A tale of two digital-video converters
Software versus hardware compression defines the difference.
By Brian Dipert, Senior Technical Editor -- EDN, February 19, 2009
Assessing the tradeoffs between implementing video-processing tasks in hardware on function-optimized silicon, in software running on a general-purpose microprocessor, or (as an intermediary approach between these two extremes) in software running on a multimedia-tuned DSP has long fascinated me. Unique application criteria in each particular design situation mean that no one approach is optimum in every possible case. Consider, for example, the endpoints of the spectrum of possible implementation options. Function-optimized hardware is often the lowest-power-consuming approach, and it can also be the most cost-effective. Conversely, you're out of luck unless you wish to implement exactly the algorithms frozen into silicon, as compared with the infinite flexibility of code executing on a CPU. And the relative cost argument is a moot point if the system you're designing is intended to be (for example) a peripheral to a computer already containing a beefy x86 µP, versus a standalone system.
This particular Prying Eyes project is emblematic of these pros and cons. It compares and contrasts two Plextor-branded products, both purchased on closeout, with similar advertised functions. The less expensive unit even offers a TV tuner absent from its pricier peer. Plextor's ConvertX PX-TV100U cost $8.99 after rebate ($38.99 before, $99 original MSRP), while the PX-M402U was $29.99 after rebate ($99.99 before, $159 original MSRP). The difference between the two centers is a software-versus-hardware compression emphasis. The PV-TV100U relies on the CPU horsepower of a USB2-tethered PC to handle the hefty audio-plus-video lossy compression duties, whereas the PX-M402U harnesses its integrated WIS Technologies G07007SB streaming-media encoder for these meaty tasks. The G07007SB natively supports multiple video inputs; conversely, with the PV-TV100U, input switching is handled by two Philips (now NXP Semiconductor) 74HC4052 analog multiplexer/demultiplexers (one on each side of the PCB).
Speaking of inputs, both devices need to translate incoming analog multimedia information into digital equivalents. With the PX-TV100U, audio and video conversion are respectively the domains of Micronas'MSP 3425G (PDF) sound processor and Texas Instruments'5150AM1 video decoder. The PX-M402U, conversely, relies respectively on AKM Semiconductor's5355 (PDF) dual-channel 16-bit audio ADC and NXP Semiconductor's SAA7115 video decoder. Whereas the SAA7115 comprehends NTSC, PAL, and SECAM video standards, the video decoder in Plextor's PX-M401U (the PX-M402U's predecessor) supported only NTSC. Another key difference between the PX-M401U and follow-on PX-M402U is that the latter, later device was explicitly certified by DivX. I suspect, however, that both products employed the same WIS G07007SB compression engine, which supports MPEG-4, MPEG-2, MPEG-1, Motion JPEG, and H.263 video codecs. The DivX format, after all, is a combination of MPEG-4 video and MP3 (MPEG-1 Audio Layer 3) audio bit streams bundled together by a proprietary “wrapper.”
On the other end of the processing chain, raw audio and video conveyance from the PV-TV100U to a PC is the responsibility of Empia Technology'sEM2860 USB2 media-capture processor, whereas the USB2 transport of already-compressed multimedia data from the PX-M402U falls to Cypress Semiconductor'sCY7C68013 embedded microcontroller. Both products include serial EEPROMs; an Atmel24C01A 1-Mbit device in the case of the PX-M402U and a Catalyst Semiconductor (now owned by ON Semiconductor) 24WC02 2-Mbit memory for the PX-TV100U. The PX-M402U also embeds a 64-Mbit PC100 SDRAM, Micron Technology's 32-bit interface 48LC2M32B2. The not-shown backside of the PX-M402U PCB, by the way, is devoid of ICs; all you'll find there (aside from traces and solder points) are a scattered few passive components, such as capacitors, diodes, inductors, and resistors.
When examining a piece of gear, I always enjoy seeing how it's been intentionally made adaptable by its designers to support multiple product proliferations spawned from a common hardware foundation, and differentiated by minor peripheral variations and unique software builds. Flexibility within the PX-M402U design is exemplified both by unpopulated primary PCB sites and by sufficient spare space in the enclosure for a supplemental PCB-housed TV tuner to transform the unit into a PX-TV402U. Reflective of this malleability, the unit's rear panel is easily removable, therefore replaceable by one incorporating a coaxial cable connector. Some of the unpopulated primary PCB sites are, I suspect, intended for primary-to-secondary PCB bridging purposes.
Note, too, the unused areas on the PX-M402U PCB intended for a second set of analog audio connections and both composite and S-video ports. Inputs or outputs; who knows? Plextor never brought to production a variant of the PX-M402U that employed them; I suspect the company leveraged a generic WIS Technologies reference design. Similar flexibility can be found in the PX-TV100U, which includes a LG Electronics “can” tuner. Plextor also sold a tuner-less variant of this same design called the PX-AV100U.
Because the PX-M402U and PX-TV402U don't rely for compression duties on the system they're connected to over USB2, they're more adept at supporting various tethered equipments' operating systems, as compared with the Windows-only PX-TV100U. They quickly became, for example, PVR hardware favorites of Linux users. Specifically, it's possible to transform a Sony PlayStation 3 into a PVR, by virtue of the game console's built-in USB ports and its capability for running Yellow Dog Linux and other open-source distros.
1. Plextor’s PX-TV100U relies on the CPU horsepower of a USB2-tethered PC to handle the hefty audio-plus-video lossy-compression duties. The PX-M402U harnesses its integrated WIS Technologies G07007SB streaming media encoder for these meaty tasks.
2. Raw audio and video conveyance to a PC is the responsibility of the PV-TV100U’s Empia Technology EM2860 USB2 media-capture processor. The USB2 transport of compressed multimedia data from the PX-M402U falls to Cypress Semiconductor’s CY7C68013 embedded microcontroller.
3. The PX-M402U design offers flexibility in both primary PCB (printed-circuit-board) unpopulated sites and sufficient enclosure space for a supplemental PCB-based TV tuner that would transform the unit into a PX-TV402U. Similarly, although the PX-TV100U includes an LG Electronics “can” tuner, Plextor also sold the PX-AV100U, a tunerless variant of this design.
4. Both devices need to translate analog multimedia information into digital equivalents. With the PX-TV100U, audio and video conversion are the domains of Micronas’ MSP 3425G sound processor and Texas Instruments’ 5150AM1 video decoder, respectively. The PX-M402U, conversely, relies on AKM Semiconductor’s 5355 dual-channel, 16-bit audio ADC and NXP Semiconductor’s SAA7115 video decoder. Whereas the SAA7115 comprehends NTSC (National Television System Committee), PAL (phase-alternating-line), and SECAM (Séquentiel Couleur Avec Mémoire) video standards, the video decoder in the PX-M401U (the PX-M402U’s predecessor) supported only NTSC.
5. Both products include serial EEPROMs: a 1-Mbit Atmel 24C01A device in the PX-M402U and a 2-Mbit Catalyst Semiconductor 24WC02 memory for the PX-TV100U. The PX-M402U also embeds a 64-Mbit, PC100 SDRAM, Micron Technology’s 48LC2M32B2 with a 32-bit interface.
-
I read Government Coupon (w.r.t. converter boxes) where it said Rebate. Just highlight what I said, right-click, and select Translate> "Social Monarchic >> Austrian-Economic" please.
Steve Nordquist - 2009-13-3 14:39:00 PDT -
So this was the article where you reminded me of the time the Bush Administration paid you to get programmable ATSC tuners to encourage your participation in ATSC communication as a citizen. You ended up getting things that thinned channels and ''sidebands,'' compressing the signal in a lossy way, and one of them ate CPU cycles.
Oddly the device did not even act as a much-discussed relay station with functions allowing appending streams on-channel and adding metadata. There was nothing about how you tied in cutdown processors and GbE cameras or encoded various sound streams, sidebands, etc.
My memory is not going to get any better on this thing...was it really about gadgets to put VHS stuff on optical media?
Steve Nordquist - 2009-13-3 14:30:00 PDT -
The converter box is to convert DIGITAL into ANALOG signal NOT THE OTHER WAY around as your artical states !!!
gene roberts - 2009-25-2 13:50:00 PST -
Nice job, Brian, but we haven't seen the last of the DTV transition delays. June 12 won't be the transition date. Also, we should follow President Obama's suggestion and read instead of watch TV.
Martin - 2009-25-2 12:07:00 PST -
I have some history with video capture myself. At one time I used a P3 and an ATI all-in-wonder. The best it could do was 352x240 due to the limits of the P3 CPU. If I tried to do DVD (720x480) it would just hang up. The P3 was fast enough to record DV from a fire-wire port at DVD resolution. I think DV is slightly compressed but not like MPEG.
Then I got an ADS DVD Express, $45 to $70 . With a Cypress chip doing the video compression, even a P3 has no problem capturing DVD resolution. The HDD light just blinks about every second.
Using the CPU for compression is common because it make the hardware cheap and software can be revised long after the sale. But CPU compression will brick-wall at some resolution. Even if the CPU is fast enough, any CPU will occasionally be distracted by other tasks and miss frames.
I have a couple HDTV cards, Pinnacle and Hauppauge. Neither work 100%. It may be due to my lack of CPU power but in both cases the UI is a mess and with each software update they work a bit better. By contrast, the tuner in my JVC DR-MV100B DVD recorder works very well. I don’t actually use it as a DVD or VHS recorder, just as a clear QAM tuner for an analog TV. I could use another one for my digital TV so that I could switch from cable to OTA instantly instead of a half hour channel scan. You can buy lots of digital TV tuners but none of them support clear QAM and none of them have a HDMI output. Unfortunately my use model and the typical marketing model do not match. 8-/ Actually I think HDMI will either have to get cheaper or be replaced by something like LAN. $40 for a cable is just too much.
I now have an AVCHD camera but my Pentium D 805 (dual 2.6GHz) is just too slow to even play full HD files. AVCHD requires a CPU with a Passmark benchmark of about 1000 or more. That means a “Core 2” 65 micron CPU or better. The SD card slot also has to be HC compatible, which my computer’s is not, although it’s cheap and easy to buy a HC compatible reader such as my $6 Ativa
Toshiba uses a media co-processor in some of their notebooks and some nVidia cards provide co-processor for the same reason. However fast the CPU, HD video requires dedicated hardware.
Cheers,
Steve
Steve Ungstad - 2009-19-2 15:14:00 PST
