Rate 12.5 Gb/s per lane
Supply Voltage Class 0.95 V (per FindIC listing for this device)
Temperature Grade Industrial (-1I suffix)
Packaging / Shipping Tray (per distributor listing)

Guide de configuration des broches du XCKU040-L1SFVA784I 31 mm x 31 mm fcbga

Configuration des broches du XCKU040-L1SFVA784I (boîtier 31 mm x 31 mm fcbga). Ce circuit intégré numérique comprend des GPIO, des interfaces de communication (UART, SPI, I2C) et des broches d'alimentation. Consultez la fiche technique du fabricant pour les fonctions alternatives des broches et les options de configuration. Indispensable pour la conception de systèmes embarqués et la conception de circuits imprimés.

Schéma de brochage du boîtier 31 mm x 31 mm fcbga pour le XCKU040-L1SFVA784I

Aucune donnée de brochage détaillée disponible pour XCKU040-L1SFVA784I.

Consultez la fiche technique pour la configuration complète des broches.

Applications typiques

XCKU040-L1SFVA784I convient pour 6 applications: 5G and Broadband Baseband Equipment, Radar and Electronic Warfare Signal Processing, High-Channel-Count Industrial Data Acquisition, Machine Vision and Industrial Inspection, Video Infrastructure and Broadcast Processing, Aerospace and Defence Avionics.

🌐

5G and Broadband Baseband Equipment

The XCKU040-L1SFVA784I suits radio baseband and small-cell line cards where a single device must absorb multiple DSP pipelines plus control logic. The 192 DSP48E2 slices and 424,200 logic cells allow a single FPGA to implement FFT, channel estimation and rate matching stages, while 468 I/O serve parallel CPRI and backhaul interfaces. The -1L grade lowers static power at a 0.90 V VCCINT setting, which matters in a line card that repeats the same part across many channels in a sealed outdoor enclosure with limited convection cooling. GTX lanes at up to 12.5 Gb/s handle the radio datapath directly, removing the need for a separate PHY device.

✈️

Radar and Electronic Warfare Signal Processing

Wideband radar receivers need many channels of ADC data captured in parallel and then conditioned digitally, and that is exactly where the 468 I/O of the XCKU040-L1SFVA784I helps. ADC samples can be captured on parallel LVDS banks into the 16.3 Mb of block RAM, while the 192 DSP48E2 slices perform digital down-conversion, matched filtering and pulse compression in parallel. The device also provides the deterministic latency needed for coherent processing chains, and the large 31 mm FCBGA land pattern supports the dense decoupling network these high-speed designs require. The -1L low-static-power screening helps in space-constrained RF modules where conducted heat cannot be removed quickly.

🖥️

High-Channel-Count Industrial Data Acquisition

Test and measurement front-ends frequently need to capture many analog channels synchronously, and the XCKU040-L1SFVA784I provides 468 I/O for parallel ADC multiplexing alongside 192 DSP48E2 slices for per-channel calibration and filtering. Because each I/O bank has its own VCCO rail, different interface standards can coexist on the same device, which removes board area otherwise spent on level translators. The 16.3 Mb block RAM allows long acquisition windows for waveform analysis, and the deterministic architecture supports repeatable, calibration-friendly measurements. For instrumentation that runs warm inside a 19-inch rack, the -1L grade is preferred because its lower maximum static power reduces the need for forced air behind the instrument.

📷

Machine Vision and Industrial Inspection

Industrial camera and inspection systems combine parallel image capture with heavy pixel-level processing, and the XCKU040-L1SFVA784I is well matched because 468 I/O can host wide parallel camera interfaces while 192 DSP slices run filter, morphology and correlation pipelines. The large 784-ball package also gives access to enough I/O for multiple cameras plus encoder, trigger and strobe lines without multiplexing. The -1L grade helps in camera housings with limited airflow because it is screened for lower maximum static power, and the industrial temperature grade supports factory-floor environments. Block RAM of 16.3 Mb provides line and frame buffers without external memory for moderate resolution formats.

📺

Video Infrastructure and Broadcast Processing

Video processing boards in broadcast and test equipment need to move pixel data at line rate across many channels, which is where the 468 I/O and 12.5 Gb/s GTX lanes of the XCKU040-L1SFVA784I become valuable. A single part can implement deinterlacing, scaling and chroma processing plus 3D-LUT management, while the transceivers feed or receive video over coax or fibre without an external link chip. The 192 DSP48E2 slices handle FIR filters and motion compensation efficiently, and the 16.3 Mb of block RAM buffers multiple frames. For chassis-mounted equipment, running the -1L grade at 0.90 V VCCINT trims the static power that dominates in lightly loaded video pipelines.

🛩️

Aerospace and Defence Avionics

Avionics and mission-computing modules need predictable timing, deterministic I/O and long program life cycles, and the XCKU040-L1SFVA784I provides an industrial-temperature Kintex UltraScale option with 468 I/O for sensor and datalink interfaces. The 192 DSP48E2 slices support modems, filters and transform engines for onboard signal processing, while GTX lanes carry high-rate sensor or recorder data internally. Using a single large device rather than several smaller ones reduces board area, weight and inter-board interconnect risk in a vibration-controlled enclosure. Qualification status for aerospace programs should be confirmed with the manufacturer, and the power plan should be based on the -1L static power screening to keep thermal margins inside a sealed box.

What is the XCKU040-L1SFVA784I FPGA and what does it do?
The XCKU040-L1SFVA784I is a Kintex UltraScale field-programmable gate array from AMD (formerly Xilinx) providing roughly 424,200 logic cells, 468 user I/O, 16.3 Mb of block RAM and 192 DSP48E2 slices on a 20 nm process node. It is used where a mid-density FPGA must combine high parallel I/O, DSP throughput and multi-gigabit transceivers in a single device, such as in 5G baseband, radar, machine vision and high-channel-count data acquisition systems.
What is the difference between the -1L and -1 speed grade in XCKU040?
The -1L grade is the low-voltage variant of the -1 grade. According to the AMD Kintex UltraScale family datasheet, -1L devices can operate at either 0.95 V or 0.90 V of VCCINT and are screened for a lower maximum static power, whereas -1 devices operate at 0.95 V VCCINT only. Both grades share the same die, package and pinout, so the two are drop-in compatible in the SFVA784 footprint.
How many I/O pins and how large is the package on the XCKU040-L1SFVA784I?
The XCKU040-L1SFVA784I provides 468 user I/O and is packaged in a 784-BFBGA (SFVA784) FCCSP. Distributor and datasheet listings identify the package as a 784-pin FCBGA, the largest ball-count option in the Kintex UltraScale family, which makes it the preferred choice for parallel sensor arrays and wide data buses.
Where can I download the XCKU040-L1SFVA784I datasheet PDF?
The XCKU040-L1SFVA784I datasheet is available from the AMD product documentation archive and is mirrored on Octopart's datasheet page for this device. Search the part number in the AMD documentation portal, or use the datasheet PDF mirrored at Digi Electronics, to obtain the full Kintex UltraScale data sheet covering DC/AC characteristics, package data and pin definitions.
Where do I find the pinout and package pin files for XCKU040-L1SFVA784I?
Pin definitions for the SFVA784 package are published in the AMD UltraScale package device pinout files, which provide ASCII and CSV package files described in UG575. For differential pairs such as DIFF_SSTL12_DCI, use the pin-placement spreadsheet together with the pinout files to confirm that the required strobe and data pins exist on the specific package before committing a layout.
What is the best drop-in replacement for XCKU040-L1SFVA784I?
The closest drop-in replacement is the XCKU040-1SFVA784I, which shares the same 784-ball SFVA784 package and the same die but uses the standard 0.95 V -1 grade instead of the dual-voltage -1L grade. For higher throughput needs the XCKU040-2SFVA784I or XCKU040-3SFVA784I are pin-compatible and faster, and XCKU040-1SFVA784C is the commercial-temperature version of the -1 part.
Can XCKU040-2SFVA784I replace XCKU040-L1SFVA784I in an existing board?
Yes, in most cases. The XCKU040-2SFVA784I uses the identical SFVA784 package, the same 468 I/O and the same supply rails as the -1L grade, so the board does not need modification. The differences are performance-related: the -2 grade delivers higher logic and routing performance, while the -1L grade is screened for lower static power and can run VCCINT at 0.90 V, so power budgets may need re-checking after the swap.
Is XCKU040-L1SFVA784I the same as XCKU040-1SFVA784I?
They are not identical, although they are mechanically and electrically drop-in compatible. The -L suffix denotes the low-voltage grade screened for lower maximum static power and able to run at 0.90 V or 0.95 V VCCINT, while the plain -1 part is a 0.95 V-only device. If your design relies on the 0.90 V VCCINT option for power savings, use the -L part; otherwise -1 provides a more widely stocked option.
What is the XCKU040-L1SFVA784I versus XCKU035-1SFVA784C?
Both parts use the same 784-ball SFVA784 package and the same -1 speed grade, so they are pin-compatible. The difference is die size and resources: XCKU040 provides about 424,200 logic cells versus roughly 371,000 for XCKU035, and XCKU035-1SFVA784C carries a commercial temperature suffix while XCKU040-L1SFVA784I is industrial grade. Use XCKU035-1SFVA784C when the smaller die is sufficient and commercial temperature is acceptable.
Why are the FFVA1156 and FBVA676 XCKU040 variants not drop-in replacements?
Those variants use different physical packages: the FFVA1156 options are in a 1156-ball FCBGA and the FBVA676 options are in a 676-ball FCBGA, whereas the XCKU040-L1SFVA784I uses the 784-ball SFVA784 package. Because the ball counts and pin maps differ, they cannot be soldered onto the same land pattern and require a PCB respin, so they are alternatives only at the board-design level.
What are the key specifications of XCKU040-L1SFVA784I that engineers should know?
Key figures are 424,200 logic cells, 468 user I/O, 192 DSP48E2 slices, 16.3 Mb block RAM, 32.5 Mb distributed RAM, GTX transceivers up to 12.5 Gb/s per lane, a 20 nm process node and a 784-ball FCBGA package. The -1L grade runs VCCINT at 0.95 V or 0.90 V and is screened for lower maximum static power, which is the main selection driver over the standard -1 part.
How much power does the XCKU040-L1SFVA784I consume and how is it supplied?
Supply is divided across several rails: VCCINT for the core, VCCAUX for the auxiliary logic, VCCBRAM for the block RAM and per-bank VCCO for the I/O, each with its own decoupling and ramp requirements. Because the -1L grade is screened for lower maximum static power, core current depends heavily on utilization; use the AMD power estimator with your actual utilization and clocking plan, and follow the prescribed power-up sequencing before releasing configuration.
Is the XCKU040-L1SFVA784I suitable for high-speed serial interfaces?
Yes, the Kintex UltraScale family integrates GTX transceivers with line rates up to 12.5 Gb/s per lane, along with dedicated PCIe Gen3 and memory interface controllers. With 468 I/O and 192 DSP slices the device can drive multiple high-speed lanes plus parallel data paths simultaneously, which suits radar, test instrumentation and telecom line-card designs. Route all differential pairs on continuous reference planes with the impedance and length matching values specified for the selected transceiver quad.
Where to buy XCKU040-L1SFVA784I online and what is the lead time?
The XCKU040-L1SFVA784I is a large, low-volume-count Kintex UltraScale device typically handled as an order-on-request part rather than a stock item. Use the Request a Quote form on this page to obtain current availability, allocation and lead time, along with the appropriate packing and tray configuration. The manufacturer supports the SFVA784 tray format for this part.
What is the price of the XCKU040-L1SFVA784I?
Pricing for the XCKU040-L1SFVA784I is quote-based because this is a high-value, low-volume FPGA order. Request a quote through the Request a Quote form on this page for the exact suffix and quantity you need; the response will reflect current channel pricing, volume breaks and the delivery schedule. Reference the manufacturer data sheet for configuration options and package details rather than catalog pricing.
Is the XCKU040-L1SFVA784I still in production?
The Kintex UltraScale family remains a supported AMD Xilinx product line with the XCKU040-L1SFVA784I listed for order alongside its -1, -2 and -3 grade siblings in the same SFVA784 package. Contact us through the Request a Quote form on this page to confirm current lead time, lifecycle status and the exact configuration suffix you need before committing a design.

Données de référence d'ingénierie pour XCKU040-L1SFVA784I — comparaison, recommandations de conception et informations de conformité.

Guide de sélection

Choose the XCKU040-L1SFVA784I when your design needs more than 400 parallel I/O and roughly 190 DSP slices and you are thermally constrained, because the -1L grade is screened for lower maximum static power and can run VCCINT at 0.90 V. Pick the XCKU040-1SFVA784I when supply certainty matters more than the low-voltage option, since the standard -1 grade is more widely stocked, and pick XCKU040-2SFVA784I or XCKU040-3SFVA784I when timing closure is the constraint and you can absorb the extra static power. Use XCKU040-1SFVA784C only in commercial-temperature enclosures, and XCKU035-1SFVA784C when the smaller die still covers your utilization. The FFVA1156 and FBVA676 variants are only sensible when you are already respinning the PCB, since they use different ball counts and pin maps.

Comparaison avec les alternatives

Paramètre Ce produit XCKU040-1SFVA784I XCKU040-2SFVA784I XCKU040-3SFVA784I XCKU040-1SFVA784C XCKU035-1SFVA784C
Package 784-BFBGA (SFVA784) FCCSP 784-BFBGA (SFVA784) FCCSP - same 784-BFBGA (SFVA784) FCCSP - same 784-BFBGA (SFVA784) FCCSP - same 784-BFBGA (SFVA784) FCCSP - same 784-BFBGA (SFVA784) FCCSP - same
Brand AMD AMD AMD AMD AMD AMD
Speed Grade -1L (0.95 V or 0.90 V VCCINT, low static power) -1 (0.95 V VCCINT) -2 (0.95 V VCCINT) -3 (0.95 V VCCINT) -1 (0.95 V VCCINT) -1 (0.95 V VCCINT)
Temperature Grade Industrial (-1I) Industrial (-1I) Industrial (-2I) Industrial (-3I) Commercial (-1C) Commercial (-1C)
Process Technology 20 nm 20 nm 20 nm 20 nm 20 nm 20 nm
Supply Voltage (VCCINT) 0.95 V or 0.90 V (-1L) 0.95 V 0.95 V 0.95 V 0.95 V 0.95 V
Shipping Format Tray Tray Tray Tray Tray Tray

Principaux différenciateurs

  • Dual-voltage low-power core operation (vs XCKU040-1SFVA784I)
  • Highest I/O count in the Kintex UltraScale family (vs XCKU040-1FBVA676I)
  • Consistent large-package footprint across speed grades (vs XCKU040-2SFVA784I)
  • One package size replaces three different board spins (vs XCKU040-1FFVA1156C)

Notes de conception

The -1L grade lets the core run at either 0.95 V or 0.90 V VCCINT, so sequence the supplies exactly as specified: VCCAUX first, then VCCINT, then VCCBRAM, then per-bank VCCO. Never enable configuration until every rail is stable and within tolerance. Place dedicated bulk and high-frequency decoupling on each rail close to the device, since the 784-ball FCBGA needs a dense via-in-pad and buried-via structure to reach the required decoupling inductance. Keep the 0.90 V option in mind when the thermal budget is tight, because lower VCCINT directly reduces core static and dynamic power.

With 468 I/O in a 31 mm x 31 mm FCBGA, the board must be a multilayer design with buried vias and a solid reference plane per VCCO bank. Route all GTX and other high-speed differential pairs on continuous reference planes, keep intra-pair skew within the limits stated for the selected transceiver quad, and avoid vias in the pair where possible. Use the AMD UltraScale package pinout files and the IO Port spreadsheet together: signals such as DIFF_SSTL12_DCI strobes may not be available on every package, so confirm pin placement before freezing the layout.

Kintex UltraScale power is dominated by dynamic consumption at high utilization and high clock rates, so estimate power with the AMD tool using the real clocking plan and utilization percentages rather than worst-case corners. Because a large FCBGA conducts heat mainly through the board and its vias, reserve a solid copper pour under the device on every layer and provide a thermal via field or a heat spreader if the calculated junction temperature margin is thin. The -1L grade is screened for lower maximum static power, which helps in passive enclosures, but it does not replace airflow planning in a dense system.

A common mistake is assuming that any XCKU040 suffix can be substituted freely: the FFVA1156 and FBVA676 options use different ball counts and pin maps, so they cannot be dropped onto an SFVA784 land pattern. Another frequent error is mixing the -1L part into a board whose power design was calculated for the standard -1 grade, or vice versa. Check the speed-grade, temperature-grade and VCCINT options in the part number against the BOM, and re-verify the static power estimate whenever the grade changes.

Informations de conformité

RoHS
Unknown
REACH
Unknown
AEC-Q100
Unknown
Sans plomb
Unknown
Sans halogène
Unknown
Minerais de conflit
Unknown

No compliance statements are printed in the device listings available for this part. RoHS, REACH, AEC-Q100 and lead-free status should be confirmed from the current AMD product environmental and qualification documentation for the exact orderable part number.

Données vérifiées le : 2026-11-18 — données vérifiées et sélectionnées par l'équipe d'ingénierie composants de XAIPART

Recherches associées

xcku040 datasheet xcku040 2ffva1156e datasheet xcku040 2ffva1156i xcku040 1ffva1156c xcku040 ffva1156 pinout xcku040 3ffva1156e xcku040 1fbva676c xcku040 2ffva1156e XCKU040-L1SFVA784I datasheet pdf XCKU040-L1SFVA784I pinout SFVA784 XCKU040-L1SFVA784I equivalent replacement XCKU040-L1SFVA784I vs XCKU040-1SFVA784I AMD Kintex UltraScale FPGA 468 I/O XCKU040-L1SFVA784I price quote xcku040 ultrascale package pinout files

Composants & termes associés

AMD Xilinx XCKU040-L1SFVA784I XCKU040-1SFVA784I XCKU040-2SFVA784I XCKU040-3SFVA784I XCKU040-1SFVA784C XCKU035-1SFVA784C Kintex UltraScale FPGA field programmable gate array logic cell 784-FCBGA SFVA784 FCBGA package family surface mount DSP48E2 block RAM GTX transceiver VCCINT VCCO VCCAUX VCCBRAM speed grade -1L 20 nm process node UltraScale architecture AXI4 interconnect PCIe Gen3 hard block RoHS REACH AEC-Q100 UG575 radar signal processing 5G baseband industrial data acquisition
RFQ de devis rapide
Renseignez tous les détails — notre équipe commerciale vous répondra sous 24 heures
Référence Fabricant Boîtier QTY Prix cible Total
✓
Devis envoyé !

Nous répondrons à votre e-mail sous 24 heures

1
RFQ envoyée
2
Devis reçu
3
Commande passée
4
Paiement
5
Expédié
6
Livré
Voir les détails de la RFQ