Post by thread:[EE] USB vs. Serial (was: Re: 'C' PIC programmer f

Peter L. Peres <spam_OUTpiclistTakeThisOuTmit.edu> wrote: > USB is a cash cow. And that's bad because...? > It is a set of protocols and hardware specifications designed to sell > chips and products made by a group of manufacturers with a specific > set of manufacturing and engineering capabilities. Kinda like Ethernet, or RS-232, or JTAG or ASCII? > The common points seem to be: most silicon that will run off usb will > be low power cmos at 3V or lower. Most silicon, period, is low-power CMOS at 3.3V or lower. Do you have a preference for older, more power-hungry technologies, or juat have a stock of 78L05s that you're trying to use up? > It will be intelligent (no $2 homemade interfaces are possible). Oh, please. Even if you buy through Digikey, you can get a Cypress CY7C63723 (USB microcontroller with 8K code space and 256 bytes RAM in an 18-pin DIP package) for $3.60 in quantity one. If you buy 100, the Digikey price drops to $1.80. > Any need to extend the network beyond basic capabilities will require > the end user to purchase more equipment (hubs), that will sell more > chips. Most new PCs come with at least 4 USB ports, and the hot-swap nature of USB means that that's usually enough no matter how many cameras, MP3 players, flash drives, or PDAs you have. Even if an end-user DOES need to buy a hub, they're cheap -- I just bought a high-speed hub (480 Mbps, 4 ports, 4 TTs) for $13, including shipping -- and I know NO ONE who needs more than one. > For all this complexity the users will receive the benefit of automatic > device identification and driver loading, and a little more speed than > serial "A little more speed than serial" is a bit of an understatement... When's the last time you pushed 50+ megabytes per second (that's, um, 500 million baud) through an RS-232 link? You can do it with USB, again using Digikey prices, for $7.20. And... Do you REMEMBER what it was like to use RS-232 peripherals? You needed: 1. The right cable. I had a box filled with DB9-to-DB25 adapters, null modems, 9- and 25-pin gender changers, straight-through cables, cables with pins 2 and 3 flipped, cables with all combinations of pins 4, 6, 11, and 20 tied to each other, one of those Radio Shack RS-232 testers with the bicolor LEDs, and a $50 breakout box so I could figure out which combination of all the above I needed for any particular serial connection. 2. An available COM port. Most PCs had two; some had only one. Really fancy PCs had four, but they were rare and most software wouldn't work with COM3 or COM4 anyway. If you only needed to use one device at a time, you could plug a whole bunch of serial devices into a $100+ switchbox built from a massive rotary switch and about a zillion little hand-soldered wires, but then you'd also need to buy duplicates of all your gender changers and adapters, etc. It was even worse a few years before USB was invented: Your 14.4k modem couldn't POSSIBLY fit on an ISA card, so it used one of the serial ports, and if you didn't buy a "bus mouse", your mouse used the other. 3. COM port configuration. Pick an IRQ, but make sure it doesn't conflict with any others. Flip a coin to decide if you want 0x378, 0x3F8, 0x278, or 0x2F8, whatever those are. If you have four COM ports, be aware that two of them will conflict with the other two. 4. RS-232 configuration. At the PC, you needed to set baud rate, parity, and number of stop bits. You also got to choose between hardware handshaking (although you couldn't usually specify what that actually meant), XON/XOFF flow control, or no flow control at all. If you were unlucky, you also got to choose some subset of those options on the device side, then guess at what the device manufacturer set the unconfigurable options to. 5. Power. Unless the device drew very little power, which was unlikely (hmm, too bad 3.3V low-power CMOS wasn't widely available at the time...), you needed a wall wart for every serial device. 6. Software. If a serial device wasn't a modem (and even, sometimes, if it was), it usually required you to install a driver on the PC. When you were finally done with all that, you had a more-or-less reliable connection to one device at a time that usually ran at around 1000 bytes per second, but could sometimes be as fast as 5000-12000 bytes per second. Since that was too slow for printers, scanners, ZIP drives, external CDROM drives, etc., THOSE devices used the parallel port. I think I've repressed most of my memories of that nightmare, but here's a little of what I recall was the usual sequence of events: 1. Configure the following alphabet soup: SPP, EPP, ECP 1.1, ECP 2.0. If your device needs a configuration that you don't have, go buy a parallel-port card. If you have a laptop, go buy a desktop. 2. Plug your ZIP drive into the PC, plug your scanner into the ZIP drive, plug your printer into the scanner. Plug all their power cords into the wall. With fire-extinguisher at the ready, turn on the PC. 3. Install drivers for all three. Spend a day configuring them and experimenting with power-sequencing until they all sorta work. 4. Try to run PCAD. Notice that it won't run without the parallel-port dongle that used to be plugged into your machine. Power everything down, plug the dongle between the PC and the ZIP drive, power it all back up. Notice that now NOTHING works. 5. Throw out the scanner and ZIP drive. Buy a SCSI scanner and interface card, and an internal ZIP drive. Open your PC and see that there's no room for the SCSI card and all your IDE connectors are already used. Punch something. Return the scanner, return the ZIP drive, sell the PC, seek employment as a shepherd. How, again, is USB not a gigantic improvement over that? > (but less than firewire, which was already available when usb was > invented). Firewire and USB both became available at around the same time (although Firewire had, it's true, been in development for over a decade). Firewire, though, is significantly more expensive than USB, even aside from the required royalty (originally $1 per port to Apple, now $0.25 per device to 1394LA). If it weren't so, more PCs would have Firewire ports and Apple wouldn't be using our USB chip in the iPod. > This is just my opinion, feel free to comment. Mine, too. -Andy === Andrew Warren - .....aiwKILLspam@spam@cypress.com === === Principal Design Engineer === Cypress Semiconductor Corporation ____________________________________________

> And... Do you REMEMBER what it was like to use RS-232 peripherals? > You needed: > > 1. The right cable..... <Snip> Wow! that's a scary horror story worthy of telling on Halloween Night. :-) -Denny ____________________________________________

> > USB is a cash cow. > And that's bad because...? I must agree with Andrew that USB is (nearly always) far easier to use that serial or parallel. OTOH when it does *not* work I don't have a clue why, with serial I could often get it to work. OTOH the fee for a USB ID is a significant hurdle for low-cost product development. The easiest way out is to buy an FT chip and use their vendor ID. I would love a company that re-selled USB IDs in small quantities! Wouter van Ooijen -- ------------------------------------------- Van Ooijen Technische Informatica: http://www.voti.nl consultancy, development, PICmicro products docent Hogeschool van Utrecht: http://www.voti.nl/hvu ____________________________________________

> Wow! that's a scary horror story worthy of telling on > Halloween Night. :-) That's no horror story, it was my typical day at work 15 years ago and I liked it! (Nowadays I like it when I plug in an extra hub - Andrew, I do have 4 hubs here, not because I need that number of ports but because it reduces the cable mess -, plug in the device, and it works.) Wouter van Ooijen -- ------------------------------------------- Van Ooijen Technische Informatica: http://www.voti.nl consultancy, development, PICmicro products docent Hogeschool van Utrecht: http://www.voti.nl/hvu ____________________________________________

On Oct 31, 2004, at 10:13 PM, Denny Esterline wrote: >> >> 1. The right cable..... > > <Snip> > > Wow! that's a scary horror story worthy of telling on Halloween Night. > :-) > > It's not too bad, once you accept the mindset that IT'S ALWAYS THE CABLE preventing correct operation. I did a presentation (to cisco's TAC) that essentially said "forget trying to figure our what you think you ought to need. Get a cable tester and see what's actually there..." BillW ____________________________________________

Wouter van Ooijen <piclistKILLspammit.edu> wrote: > I would love a company that re-selled USB IDs in small quantities! I actually thought of doing that shortly after I heard the USB-IF had increased their non-member Vendor ID fee from $200 to $1500, but I decided against it for two reasons. First, people who sublet my ID would be extremely unlikely to get their devices properly tested... Which would be fine for each of them (in the near term, at least), but bad for the end-user community as a whole. Second, Vendor ID/Product ID uniqueness really couldn't be guaranteed anyway. The current system self-enforces VID/PID uniqueness, because no vendor needs more than the one unique ID assigned to him by the USB-IF, and it's in his best interest to keep his Product IDs -- which he selects, and which cost him nothing -- unique. It's easy (and free) for him to assign a guaranteed-unique VID/PID combination to each product he designs, even if he plans only to build one or two of that product. If I let people share my Vendor ID, but sold and kept track of unique Product IDs, a vendor who knew he was only going to make a few different devices might be tempted to save the cost of multiple Product IDs by just purchasing one from me and reusing it... Or worse, by just picking an arbitrary Product ID for each of his products. At that point, which I figured would happen about a day after I first offered Product IDs for sale, the guarantee of uniqueness would be gone and we might as well all be picking random VIDs. -Andy === Andrew Warren - .....aiwKILLspam.....cypress.com === === Principal Design Engineer === Cypress Semiconductor Corporation ____________________________________________

>5. Throw out the scanner and ZIP drive. Buy a SCSI scanner and interface card, and an internal ZIP drive. Open your PC and see that there's no room for the SCSI card and all your IDE connectors are already used. Punch something. Return the scanner, return the ZIP drive, sell the PC, >seek employment as a shepherd. > How, again, is USB not a gigantic improvement over that? I don't know, sometimes I thought working as a shepherd would be a peaceful and satisfying existence, especially when one of the above-mentioned devices belches it's guts at 2:00 AM when I am frantically working on a 8:00 AM deadline project. Just think - sitting out in a field watching the sun rise and set....a jug of grape juice and a loaf of bread at the handy, a trusty sheep dog doing all the real work while you sit against a tree... -- Lawrence Lile, P.E. Electrical and Electronic Solutions Project Solutions Companies http://www.projsolco.com --- Outgoing mail is certified Virus Free. Checked by AVG anti-virus system (http://www.grisoft.com). Version: 6.0.762 / Virus Database: 510 - Release Date: 9/13/2004 ____________________________________________

On Oct 31, 2004, at 9:00 PM, Andrew Warren wrote: > Oh, please. Even if you buy through Digikey, you can get a Cypress > CY7C63723 (USB microcontroller with 8K code space and 256 bytes RAM > in an 18-pin DIP package) for $3.60 in quantity one. The development environment is a bit steep at $500 for an emulator. No flash, not even any UV-erasables :-( So who's going to implement a "USB Stamp"? Smallish USB controller, like the above, and an eeprom for "program storage"; download your interpretable bytecodes into eeprom via USB, and include USBin and USBout "instructions" for each of the possible endpoints? BillW ____________________________________________

On Mon, 1 Nov 2004, Andrew Warren wrote: > Wouter van Ooijen <EraseMEpiclistspam_OUTTakeThisOuTmit.edu> wrote: > >> I would love a company that re-selled USB IDs in small quantities! > > I actually thought of doing that shortly after I heard the USB-IF had > increased their non-member Vendor ID fee from $200 to $1500, but I > decided against it for two reasons. > > First, people who sublet my ID would be extremely unlikely to get > their devices properly tested... Which would be fine for each of > them (in the near term, at least), but bad for the end-user community > as a whole. Agreed, but you can't fish in a frozen lake without holes in the ice. Also, please see my other posting wrt. VID/PID selling. Maybe someone would be able to buy a few vendor id's and sell cheap non-exclusive vendor id tickets for small developers and protyping work. After all, it's a business idea like any other. Who wouldn't pay $50 to be allowed the use of a 'development id' and associated device id, reserved for him for a limited period (say 2 years). This is almost in line with prices for DNS. > Second, Vendor ID/Product ID uniqueness really couldn't be guaranteed > anyway. The current system self-enforces VID/PID uniqueness, because > no vendor needs more than the one unique ID assigned to him by the > USB-IF, and it's in his best interest to keep his Product IDs -- > which he selects, and which cost him nothing -- unique. It's easy > (and free) for him to assign a guaranteed-unique VID/PID combination > to each product he designs, even if he plans only to build one or two > of that product. The VID/PID uniqueness could be guaranteed in the same way the current DNS name allocation works. A number of private entities sell them by delegation, and there are registries. Some ids (like .gov .mil etc) are clearly reserved and not managed by private entities. > If I let people share my Vendor ID, but sold and kept track of > unique Product IDs, a vendor who knew he was only going to make a few > different devices might be tempted to save the cost of multiple > Product IDs by just purchasing one from me and reusing it... Or > worse, by just picking an arbitrary Product ID for each of his > products. Yes, it's called domain squatting in the DNS world. Nobody died of it although someone (usually lawyers) occasionally made some money from it. > At that point, which I figured would happen about a day after I first > offered Product IDs for sale, the guarantee of uniqueness would be > gone and we might as well all be picking random VIDs. Yes, keep the unwashed masses where they belong ;-) (just kidding) no holes in the ice, no fishing. Peter ____________________________________________

On Mon, 1 Nov 2004, Wouter van Ooijen wrote: >>> USB is a cash cow. >> And that's bad because...? > > I must agree with Andrew that USB is (nearly always) far easier to use > that serial or parallel. OTOH when it does *not* work I don't have a > clue why, with serial I could often get it to work. > > OTOH the fee for a USB ID is a significant hurdle for low-cost product > development. The easiest way out is to buy an FT chip and use their > vendor ID. I would love a company that re-selled USB IDs in small > quantities! As an alternative, how about usb silicon makers offering a 'ride' on their id (or on specific id's they's purchase for this specific purpose), for a few pennies more. E.g. you buy a Cypress USB client controller cpu and you have the option to get an id and uid from Cypress (you share the id with many others but the device id is uniquely guaranteed by the database consolidated application process), for a few $, through a web form and PayPal payment. If there would be enough demand this scheme would pay for itself very fast imho. But is there enough demand ? And what about liability ? I.e. if something with a id from Cypress f.ex. would catch fire, who would be liable ? Peter ____________________________________________

On Sun, 31 Oct 2004, Andrew Warren wrote: > Peter L. Peres <piclistspam_OUTmit.edu> wrote: > >> USB is a cash cow. > > And that's bad because...? Because I am not among those riding it of course. >> It is a set of protocols and hardware specifications designed to sell >> chips and products made by a group of manufacturers with a specific >> set of manufacturing and engineering capabilities. > > Kinda like Ethernet, or RS-232, or JTAG or ASCII? JTAG, ASCII and RS232 cannot be considered closed standards in the context I was pointing out. The context was pretty clear I think, but I will point it out again: I was referring to the current trend to lock customers in by requiring specific silicon to be used (besides id fees etc). This is an old trend but it has raised to an unprecedented level with usb imho. >> The common points seem to be: most silicon that will run off usb will >> be low power cmos at 3V or lower. > > Most silicon, period, is low-power CMOS at 3.3V or lower. Do you > have a preference for older, more power-hungry technologies, or juat > have a stock of 78L05s that you're trying to use up? In the non-smd player area, with mixed analog-digital off the shelf components, where most small projects are prototyped and manufactured when done outside a large enterprise, 3.3V is not the preferred supply voltage, and 0.8W is really low power. The additional requirement to negotiate power with the host is an extra hurdle imho (from the point of view of small projects). Whyever they did this is beyond my comprehension. The PCs that were to run as usb hosts had power supplies exceeding 250Watts since before the usb trend started. I see it as an attempt to limit potential liability for fire and other damages and a way to save another few cents (to be poined up by users for external power supplies later). Compare this to the power specs for firewire. >> It will be intelligent (no $2 homemade interfaces are possible). > > Oh, please. Even if you buy through Digikey, you can get a Cypress > CY7C63723 (USB microcontroller with 8K code space and 256 bytes RAM > in an 18-pin DIP package) for $3.60 in quantity one. If you buy 100, > the Digikey price drops to $1.80. I accept your explanation, and I will likely go that way to see how much it gets to cost after the dust settles (remember I'm not in the us). This means, prototype board (probably 2 or 3 editions until the bugs all lie low), software (I have not programmed a Cypress cpu before), computer issues (does the Cypress ide run un linux or *bsd ? Do I need to buy a programmer or can i build one ? What's a reasonable budget for a complete ide including programmer and learning curve for the new cpu and its special functions ?), and the project itself, plus interesting money transfer and post office/customs issues you probably don't want to know about since you are fortunate enough to live in the us. >> Any need to extend the network beyond basic capabilities will require >> the end user to purchase more equipment (hubs), that will sell more >> chips. > > Most new PCs come with at least 4 USB ports, and the hot-swap nature > of USB means that that's usually enough no matter how many cameras, > MP3 players, flash drives, or PDAs you have. But you cannot use any of the USB peripherals with a microcontroller project. Well, almost. Anyway, the price of putting something that emulates a USB host controller into your project will likely double or triple its price, either in development time (f.ex. using a scenix chip to bit-bang usb 1.0 as I have read on the Internet), or in purchased silicon and prototype smd board manufacturing. This would wipe out the advantage of trying to use an existing USB product to do half the job of your project. Compare this to using TTL level RS232 which can be implemented for $0.5 and works both as host and as client. Don't get me wrong, I am not missing the serial setup nightmares, I am just saying that the present situation with USB is not good for small developers. I.e. USB was technical progress, but it lost some of the people who contributed most of the innovation in the field so far. > Even if an end-user DOES need to buy a hub, they're cheap -- I just > bought a high-speed hub (480 Mbps, 4 ports, 4 TTs) for $13, including > shipping -- and I know NO ONE who needs more than one. I agree, but for $13 you could build at least 4 devices according to the price you gave above - and they cannot interconnect using usb. Also the $13 hub is the end user price for something that is mass manufactured in the Far East. I'm pretty sure that a one off hub would run into $50ish without trying hard (even without the case), from smd board manufacture, shipping costs, assembly and parts purchased in 1-offs alone. Maybe someone could better me on this price, but likely not by much. {Quote hidden}>> For all this complexity the users will receive the benefit of automatic >> device identification and driver loading, and a little more speed than >> serial > > "A little more speed than serial" is a bit of an understatement... > When's the last time you pushed 50+ megabytes per second (that's, um, > 500 million baud) through an RS-232 link? You can do it with USB, > again using Digikey prices, for $7.20. > > And... Do you REMEMBER what it was like to use RS-232 peripherals? -- snip -- of excellent and true arguments. > You needed: > connectors are already used. Punch something. Return the > scanner, return the ZIP drive, sell the PC, seek employment as a > shepherd. > > How, again, is USB not a gigantic improvement over that? Yes, but it is one that is beyond economic feasability for 99% of the one-off people until they get the technology and licensing sorted out. Apropos licensing, what does one do if one aquires a Cypress or Microchip USB capable chip and wishes to make a product based on it. To simplify, suppose one makes something that uses the generic serial-usb driver and appears as a serial port. What ID does one put in there ? Fake it and get a number that's unassigned ? I don't think so. Copy one that's assigned to an existing manufacturer ? Not so good. So what ? $1500 or forget it ? Of course you realise what the economics of low count, low cost devices look like with a $1500 tax to be amortised within two years or less. {Quote hidden}>> (but less than firewire, which was already available when usb was >> invented). > > Firewire and USB both became available at around the same time > (although Firewire had, it's true, been in development for over a > decade). Firewire, though, is significantly more expensive than USB, > even aside from the required royalty (originally $1 per port to > Apple, now $0.25 per device to 1394LA). If it weren't so, more PCs > would have Firewire ports and Apple wouldn't be using our USB chip in > the iPod. Apple/Sony firewire was available a relatively long time before usb came out afaik (by this I mean, being deployed and affordable for a reasonable number of pc's or macs). Firewire was specialised use (mostly cameras, portable digital media storage and such). Usb targeted a different group of users and only became ubiquitous by 1999 or so when Windows 98SE started supporting it 'out of the box'. After that it took a small while to spread and for the prices to drop to the intended (?) consumer levels. The interesting part is that firewire equipment has now also dropped to retail/consumer price levels. >> This is just my opinion, feel free to comment. > > Mine, too. ok, Peter ____________________________________________

>The additional requirement to negotiate power with the host >is an extra hurdle imho (from the point of view of small projects). It does not get negotiated, in the case of the FTDI devices it is a field in the eeprom where you enter your maximum required current, and the host can set a current limit at that. The host just reads out the value. >Whyever they did this is beyond my comprehension. ... >I see it as an attempt to limit potential liability >for fire and other damages which is probably a "good thing" when running on a laptop, to limit mistreatment of the battery, which may cause exactly that in a worst case situation. Alternatively it allows the host device to detect a fault condition on a device and turn it off before it takes down the whole system by shorting the power supply. Would you be happy if a short on a USB device caused your desktop machine to shut down, potentially loosing work in progress? ____________________________________________

>As an alternative, how about usb silicon makers offering a >'ride' on their id (or on specific id's they's purchase for >this specific purpose), for a few pennies more. E.g. you buy >a Cypress USB client controller cpu and you have the option >to get an id and uid from Cypress (you share the id with >many others but the device id is uniquely guaranteed by the >database consolidated application process), for a few $, >through a web form and PayPal payment. If there would be enough >demand this scheme would pay for itself very fast imho. But >is there enough demand ? And what about liability ? I.e. >if something with a id from Cypress f.ex. would catch >fire, who would be liable ? In effect I believe the FTDI devices allow this, within the limits of the type of chip they represent. ____________________________________________

This is exactly what FTDI do, and they issued me with a block of 8 ID's. (of which I used 2!) -----Original Message----- From: @spam@piclist-bouncesKILLspammit.edu [KILLspampiclist-bouncesKILLspammit.edu] On Behalf Of Alan B. Pearce Sent: 02 November 2004 12:12 To: Microcontroller discussion list - Public. Subject: Re: [EE] USB vs. Serial (was: Re: 'C' PIC programmer from USB Portpower) {Quote hidden}>As an alternative, how about usb silicon makers offering a 'ride' on >their id (or on specific id's they's purchase for this specific >purpose), for a few pennies more. E.g. you buy a Cypress USB client >controller cpu and you have the option to get an id and uid from >Cypress (you share the id with many others but the device id is >uniquely guaranteed by the database consolidated application process), >for a few $, through a web form and PayPal payment. If there would be >enough demand this scheme would pay for itself very fast imho. But is >there enough demand ? And what about liability ? I.e. >if something with a id from Cypress f.ex. would catch fire, who would >be liable ? In effect I believe the FTDI devices allow this, within the limits of the type of chip they represent. ____________________________________________

> This is exactly what FTDI do, and they issued me with a block of 8 ID's. > (of > which I used 2!) How do you go about requesting these ID's? Bob Ammerman RAm Systems ____________________________________________

William Chops Westfield writes: > So who's going to implement a "USB Stamp"? Smallish USB controller, > like the above, and an eeprom for "program storage"; download your > interpretable bytecodes into eeprom via USB, and include USBin and > USBout "instructions" for each of the possible endpoints? There are a few people making FTDI USB -> TTL (serial and parallel) solutions on DIP headers, but at $20-$30/ea they aren't cheap. I think everyone else has just been waiting anxiously for the PIC18F devices with USB support to hit the shelves. Either that, or the Atmel AT91SAM7 series, which just might become my universal hobby chip of choice... -p. ____________________________________________

Linx has one (SDM-USB-QS-S) which is $10 (qty 200) to $15 (qty 1) at DigiKey. I bet it can be found for less with some searching. {Original Message removed}

On Tue, 2 Nov 2004, Alan B. Pearce wrote: >> The additional requirement to negotiate power with the host >> is an extra hurdle imho (from the point of view of small projects). > > It does not get negotiated, in the case of the FTDI devices it is a field in > the eeprom where you enter your maximum required current, and the host can > set a current limit at that. The host just reads out the value. > >> Whyever they did this is beyond my comprehension. > ... >> I see it as an attempt to limit potential liability >> for fire and other damages > > which is probably a "good thing" when running on a laptop, to limit > mistreatment of the battery, which may cause exactly that in a worst case > situation. Alternatively it allows the host device to detect a fault > condition on a device and turn it off before it takes down the whole system > by shorting the power supply. Would you be happy if a short on a USB device > caused your desktop machine to shut down, potentially loosing work in > progress? No, but all external supplies must be current limited and/or protected against shorts. Unlike the keyboard +5V which is unfused on some boards, unlike the mouse +% ditto, unlike the joystick +5 ditto etc etc. F.ex. firewire pci cards use polyfuses, 1 on each external connector. So it's not that hard. Peter ____________________________________________