|
This is a selection of some of the most frequently asked questions that arise. If these do not resolve your query, then please contact our sales team using the contacts page, who will be happy to help.
|
Commercial |
|
|
|
|
1 How much do RFEL's signal processing cores cost? |
top |
|
|
Prices for RFEL's standard signal processing cores start at around $2,000 for a single use licence, with prices increasing depending on the complexity and performance of the core, but falling with volume of use.
|
|
2 What are the licensing/royalty arrangements? |
top |
|
|
RFEL's standard FPGA cores are licensed on a per-use basis, where each "use" is defined as a single instantiation of the core on a physical FPGA device. Licence costs can be viewed in a similar way to the cost of physical components.
Licence prices fall on a sliding scale as greater volumes are purchased. Volume discounts are calculated based on the total number of licences purchased, rather than on each single order, which means that licences only need to be purchased as and when required.
Custom designs or complex modification work is generally developed under a Modification Contract, with a royalty arrangement for subsequent uses of the resulting core.
|
|
3 What are the typical lead-times? |
top |
|
|
Lead-times depend on the complexity of the core, and how much customisation work is required. Many of the standard signal processing cores can be delivered within 2-4 weeks from receipt of order.
|
|
4 What types of warranty and support contract do RFEL provide? |
top |
|
|
RFEL provide a bit-true Matlab model for each core, which acts as the reference design for warranty purposes.
In addition to the standard warranty, RFEL can also provide an optional Support Contract arrangement. This includes provision for support during customer integration of the core, but also enables the customer to request minor design changes to the core that may result from initial testing and evaluation. The Support Contract is based on a fixed price arrangement whereby the customer effectively purchases RFEL engineering man-hours at a preferential rate.
|
|
5 Do RFEL provide custom FPGA designs? |
top |
|
|
Yes. A large proportion of the development work undertaken at RFEL is to create complete FPGA system designs to meet specific customer requirements. In many cases these designs are based on a collection of standard FPGA products, which are modified and/or integrated together.
Our team of experienced systems engineers are able to specify and design complete radio systems from antenna through to application, with an emphasis on developing signal processing algorithms that are optimised for FPGAs. By drawing on our large portfolio of existing IP we are able to rapidly implement these designs on FPGAs with minimal risk.
|
|
6 Who are RFEL's customers? |
top |
|
|
RFEL supply their FPGA cores and system services across a wide range of markets throughout the world, including major defence contractors in the United States and Europe, test equipment manufacturers, and telecommunications companies. Recent press releases from RFEL can be viewed here. If you require more information about our previous projects please contact us to receive a copy of our Capability Statement.
|
|
7 Do RFEL sell complete products? |
top |
|
|
Yes. A lot of effort has been committed over several years to the development of signal processing algorithms, and the engineering design team have a broad range of capabilities in the design of radio systems, digital hardware, signal processing algorithms, FPGA implementations and application software. Also, our experience spans a range of high-tech industries including satellite communications, military RADAR, commercial telecoms and instrumentation.
So although RFEL continues to sell significant volumes of cores and system-on-chip designs, a major part of the company’s business is also now involved in developing and selling finished products that incorporate these DSP techniques. These products are then either supplied direct to customers, or are supplied to the OEM (original equipment manufacturer) who then sells on to the end customer. RFEL has had many years experience in the design of complete products that include RF (radio frequency) and DSP techniques, and so have developed products including:
|
|
• |
digital receivers that are able to acquire and process signals (used in communications and some military and governmental applications) |
|
|
• |
spectrum analysers (that are used to measure, analyse and observe signals) |
|
|
• |
satellite communication and basestation channelisation designs |
More information on these products is available here
|
|
8 What is the benefit of licensing an RFEL core, why can't I just buy it and have the source code? |
top |
|
|
The development of complex cores is a costly exercise and in order to recoup the cost of developing the IP and also to protect the company’s IP, then a licensing model is applied in most cases. This licensing model also benefits the customer in that he needs to only pay for a set number of uses of the core, and if the forecast product volumes are not realised, then no further payments are necessary. This approach therefore does share some of the risk between RFEL and the customer.
Although the IP is retained by RFEL, the use of the design should be considered as little different to buying any other component. For example there is no significant difference between buying say 100 amplifiers or 100 uses of a particular IP core – in both cases the customer is buying the product and not the IP.
As referred to earlier, in question 2, a sliding scale per use is applied, and we do have a flexible approach to licensing. For example an agreement can be reached for a one-off payment for an unlimited use licence for a specific application.
If source code is essential, then we would seek to reach agreement on this, including whether ESCROW was a viable way forward. In summary, always ask and hopefully we should be able to reach an agreement. |
|
General Technical |
|
|
|
|
11 What FPGA devices do RFEL support? |
top |
|
|
RFEL are able to provide designs for the complete range of FPGA devices provided by Xilinx and Altera. We are able to support other devices by agreement.
|
|
12 What format is used for delivering FPGA cores? |
top |
|
|
In general RFEL supply FPGA cores using the EDIF netlist format, which allows the user to quickly integrate the core into the rest of their FPGA design. Alternatively, the core can be supplied as a binary bit-stream for downloading to the device, in which case RFEL take complete responsibility for the FPGA design.
|
|
13 Does RFEL recommend hardware platforms for their designs? |
top |
|
|
As part of our initial quotation process, we will usually be able to recommend a target FPGA device for the core, and the likely FPGA resources required by the core. At this stage we can also provide systems advice that includes interfacing requirements, analogue-to-digital converters, and other peripherals.
RFEL is independent from FPGA board companies, but does have close relationships with many of the major manufacturers. As part of our turnkey development services we are able to specify and procure complete hardware systems, and integrate the FPGA core.
|
|
14 Does RFEL produce its own hardware designs? |
top |
|
|
In instances where a suitable FPGA platform is not available commercially, RFEL can provide bespoke hardware design services. Generally our hardware developments are geared to high performance requirements, such as multiple FPGA designs or large memory requirements. More information about our hardware designs can be viewed here.
|
|
15 How can I find out what FPGA resources are required for RFEL's cores? |
top |
|
|
Examples of silicon usage for many of the standard cores can be found in the individual datasheets from our Product List. We also provide resource and power estimates as part of our initial quotation service.
|
|
16 What sort of simulations do RFEL provide? |
top |
|
|
All RFEL cores are supplied with a fixed-point Matlab model in MEX format allowing the customer to quickly incorporate the RFEL core into any existing Matlab simulation. The fixed-point model is a bit-precise reference for the final FPGA implementation, and acts as a warranty for the final delivery.
VHDL code is simulated at RTL level using ModelSim and then validated against the Matlab model. At the RTL level, the simulator represents the behaviour of lines of VHDL code. This is used to demonstrate that in principle the code carries out the correct signal processing.
The generated VITAL netlist is simulated at gate-level using ModelSim and is again validated against the behavioural model. In the VITAL simulation actual logic gates are simulated with all the associated delays generated by the place and route tools (.sdf file). The VITAL simulation accurately represents the performance of the logic gates on the FPGA. This simulation therefore forms the basis of the final design sign-off. Power estimates are optionally available at this stage.
|
|
17 Why do RFEL use fixed-point arithmetic? Doesn't this result in inferior performance than floating-point arithmetic? |
top |
|
|
RFEL use fixed-point arithmetic for the majority of their algorithms as it results in designs that are far more efficient in terms of their silicon requirements and power consumption. Fixed-point implementations also allow for greater parallelism within the algorithm and enable higher speeds to be achieved.
Careful design of the arithmetic processes when implementing the algorithms can ensure that numerical precision is equivalent to that of a floating-point design. "Growing" integer bit-widths gradually through algorithms can ensure no loss of precision, whilst keeping the design as silicon efficient as possible.
In some cases it may be desirable to accept some loss of precision in order to reduce the silicon requirement for the design. RFEL's core generation process is extremely flexible in this regard and allows integer bit-widths to be specified at each stage through the algorithm. When this is necessary our experienced system engineers are able to make recommendations on the best approach to take, and our bit-true Matlab models allow the effects to be quickly evaluated against a floating-point model.
|
|
Channelisers |
|
|
|
|
21 What is a channeliser? |
top |
|
|
In many radio applications it is necessary to take a wideband input spectrum and extract from it much smaller bands (channels), where each channel can potentially contain a signal that must be demodulated or analysed. An example is terrestrial FM radio, which operates in the band 88 to 108 MHz, divided into a number of 200kHz channels for the individual radio stations. A radio receiver must be capable of tuning to a particular radio channel of interest, and "channelising" the 200kHz band for subsequent demodulation of the audio signal.
In the FM radio example, the receiver only needs to receive one channel at any one time, but there are other systems where the receiver must be capable of receiving many channels at once. Examples include: satellite communications receivers, telecoms base stations, monitor receivers and relay nodes for defence receivers. RFEL's channelisation products are aimed at complex multi-channel applications, and can handle simultaneous reception of 1000 or more channels.
This article, reprinted from Wireless Design Systems magazine, describes the different techniques that can be used for wideband channelisation.
|
|
22 How can frequency transforms be used for channelisation? |
top |
|
|
Frequency transforms, such as the FFT, are special cases of channeliser designs. They divide the input bandwidth into a number of evenly spaced frequency bands, commonly referred to as "bins", in order to allow the frequency content of an input signal to be analysed. In fact, the operation of the FFT is analogous to a stack of evenly spaced band-pass filters, where each of the frequency bins represents the output of a filter. When viewed in this way, an FFT can be considered as a simple channeliser that converts an input signal into N evenly spaced channels, where N is the length of the FFT.
Like all physically realisable filters, the bins of an FFT do not provide perfect "brick-wall" filter performance - there is always spectral power that leaks from its actual bin into the bins on either side. These filtering characteristics can be controlled by the use of a windowing function at the input of the FFT.
|
|
23 What are the advantages and disadvantages of the Polyphase DFT? |
top |
|
|
The Polyphase DFT can provide significantly superior filter performance when compared to a standard FFT. In fact adjacent bin rejection can designed to be as much as 90dB or more. More information about the Polyphase DFT can be found in the datasheet.
The disadvantage of the improved filter performance offered by the Polyphase DFT is the increased transient response time when compared to an FFT of the same length. Whether or not this presents an issue for a particular application will depend on the types of signals that are being monitored. This paper compares the transient responses of the various filter banks offered by RFEL.
|
|
24 What is the PFT and what are its advantages? |
top |
|
|
The PFT (Pipelined Frequency Transform) is a proprietary architecture for real-time channelisation. Unlike the FFT, the PFT can provide outputs at intermediate frequency resolutions (e.g. 2, 4, 8, 16 bins, etc.) up to the length of the PFT . This is equivalent to having simultaneous Polyphase DFT's of different lengths running on the same input data. Furthermore, the PFT allows a separate filter to be defined for each bin at each output resolution.
There are two main uses for the PFT. Firstly, use of the intermediate frequency resolutions allows for simultaneous time and frequency decimation. This can be advantageous in complex signal environments where it is desirable to maximise receiver sensitivity for a variety of different pulse lengths.
The second use of the PFT is for flexible channelisation systems whereby a number of different size channels are required with different filter characteristics. This is often a requirement for multi-standard and multi-mode radio systems.
|
|
25 Is the PFT more efficient than an FFT? |
top |
|
|
The PFT is a multiplier-less design, and for some target devices may result in a more efficient implementation than an FFT. However, if the target is a modern FPGA device with multipliers, then the FFT algorithm is likely to result in the most efficient implementation. In this case the PFT should only be considered for applications where its specific features are required.
|
|
26 How do RFEL's channelisers compare to Digital Downconverter (DDC) solutions? |
top |
|
|
Fundamentally, RFEL's channeliser designs are based on frequency transforms such as the FFT, Polyphase DFT, PFT and Mixed-Radix designs. These techniques are extremely efficient at simultaneously channelising a number of channels. As a general rule, these techniques become more efficient than DDC solutions when there are eight or more channels to be down-converted.
Frequency-transform techniques can be highly optimised when the actual requirements of each design can be constrained in some way (see FAQ 20 here). Through these optimisations it is possible to develop 'single-chip' channelisers capable of down-converting thousands of channels simultaneously. Attempting such a design with a DDC-based architecture would not be economically viable.
|
|
27 What are the critical performance requirements that determine the complexity of a channeliser? |
top |
|
|
Rather than supplying a standard channeliser core, RFEL has a collection of techniques and FPGA designs that can be configured to meet just about any channeliser requirement. Configuring our designs in this way means that the most efficient implementation can be achieved for each particular application.
The major factors that determine how efficiently a channeliser can be implemented include: input bandwidth, number of channels, required filter performance, independent channel tuneability, channel bandwidths and output sample rates. RFEL's sales team can help to define channeliser requirements, and can then quickly provide silicon estimates.
|
|
Frequency Transforms |
|
|
|
|
31 What is the difference between HiSpeed and QuadSpeed FFTs? |
top |
|
|
RFEL provide a range of FFT architectures that have been optimised for different transform lengths and input sample rates. The HiSpeed range of cores is intended for sample rates up to 100 MSps complex, and transform lengths up to 128K complex samples. QuadSpeed solutions are less efficient in terms of their silicon usage, but will run at speeds up to 400 MSps complex.
RFEL are continuously expanding their FFT capabilities with new architectures, and to exploit new FPGA devices. Currently in development is the next generation of FFT cores, which will be capable of running at speeds of 3.2 GSps complex, or lengths up to 64M complex samples, using currently available FPGA devices.
|
|
32 What are Mixed-Radix DFTs and what are they used for? |
top |
|
|
The FFT is a highly efficient implementation of the Discrete Fourier Transform, but it is limited to power-of-two transform lengths (2, 4, 8, 16, etc.). RFEL's mixed-radix cores allow arbitrary length DFTs to be constructed..
Arbitrary length DFTs have applications in channeliser designs where it is necessary to precisely set the output sample rate of each channel. They have also been used in spectral analysis designs where there are fixed constraints on the sample rate, transform rate, or frequency resolution.
|
|
33 What is a distributed half-band filter (DHBF)? |
top |
|
|
A DHBF is generally used at the input to a digital processing system. It takes as input a digitised stream of real samples, for example the output of an ADC, and converts this into a stream of complex samples at half the sample rate. In effect, the DHBF is a highly efficient down-converter that performs down-conversion of the input signal to a zero-IF, followed by filtering to remove aliases, and then decimation-by-two. Reducing the sample rate and converting to a complex signal can significantly reduce the subsequent processing requirements.
RFEL are also able to offer third-band, quad-band and other variants of the design to suit particular input signal bandwidths.
|
|
34 What is special about RFEL's frequency transform products? |
top |
|
|
RFEL use highly efficient pipelined architectures to achieve solutions that are highly optimised in terms of silicon usage and able to support high sample rates and long transforms. FFT cores are available as Standard Products that can be delivered to order within very short timescales.
A range of associated products is available which can be pre-integrated with the FFT to create a single self-contained spectral analysis FPGA core. Examples of associated products include, input buffer, DHBF, windowing function, phase-magnitude calculation and fixed-to-floating-point conversion.
RFEL also provide custom development services to tailor standard designs to meet particular specialist requirements.
|
|
35 What are the bit-width considerations? |
top |
|
|
All of RFEL's frequency transform solutions are tailored to the particular precision requirements of each application. This ensures that the design is as silicon efficient as possible whilst guaranteeing that the particular dynamic range and precision requirement of the application can be met.
Based on the particular bit-width of the input signal (determined by the ADC), and the required precision at the output, RFEL will carefully select bit-widths for the windowing coefficients and twiddle factors, and control bit-growth through the transform.
|
|
SpectraChip™ |
|
|
|
|
41 What is SpectraChip™? |
top |
|
|
SpectraChip™ is a range of FPGA-based solutions for spectrum analyser instruments. The designs provide replacements for traditional analogue style spectrum analyser circuit designs, and include IF-strip replacement, digital sweep, and digital frequency transforms.
|
|
42 What are the benefits of SpectraChip? |
top |
|
|
Updating spectrum analyser designs to replace traditional analogue components with modern digital techniques helps to make the systems easier to manufacture, more reproducible and more reliable. In particular, the calibration requirements for the system can be substantially reduced.
New features also become possible. Most existing spectrum analyser designs are based on a swept approach to monitoring a band of the spectrum, which means that the system is only looking at each part of the spectrum for a finite period of time. Using FFT techniques to perform the spectral analysis means that the entire band of interest can be analysed at all times, and that even very short duration pulses can be captured by the instrument. The addition of complex triggering allows storage of spectral data to occur upon a particular event in the frequency domain. This can then be presented to the user using 3-dimensional "waterfall" plots or 2-dimensional spectragrams to show how the signal changes over time.
|
|
Products, System Design and Consultancy |
|
|
|
|
51 In what technical areas is RFEL able to provide consultancy? |
top |
|
|
RFEL's engineers have a broad range of capabilities in the design of radio systems, digital hardware, signal processing algorithms, FPGA implementations and application software. Our experience spans a range of high-tech industries including satellite communications, military RADAR, commercial telecoms and instrumentation.
Our industry leading FPGA cores are often just the start of the contribution we can make to a development project. Examples of how RFEL has been involved in previous projects include: design of radio system architectures, development of RF, mixed signal and FPGA hardware platforms, assistance with the selection of COTS hardware platforms, development of Windows-based application software and integration of the FPGA core onto the target hardware.
|
|
52 What products do RFEL manufacture? |
top |
|
|
RFEL continues to supply significant volumes of IP cores and system-on-chip designs, but a major part of the company’s business is also now in developing and selling finished products that incorporate these specialist DSP techniques. These products are then either supplied direct to customers, or are supplied to the OEM (original equipment manufacturer) who then sells on to the end customer.
RFEL has had many years experience in the design of complete products that include RF (radio frequency) and DSP techniques, and so have developed products for the high-growth markets of wireless communications, government services and defence, that include:
|
|
• |
digital receivers that are able to acquire and process signals (used in communications and some military and governmental applications) |
|
|
• |
spectrum analysers (that are used to measure, analyse and observe signals) |
|
|
• |
satellite communication and basestation channelisation designs |
More information on these products is available here |
|