Some questions for RussSchultz regarding Sigmatel chips.
- Thread starter PC-Engine
- Start date
Aha! I'm back!
First off, it wouldn't suprise me that the new generation (35xx) has slightly worse theoretical maximum SNR than the older generation (34xx). From what I remember from meetings, etc, there was ongoing development on the analog side of the design that removed external components (reduced cost) and kept the SNR very high. (95dB vs 90dB is pretty similar and well above any compressed audio content)
Next, there's a lot of ways to fudge SNR, and a lot of different measurement techniques that give different values. We've had a lot of trouble with customers (and potential customers) trying to beat us up on cost based on competitors numbers, so we publish our and test methodology in the marketting glossies.
Not suprisingly, the PCB layout has a big impact on the audio quality and stability of the device, at least for our part. As such, we work very closely with our customers (that will let us) to ensure a layout that gives them as close to the theoretical performance as possible.
Suffice to say, there aren't any (I believe) integrated solutions (ones that don't require an external DAC) that beats our numbers. I can't remember the exact numbers, but ours has beaten any of our competitors in our in house testing. While I'm obviously biased, I've seen the testing methodology and it wasn't an exercise in finding the "perfect benchmark" that let us beat our competitors. We are better than our competitors products in real world results.
Finally, I believe the 'real world' numbers, on a typical player, are about 86dB. This is measured on a 1kHz, 0dB sine wave, I believe.
Ones I know of that are public knowledge (and recognizable) are SonicBlue , Philips, and Creative (excepting large HD players).
And I think you're confused about iRiver. I believe all of their products on the shelf (in the US, at least) use the Philips Symphony chip.
My personal favorite is the iBead, from EXATelecom, Korea. They never had much luck finding a US distributer for some reason, but their size/form factor and firmware useability were just what I was looking for. Also they did a fantastic job with the layout and build quality (in my opinion). These were the products we purchased to give out as freebees as they highlighted the market that we excelled in--small form factor and high quality. The only non grey market place to buy it in the US is http://www.hammacher.com/publish/70548.asp?promo=el_audio, though I've seen it on ebay quite a bit.
SonicBlue has done most of their own software, and they're all smart, top notch engineers. They however, aren't USB mass storage class, so you need drivers to connect to a computer.
NAND flash comes in flavors and different flavors have different characteristics. Even the same flavor has different capabilities when it comes from different manufacturers. Some have internal caching mechanisms, some have larger erase blocks, some require more error correction than others, etc.
Generally, we've grouped NAND flash into 3 (and now more) groups:
Type I -- older (<128 MB flash chips). Erase block size of 16KB, page size of 512 bytes.
Type II -- newer (all 128MB chips or bigger). Erase block size of 128KB, page size of 2KB, most have internal caches.
Type III -- newer, low cost (128MB or bigger). Smaller erase block size, longer erase times per block. (This stuff is beastly slow)
Assuming one erase per block, and one write per page, the theoretical maximums on Type I NANDs is about 12 MB/s. Type IIs is slightly higher (approaching 18MB/s, I think), Type IIIs are horrifically slow (but apparently 20% cheaper per bit).
Of course, windows USB mass storage host driver doesn't always help us out, so I think our current max speeds are about 7MB/s read, and 5MB/s write for our best base. Even better, each of the operating systems have slightly different implementations of USBMSC, so the numbers change quite a bit.
Writing to an MMC/SD card is mostly dominated by the SPI bus. I think its theoretical max is 2MB/s or something like that.
Truthfully, this is the area where I'm not so strong.
First off, it wouldn't suprise me that the new generation (35xx) has slightly worse theoretical maximum SNR than the older generation (34xx). From what I remember from meetings, etc, there was ongoing development on the analog side of the design that removed external components (reduced cost) and kept the SNR very high. (95dB vs 90dB is pretty similar and well above any compressed audio content)
Next, there's a lot of ways to fudge SNR, and a lot of different measurement techniques that give different values. We've had a lot of trouble with customers (and potential customers) trying to beat us up on cost based on competitors numbers, so we publish our and test methodology in the marketting glossies.
Not suprisingly, the PCB layout has a big impact on the audio quality and stability of the device, at least for our part. As such, we work very closely with our customers (that will let us) to ensure a layout that gives them as close to the theoretical performance as possible.
Suffice to say, there aren't any (I believe) integrated solutions (ones that don't require an external DAC) that beats our numbers. I can't remember the exact numbers, but ours has beaten any of our competitors in our in house testing. While I'm obviously biased, I've seen the testing methodology and it wasn't an exercise in finding the "perfect benchmark" that let us beat our competitors. We are better than our competitors products in real world results.
Finally, I believe the 'real world' numbers, on a typical player, are about 86dB. This is measured on a 1kHz, 0dB sine wave, I believe.
I know we have about 200 individual companies (mostly out of China/Taiwan/Hong Kong) using our chips. Everytime I see a player advertised, I have to go to the website and find their drivers/firmware update application to see if its ours--I just can't keep track of the many baziliions of players and customers we have.
Ones I know of that are public knowledge (and recognizable) are SonicBlue , Philips, and Creative (excepting large HD players).
And I think you're confused about iRiver. I believe all of their products on the shelf (in the US, at least) use the Philips Symphony chip.
Technically, I don't know the answer. I don't get exposed to that side of the business much, and even then we only test our customers products when they have issues that need to be resolved.
My personal favorite is the iBead, from EXATelecom, Korea. They never had much luck finding a US distributer for some reason, but their size/form factor and firmware useability were just what I was looking for. Also they did a fantastic job with the layout and build quality (in my opinion). These were the products we purchased to give out as freebees as they highlighted the market that we excelled in--small form factor and high quality. The only non grey market place to buy it in the US is http://www.hammacher.com/publish/70548.asp?promo=el_audio, though I've seen it on ebay quite a bit.
SonicBlue has done most of their own software, and they're all smart, top notch engineers. They however, aren't USB mass storage class, so you need drivers to connect to a computer.
Those numbers aren't completely correct (I'll address our current numbers later on), but the biggest factor on high speed USB download speed is the type of flash.
NAND flash comes in flavors and different flavors have different characteristics. Even the same flavor has different capabilities when it comes from different manufacturers. Some have internal caching mechanisms, some have larger erase blocks, some require more error correction than others, etc.
Generally, we've grouped NAND flash into 3 (and now more) groups:
Type I -- older (<128 MB flash chips). Erase block size of 16KB, page size of 512 bytes.
Type II -- newer (all 128MB chips or bigger). Erase block size of 128KB, page size of 2KB, most have internal caches.
Type III -- newer, low cost (128MB or bigger). Smaller erase block size, longer erase times per block. (This stuff is beastly slow)
Assuming one erase per block, and one write per page, the theoretical maximums on Type I NANDs is about 12 MB/s. Type IIs is slightly higher (approaching 18MB/s, I think), Type IIIs are horrifically slow (but apparently 20% cheaper per bit).
Of course, windows USB mass storage host driver doesn't always help us out, so I think our current max speeds are about 7MB/s read, and 5MB/s write for our best base. Even better, each of the operating systems have slightly different implementations of USBMSC, so the numbers change quite a bit.
Writing to an MMC/SD card is mostly dominated by the SPI bus. I think its theoretical max is 2MB/s or something like that.
