Ask a number of marketing professionals about upcoming trends, and you’ll get the same predictions year after year—and those predictions will be, well, predictable. For truly surpising and valuable insights into the industry’s future, you have to go to the guys in the trenches.
For example, a closer look at eight-year-old analog startup Samplify Systems reveals the rocky road that entrepreneurs must travel. Success indicates strategy, though, and how one company moves can herald how the industry will move.
Also, Maxim Integrated Products and its CEO Tunç Doluca have shown that the company knows how to weather hard times and prepare for what’s next (see “Power R&D Will Play A Key Role In Next-Generation Analog Design”). So have National Semiconductor and its technology guru Rick Zarr (see “Digital Convergences Will Drive Analog Semiconductor Technology”).
Linear Technology is another industry giant with a proven track record and an optimistic future. Of course, a few words with its founder and CTO (and Electronic Design Engineering Hall of Famer) Robert Dobkin are always valuable, especially when he’s discussing issues related to integration and the use of outside foundries. Check out his thoughts in an exclusive video series available at www.engineeringTV.com.
EXPECTATIONS MUST EVOLVE
Circa 1990, the mantra was “real men have fabs.” And that axiom applied with particular force in the analog realm, where process secrets were a key element to achieving top performance. Over the last two decades, foundries have severely eroded if not entirely erased that position.
So why follow the entrepreneurial route at all? Why not work in or be a design house? Independent chip designers say there’s no longer a lot of money to be made as an independent analog chip designer, even if you have respectable experience and some good ideas—or even if you drop your rates. But designers who don’t like working for a big company can find success as intellectual property (IP) entrepreneurs if they can shift focus as opportunities arise.
Samplify Systems is a fabless analog company that has evolved from an IP supplier to a chip vendor to a provider of complete reference designs (see “Hardware Algorithm Fine-Tunes Converters For Best Compression” at www.electronicdesign.com).
Back in 2004, with the algorithm instantiated in Samplify’s IP, designers could stream compressed data at rates significantly below the data converter’s sampling rate, as well as store more compressed data in the same memory space.
Also, storage-scope users could manually trade off between capture duration and signal quality. Implementing the new compression algorithm in silicon only added a few cents to the manufacturing cost of analog-to-digital converters (ADCs) and digital-to-analog converters (DACs).
The IP was (and is) a patented lossless compression engine for band-limited signals. The lossless compression ratio is reduced as the occupied signal bandwidth increases, but 2:1 lossless compression is often possible. The center frequency of the signal of interest can be anywhere from dc to Nyquist.
Continue on next page
For a 20% bandwidth-limited signal with a 75-dB signal-to-noise ratio (SNR), Samplify’s algorithm yielded roughly 2:1 lossless compression. The amount of compression be further increased by analyzing the noise floor and headroom requirements of the signal.
Obviously, a lot of that was analytical and not yet demonstrated in a commercial product. So, fast-forward to 2008 when Samplify had seen the necessity of moving away from the IP-licensing model for revenue generation to becoming a fabless semiconductor supplier (see “Scrappy IC Startup Challenges Big Guns With Lossy-Compression ADC For Ultrasound”).
Back then, with Texas Instruments and Analog Devices duking it out for dominance in the front ends of the world’s ultrasound medical devices, it took audacity for a fabless startup to design its first product for the same market. But that’s what Samplify did.
Samplify’s SAM1600 family of 16-channel, 12-bit, 65-Msample/s ADCs integrated the company’s Prism data-compression technology. The company also offered full 16-channel versions and versions that shared the ADCs between clock cycles.
The SAM1600 devices came in a 12- by 12-mm ball-gird array (BGA) and consumed 44 mW per channel. The Prism technology reduced the number of low-voltage differential signal (LVDS) I/O pairs, and thus the power they consume, by up to 75%. While their uses focused on ultrasound, potential applications included 4G wireless basestations, automated test equipment, and radar/sonar receivers.
To crack the ultrasound market, Samplify had been working with FPGA vendors on the decompression side, so the FPGA vendors already have Samplify decompression blocks in their libraries.
At introduction, samples of the ADC cost between $39.50 and $79.00, in 1000-unit quantities, depending on features. A complete ADC evaluation board with data analysis software was available for $1995.00. Now, two years later, the evaluation board has evolved into a complete reference design, targeting low-end, high-volume, portable ultrasound devices (Fig. 1) for medical professionals in developing countries.
The new product is being offered as a development kit. But in practical terms, any Asian original device manufacturer (ODM) that chooses to adopt the “kit” can easily turn it into the electronics module for a complete pulse and Doppler (blood-flow) ultrasound diagnostic device.
According to Allan Evans, Samplify’s vice president of marketing, the enabling Prism compression technology has been augmented to produce Prism CT (computed tomography) and Samplify’s AutoFocus beamforming technology, implemented in a separate chip. That work was carried out by Samplify employees, including founder Al Wegener, who created the original compression IP, and outside experts in beamformng (see “As Compression Technologies Reach Their Limits, What’s Next?”).
The ultrasound market is fragmenting into specialty applications other than traditional radiology and OB/GYN, Evans said, and that product differentiation is all in the image processing algorithms. The value proposition for Samplify is that it is enabling ultrasound manufacturers to focus R&D on their core image processing algorithms while leaving the collection of images to the Samplify hardware and software.
Samplify’s SMK9130, a 64-channel ultrasound front-end platform, incorporates two 32-channel ultrasound analog front-end (AFE) receiver modules for full continuous-wave Doppler (CWD) (Fig. 2) capability. The company’s AutoFocus beamforming technology and QuadBeam phased array processing combine the received signals from each channel to provide four scan-line outputs for each transmission.
Continue on next page
Tri-level pulsers and a transmit beamformer support transmission for all black-and-white and color modes. There is also a power supply for the high-voltage path. Output is via USB 2.0 or 4x PCI Express 1.1. The kit costs $60,000, with additional pricing options to support volumes for clinical trials and production.
Underlying all this is the question of how much designers should trust lossy compression of their diagnostic medical images. Going back to Wegener’s original algorithm, the point, from an information theory standpoint, was to take advantage of what they knew about the signals—they aren’t totally random—and use that to steer the compression.
More concrete reassurance can be found in an International Society for Optics and Photonics (SPIE) paper called “Real-time compression of raw computed tomography data: technology, architecture, and benefits” by Wegener, Herfkens, et al, available at spie.org/x648.html?product_id=810599. Among other things, it describes double-blind testing in which they showed images to practicing radiologists (see “Compress CT Samples At 64 Gbits/s” at www.electronicdesign.com).
The lesson for other analog designers in Samplify’s story is simple: leverage, focus, and guts. At first, Wegener had an algorithm that could magically compress certain kinds of data losslessly. Four years later, he’d seen what the big analog semi companies were doing in medical electronics and challenged them with a trio of chips and FPGA algorithms. Two years after that, he has enough credibility with Chinese ODMs to be in the chip business with unique hardware that would be darned hard to reverse-engineer.
Samplify’s journey reflects the direction the whole industry is going: ever-higher mixed-signal integration, specilization with respect to applications, and close engagement with customers.