Nah, we made money on all of them; the real money is in licencing through existing manufacturers though.
ARM doing a core that did MP3/WMA? No, this isn't true; the ARM7TDMI core was there all along - and the solution is 100% software in the rio 600/800 and nike players (plus the rio volt, intel concert, etc etc).
In fact, the Rio Receiver we also did at empeg uses the same CPU as the rio 600/800, but with a DRAM interface.
The same software decoder core is used in the empeg and the rio. We never wrote any mp3 decoders, we just did the surrounding stuff - which is a lot more complex than an mp3 decoder!
Anyway, you'll see what the empeg team has been up to all this time in a few weeks:)
...was using the cassette relay to do pulse-dialling when on holiday in portugal. The house we were staying in had a lock on the phone dial (and no DTMF back then!) so I broke out the line, hooked it through the cassette relay and played with timings in basic to get the pulse lengths right.
Unfortunately not; the driver isn't really a winmodem driver:(
The mwave modem appears to be a DSP-based solution (coltrollerless modem?) and a lot of the source is some very non-open-source binary DSP files that get uploaded to the onboard DSP. Seeing as how v34 (etc) are all pretty tied up in patents & licencing, this explains the release. They've released a soft-ish modem driver that will work with some machines. It's not a generic AC'97 "let's do DSP stuff on the actual PCM audio data" modem, and as such isn't a huge amount of use for many people:(
It's great if you have an IBM 600E laptop, though. I've got a Vaio...
This is called "universal plug'n'play" - if DHCP times out, it will still carry on DHCP'ing (in case a server appears) but will ARP random addresses in the 169.254 class B, pick a free one and sit on it. See www.upnp.com - and yes, it's MS sponsored but an open multiparty standard.
It's done in MacOS 8.5+, win98, win2000, and we did some stuff on this for the embedded linux in the Rio Receiver (mp3 thin-client - www.riohome.com). Source comes with the box, though we suspect such hackery wouldn't get into the main kernel tree;)
You can't compare the Pentium to the new SA - the die is many, many times smaller on the SA which means the yields will be much better.
It should also be a lot cheaper, but it won't be as fast as a PIII at 1Ghz.
Still, it *will* give the Transmeta stuff a run for its money - 1GHz at 1.5W is pretty good, and SA's run linux very nicely (as empeg owners, netwinder owners, and iPaq owners know)
The problem is, in the commerical world the product is driven by tight deadlines and getting the product out before you get eaten alive by your competitors (who are also doing the same thing).
If your company took the time to write very stable, near-bug-free code, they'd take so long doing it they'd go out of business - their competitor would get the business with a flakey but shipping product and by the time you turned up with your perfect product, everyone would be locked into their stuff (and most likely would have been using it for a couple of years).
Noone wants to write buggy code, we all try to do our best; logical & clear design, defensive programming & good documentation give a good base. Peer review and experience (been there, don't want to do that again) help a lot too. Just writing the comments first (saying what you're *going* to do before doing it) helps.
Another problem is that writing bug free apps on (say) windows is almost pointless as the app will still fall over when some bit of buggy OS/windows API code falls over. Things have to be stable and bug-free from the hardware upwards to give an impression of stability to the user - the problem is, the average user can't tell the difference (and couldn't give a toss) whether the app or the os fell over, it's just "my WP crashed and I lost my work".
Welcome to the real world. Software can be flakey because it was written to be useful before the hardware went out of date - not exactly a problem with the shuttle. You can spend ages hand-crafting efficient code to be overtaken by crap code on a faster CPU. Blame the chip companies for moving so quickly:)
The main problem is 4xUSB - the SA11[01]0 has a companion chip with USB host, the SA11[01]1. Only a single one, you'd then need a built in hub to give you more ports.
Multiple ethernet isn't hard either, but remember that the SA11[01]0 has *no external DMA* and so everything is done with programmed I/O. Systems based on the SA110, like the Netwinder, have a PCI bus and so you can get DMA to peripherals.
The problem is, unless you're building in *really* huge quantities (10k+) you'll find that making a StrongARM based desktop machine costs more than a more powerful Celeron/K6/whatever machine. One-off, the SA1100 at 220Mhz is more expensive than a K6-2/500Mhz. Ok, it has a lot of on-board I/O and uses no power at all, but these aren't compelling reasons to use the ARM in a system with a mains cable attached!
The HP Jornada 820 has a SA1100 in it too, but noone has tried getting linux on this as far as I know. It'll run for 10+ hours whilst powering a colour LCD too.
Hugo empeg (we use the SA1100 in the empeg in-dash MP3 player - which runs linux too and you can make your own apps for it)
No, the Ford Focus player doesn't play mp3-cds. Visteon have shown a prototype indash player which plays MP3 CDs though (CES, etc) and also an MP3-changer which uses HDD technology - we developed this unit for them.
It wasn't actually ethernet, but it was the first multiple-access network. ISTR that it didn't actually use CSMA/CD (carrier sense multiple access - ie, you don't transmit if you can hear someone else) / (collision detect - if someone else collides with me I know about it).
I think it used time slots to regulate access, and outstations couldn't talk to each other, only to the central station (which controlled the sequencing, and would perform routing).
Remember, there are no linear amps in the equation - they're not going to be bleeding all over adjacent channels & that sort of thing. Burners got a bad name because most were made with no respect for quality.
They're getting more range by concentrating the power into a beam - as opposed to the normal omnidirectional spread. The antennas they use have great gain in one direction, and sod all in others - you won't even hear them 5 degrees off-line.
There's nothing illegal about this - lots of other (read: more expensive) 802.11 units have been around for years sold with a pair of yagis for the express purpose of linking two distant sites.
I'm probably completely wrong, but from memory the maximum gain allowable for the entire link (both ends) was 24dB. Full kits with directional antennae (don't forget the lightning surge arrestor if you're mounting antenna where they might get touched by the hand of zeus) are available, but they usually cost.
Problem is, the normal "budget" 802.11 cards were crippled by just having an internal antenna and no place to connect a "real" one. Nice to see apple have done the right thing:)
Hugo (just running out to buy a couple of airports to link the empeg mk2 in my car to the office;) )
Ok, I don't like it particularly, but most (all?) of Europe has VAT - Value Added Tax. It's just sales tax on everything with a few exceptions (in the UK at least - books, basic foods, sanitary products, etc). You get charged it on everything you buy from anywhere, and it's a uniform rate for almost all goods (electricity & other utilities an exception at the moment). The UK has VAT at 17.5%.
If you import stuff from outside Europe, you get charged VAT by customs & excise (as well as import duty). If you buy something from inside Europe but outside your country, you pay VAT at source - ie, where you got the stuff from. Businesses get to claim back their VAT, but it's still charged when (eg) a distributor buys a whole load of PCs from HP.
The point is - it's uniform. The playing field is fairly level and will get more level as the countries in the EU move their VAT levels into agreement. It's a slight headache for businesses, but hey - if you run a business you have a pile of paperwork to deal with anyway. Noone has to worry about net businesses having an unfair advantage (apart from having virtual premises and often virtual stock) - you can't avoid the VAT.
If the US took up such an option, each business would have to account for sales to other states (like you have to do when selling to other countries in the EU) and then it all would get balanced up by the federal government.
Errr, I don't think so. If you run MP3 at the sort of bitrates that ATRAC currently runs at (circa 256kbit) MP3 sounds at least as good.
In blind listening tests with trained listeners at a recent perceptual audio coding conference, 192kbit mp3 was statistically undistinguishable from the source cd. Going to 256k only hammers the point in more - mp3 is good, but isn't at its best at 128k.
Problem is, you've probably not heard MP3s though good equipment, whereas MD players are mature products which have to compete on sound - most mp3 players compete on price alone.
However, you don't need any of that to make a SA1100 based system. The empeg is SA1100 based. We didn't use any development boards, we just wired up the cpu and it worked - as the SA1100 is so highly integrated, there's only *one* way to wire ram chips to it, *one* way to wire a flash chip to it and *one* way to hook up serial ports and LCD. There's nothing magic about the itsy which anyone competent enough to fabricate & build an itsy from the designs couldn't simply do themselves. I mean, if they gave you the Orcad files for it, what do you plan to do - buy orcad? The schematic editor alone costs over $1000. The problem is, building computers simply isn't hobbyist compatible anymore. 0.5mm pitch SMT components? You need a binocular microscope really, even if you can mount them chips using a fine tipped soldering iron, SMT solder (silver loaded stuff), and lots of SMT rework flux. That's not even considering how hard it is to *get* the parts - tried getting EDO ram? 3.3v stuff? In small quantities? You can't get these bits (including the SA) just off the shelf, though you can get most of the discretes through some of the better (catalogue-publishing) distributors such as Farnell. PCBs? You're looking at maybe $1000 minimum for one board set on long turnaround. Compaq have obviously been asked by a lot of universities for prototypes, so they've released the designs so research departments can make a few and do software development. It's not a hardware development platform - it's too small to prototype new stuff on. Hugo empeg
StrongARM is *already* much lower power consumption than the Crusoe 3120: some of the Crusoe figures are quite funny. Looking at the 3120 datasheet, it takes a mere (!) 1.4W whilst playing mp3's.
Compare this to around 0.075W (30% usage, CPU static when not doing decode) for a StrongARM doing MP3 decode. Hey, even *flat out* a 200Mhz, 1.5v core StrongARM on a 3-year old 0.25 micro process takes less than 0.25W. Just wait for the SA-2 stuff coming out this year - they're claiming more than 2x the performance whilst using less power.
Transmeta pushing mobile linux (and OEMs making 1st-gen designs around it, application support, etc) will probably ultimately benefit *real* low-power chips like the ARM, as linux applications (to a large extent) are pretty architecture-agnostic.
Looking at embedded machines, the Crusoe CPU also needs a lot of support circuitry. As an example, the SA1100 includes (on chip) a PCMCIA interface, real time clock, TFT/passive LCD drive, sound codec port, MMU, DRAM controller, USB slave, IrDA (SIR/FIR), sync serial port, a couple of async serials, and plenty of GPIO pins. Heck, it's almost a webpad on its own - which is why the Itsy can be so small, so cute, and still run off a couple of AA cells while playing.mov's. Add the 1101 companion and you get VGA, PS/2 mouse/kb, USB host and more besides.
The crusoe is impressive as an x86 chip. Once you leave that narrow (in embedded terms) slot, the performance doesn't really measure up to offerings from ARM, MIPS, etc.
Errr, the Cirrus 7209 (ARM7T core) CPU uses only 0.087W when *playing* mp3's, and *less than* 0.001W when in standby.
The Transmeta chips are low power compared to other x86's, but not low power when compared to the best of other architectures in terms of mips/w (eg, StrongARM, ARM10, etc).
If you read the site, it uses an internal rechargeable battery to give 10 hours of life. The HDD only runs once every 10 minutes, for (my guess) around 15 seconds.
Completely the wrong comparison. The smaller the drive, the lower mass the heads, which means the higher the shock rating. Hence why the IBM drives have an *operating* shock rating of 150g.
We use similar 2.5" drives in the empeg, and it's been through full vibration testing - we also keep them in a shock-mounted cradle though.
USB slave (as required on a device like this) is not too hard, really. Nothing like the complexity of a USB host (as is currently in 2.3 kernels). The driver I did for the empeg is in the kernel source, but it's not very useful unless you use a USBN9602... which is found in zero PCs.
Quite possibly, 64Mb of DRAM in an active state (ie, you need to be accessing *some* of it) will take more power than nothing, which is what a HDD which has had the power removed takes.
They won't just be putting the HDD into standby - they still take a non-insignificant amount of juice for a battery-powered player.
Slashdot Mirror
Nah, we made money on all of them; the real money is in licencing through existing manufacturers though.
:)
ARM doing a core that did MP3/WMA? No, this isn't true; the ARM7TDMI core was there all along - and the solution is 100% software in the rio 600/800 and nike players (plus the rio volt, intel concert, etc etc).
In fact, the Rio Receiver we also did at empeg uses the same CPU as the rio 600/800, but with a DRAM interface.
The same software decoder core is used in the empeg and the rio. We never wrote any mp3 decoders, we just did the surrounding stuff - which is a lot more complex than an mp3 decoder!
Anyway, you'll see what the empeg team has been up to all this time in a few weeks
Hugo
empeg
...was using the cassette relay to do pulse-dialling when on holiday in portugal. The house we were staying in had a lock on the phone dial (and no DTMF back then!) so I broke out the line, hooked it through the cassette relay and played with timings in basic to get the pulse lengths right.
:)
Free phone calls
Hugo
Unfortunately not; the driver isn't really a winmodem driver :(
:(
The mwave modem appears to be a DSP-based solution (coltrollerless modem?) and a lot of the source is some very non-open-source binary DSP files that get uploaded to the onboard DSP. Seeing as how v34 (etc) are all pretty tied up in patents & licencing, this explains the release. They've released a soft-ish modem driver that will work with some machines. It's not a generic AC'97 "let's do DSP stuff on the actual PCM audio data" modem, and as such isn't a huge amount of use for many people
It's great if you have an IBM 600E laptop, though. I've got a Vaio...
Hugo
Actually, the first *stored program* computer was binary - the Manchester "Baby":
http://www.computer50.org/
This used CRT tubes to provide the RAM - refresh involved a photosensor on the surface of the CRT feeding back to the CRT itself. Very cute.
Hugo
This is called "universal plug'n'play" - if DHCP times out, it will still carry on DHCP'ing (in case a server appears) but will ARP random addresses in the 169.254 class B, pick a free one and sit on it. See www.upnp.com - and yes, it's MS sponsored but an open multiparty standard.
;)
It's done in MacOS 8.5+, win98, win2000, and we did some stuff on this for the embedded linux in the Rio Receiver (mp3 thin-client - www.riohome.com). Source comes with the box, though we suspect such hackery wouldn't get into the main kernel tree
Hugo
empeg
You can't compare the Pentium to the new SA - the die is many, many times smaller on the SA which means the yields will be much better.
It should also be a lot cheaper, but it won't be as fast as a PIII at 1Ghz.
Still, it *will* give the Transmeta stuff a run for its money - 1GHz at 1.5W is pretty good, and SA's run linux very nicely (as empeg owners, netwinder owners, and iPaq owners know)
Hugo
The problem is, in the commerical world the product is driven by tight deadlines and getting the product out before you get eaten alive by your competitors (who are also doing the same thing).
:)
If your company took the time to write very stable, near-bug-free code, they'd take so long doing it they'd go out of business - their competitor would get the business with a flakey but shipping product and by the time you turned up with your perfect product, everyone would be locked into their stuff (and most likely would have been using it for a couple of years).
Noone wants to write buggy code, we all try to do our best; logical & clear design, defensive programming & good documentation give a good base. Peer review and experience (been there, don't want to do that again) help a lot too. Just writing the comments first (saying what you're *going* to do before doing it) helps.
Another problem is that writing bug free apps on (say) windows is almost pointless as the app will still fall over when some bit of buggy OS/windows API code falls over. Things have to be stable and bug-free from the hardware upwards to give an impression of stability to the user - the problem is, the average user can't tell the difference (and couldn't give a toss) whether the app or the os fell over, it's just "my WP crashed and I lost my work".
Welcome to the real world. Software can be flakey because it was written to be useful before the hardware went out of date - not exactly a problem with the shuttle. You can spend ages hand-crafting efficient code to be overtaken by crap code on a faster CPU. Blame the chip companies for moving so quickly
Hugo
The main problem is 4xUSB - the SA11[01]0 has a companion chip with USB host, the SA11[01]1. Only a single one, you'd then need a built in hub to give you more ports.
Multiple ethernet isn't hard either, but remember that the SA11[01]0 has *no external DMA* and so everything is done with programmed I/O. Systems based on the SA110, like the Netwinder, have a PCI bus and so you can get DMA to peripherals.
Hugo
empeg
(cough)
;) )
www.empeg.com
SA1100, Linux, MP3, ethernet, USB, up to 64Gb internally, voice recognition, graphic screen, 4ch out with parametric eq, etc etc.
Hugo
(from empeg
The problem is, unless you're building in *really* huge quantities (10k+) you'll find that making a StrongARM based desktop machine costs more than a more powerful Celeron/K6/whatever machine. One-off, the SA1100 at 220Mhz is more expensive than a K6-2/500Mhz. Ok, it has a lot of on-board I/O and uses no power at all, but these aren't compelling reasons to use the ARM in a system with a mains cable attached!
The HP Jornada 820 has a SA1100 in it too, but noone has tried getting linux on this as far as I know. It'll run for 10+ hours whilst powering a colour LCD too.
Hugo
empeg
(we use the SA1100 in the empeg in-dash MP3 player - which runs linux too and you can make your own apps for it)
No, the Ford Focus player doesn't play mp3-cds. Visteon have shown a prototype indash player which plays MP3 CDs though (CES, etc) and also an MP3-changer which uses HDD technology - we developed this unit for them.
Hugo
empeg
It wasn't actually ethernet, but it was the first multiple-access network. ISTR that it didn't actually use CSMA/CD (carrier sense multiple access - ie, you don't transmit if you can hear someone else) / (collision detect - if someone else collides with me I know about it).
I think it used time slots to regulate access, and outstations couldn't talk to each other, only to the central station (which controlled the sequencing, and would perform routing).
Hugo
Remember, there are no linear amps in the equation - they're not going to be bleeding all over adjacent channels & that sort of thing. Burners got a bad name because most were made with no respect for quality.
:)
;) )
They're getting more range by concentrating the power into a beam - as opposed to the normal omnidirectional spread. The antennas they use have great gain in one direction, and sod all in others - you won't even hear them 5 degrees off-line.
There's nothing illegal about this - lots of other (read: more expensive) 802.11 units have been around for years sold with a pair of yagis for the express purpose of linking two distant sites.
I'm probably completely wrong, but from memory the maximum gain allowable for the entire link (both ends) was 24dB. Full kits with directional antennae (don't forget the lightning surge arrestor if you're mounting antenna where they might get touched by the hand of zeus) are available, but they usually cost.
Problem is, the normal "budget" 802.11 cards were crippled by just having an internal antenna and no place to connect a "real" one. Nice to see apple have done the right thing
Hugo
(just running out to buy a couple of airports to link the empeg mk2 in my car to the office
Ok, I don't like it particularly, but most (all?) of Europe has VAT - Value Added Tax. It's just sales tax on everything with a few exceptions (in the UK at least - books, basic foods, sanitary products, etc). You get charged it on everything you buy from anywhere, and it's a uniform rate for almost all goods (electricity & other utilities an exception at the moment). The UK has VAT at 17.5%.
If you import stuff from outside Europe, you get charged VAT by customs & excise (as well as import duty). If you buy something from inside Europe but outside your country, you pay VAT at source - ie, where you got the stuff from. Businesses get to claim back their VAT, but it's still charged when (eg) a distributor buys a whole load of PCs from HP.
The point is - it's uniform. The playing field is fairly level and will get more level as the countries in the EU move their VAT levels into agreement. It's a slight headache for businesses, but hey - if you run a business you have a pile of paperwork to deal with anyway. Noone has to worry about net businesses having an unfair advantage (apart from having virtual premises and often virtual stock) - you can't avoid the VAT.
If the US took up such an option, each business would have to account for sales to other states (like you have to do when selling to other countries in the EU) and then it all would get balanced up by the federal government.
Might work. Works here.
Hugo
Errr, I don't think so. If you run MP3 at the sort of bitrates that ATRAC currently runs at (circa 256kbit) MP3 sounds at least as good.
In blind listening tests with trained listeners at a recent perceptual audio coding conference, 192kbit mp3 was statistically undistinguishable from the source cd. Going to 256k only hammers the point in more - mp3 is good, but isn't at its best at 128k.
Problem is, you've probably not heard MP3s though good equipment, whereas MD players are mature products which have to compete on sound - most mp3 players compete on price alone.
Hugo
empeg
However, you don't need any of that to make a SA1100 based system. The empeg is SA1100 based. We didn't use any development boards, we just wired up the cpu and it worked - as the SA1100 is so highly integrated, there's only *one* way to wire ram chips to it, *one* way to wire a flash chip to it and *one* way to hook up serial ports and LCD. There's nothing magic about the itsy which anyone competent enough to fabricate & build an itsy from the designs couldn't simply do themselves. I mean, if they gave you the Orcad files for it, what do you plan to do - buy orcad? The schematic editor alone costs over $1000. The problem is, building computers simply isn't hobbyist compatible anymore. 0.5mm pitch SMT components? You need a binocular microscope really, even if you can mount them chips using a fine tipped soldering iron, SMT solder (silver loaded stuff), and lots of SMT rework flux. That's not even considering how hard it is to *get* the parts - tried getting EDO ram? 3.3v stuff? In small quantities? You can't get these bits (including the SA) just off the shelf, though you can get most of the discretes through some of the better (catalogue-publishing) distributors such as Farnell. PCBs? You're looking at maybe $1000 minimum for one board set on long turnaround. Compaq have obviously been asked by a lot of universities for prototypes, so they've released the designs so research departments can make a few and do software development. It's not a hardware development platform - it's too small to prototype new stuff on. Hugo empeg
StrongARM is *already* much lower power consumption than the Crusoe 3120: some of the Crusoe figures are quite funny. Looking at the 3120 datasheet, it takes a mere (!) 1.4W whilst playing mp3's.
.mov's. Add the 1101 companion and you get VGA, PS/2 mouse/kb, USB host and more besides.
Compare this to around 0.075W (30% usage, CPU static when not doing decode) for a StrongARM doing MP3 decode. Hey, even *flat out* a 200Mhz, 1.5v core StrongARM on a 3-year old 0.25 micro process takes less than 0.25W. Just wait for the SA-2 stuff coming out this year - they're claiming more than 2x the performance whilst using less power.
Transmeta pushing mobile linux (and OEMs making 1st-gen designs around it, application support, etc) will probably ultimately benefit *real* low-power chips like the ARM, as linux applications (to a large extent) are pretty architecture-agnostic.
Looking at embedded machines, the Crusoe CPU also needs a lot of support circuitry. As an example, the SA1100 includes (on chip) a PCMCIA interface, real time clock, TFT/passive LCD drive, sound codec port, MMU, DRAM controller, USB slave, IrDA (SIR/FIR), sync serial port, a couple of async serials, and plenty of GPIO pins. Heck, it's almost a webpad on its own - which is why the Itsy can be so small, so cute, and still run off a couple of AA cells while playing
The crusoe is impressive as an x86 chip. Once you leave that narrow (in embedded terms) slot, the performance doesn't really measure up to offerings from ARM, MIPS, etc.
Hugo
Errr, the Cirrus 7209 (ARM7T core) CPU uses only 0.087W when *playing* mp3's, and *less than* 0.001W when in standby.
The Transmeta chips are low power compared to other x86's, but not low power when compared to the best of other architectures in terms of mips/w (eg, StrongARM, ARM10, etc).
Hugo
It uses a SA-1100 CPU (not sure on clock speed, but 133Mhz is the lowest they go) and VxWorks.
:)
It's capable of running linux, though - that's the same CPU we use in the empeg
Hugo
Fewer licence fees:
A clue. The licence fee for MP3 decoding hardware is $1.
Hugo
If you read the site, it uses an internal rechargeable battery to give 10 hours of life. The HDD only runs once every 10 minutes, for (my guess) around 15 seconds.
Hugo
Completely the wrong comparison. The smaller the drive, the lower mass the heads, which means the higher the shock rating. Hence why the IBM drives have an *operating* shock rating of 150g.
We use similar 2.5" drives in the empeg, and it's been through full vibration testing - we also keep them in a shock-mounted cradle though.
Hugo
It's not an ARM, it's a DSP @ 33Mhz.
USB slave (as required on a device like this) is not too hard, really. Nothing like the complexity of a USB host (as is currently in 2.3 kernels). The driver I did for the empeg is in the kernel source, but it's not very useful unless you use a USBN9602... which is found in zero PCs.
Hugo
Hmm, maybe not with a stressed-out 33Mhz DSP. The Kodak 265 has a powerPC and runs something like VXworks, so it's a decent CPU and a decent RTOS.
Hugo
Quite possibly, 64Mb of DRAM in an active state (ie, you need to be accessing *some* of it) will take more power than nothing, which is what a HDD which has had the power removed takes.
They won't just be putting the HDD into standby - they still take a non-insignificant amount of juice for a battery-powered player.
Hugo