Shop OBEX P1 Docs P2 Docs Learn Events
P2 with custom ROM for $5000? — Parallax Forums

P2 with custom ROM for $5000?

I think during the live-on-YouTube bringup of the P2, either Chip or Ken mentioned that if there was something wrong with the ROM, it would basically cost $5000 to create a new mask and do a respin. (I realize that "respin" probably represents at least a 6 digit dollar amount).

That makes me wonder, is Parallax going to consider the possibility of offering a Propeller 2 with a custom ROM?

Just asking out of curiosity.

===Jac

Comments

  • cgraceycgracey Posts: 14,133
    edited 2018-10-01 03:02
    If it turns out that we can make unique ROM-variant versions for a $5,000 mask fee and a special wafer run (~$20k), we could do lots of interesting things.

    My friend Teva today said that we could make ROM programs to do things like control nearly 64 DACs/ADCs and have them selectable by resistor pull-up/pull-down combos. That would make the chip almost instantly sellable to people who wanted analog functions, without getting involved in programming.
  • jmgjmg Posts: 15,140
    cgracey wrote: »
    If it turns out that we can make unique ROM-variant versions for a $5,000 mask fee and a special wafer run (~$20k), we could do lots of interesting things.

    My friend Teva today said that we could make ROM programs to do things like control nearly 64 DACs/ADCs and have them selectable by resistor pull-up/pull-down combos. That would make the chip almost instantly sellable to people who wanted analog functions, without getting involved in programming.

    Can't you get close to that now, with TAQOZ strings ? You just need to provide the config scripts, and some report scripts ?
    I guess a ROM could give faster reports than TAQOZ could manage ?
  • potatoheadpotatohead Posts: 10,253
    edited 2018-10-01 03:36
    Jmg, those people want canned functions, controlled or responding to simple switches, pots, etc...

    Big market. I once ran into a servo done this way. A whole set of easy functions. It was a lot of money. The guys using it do not program. No interest. They have signals and just want actions based on them.

    That unit had a little micro coupled to a driver board. You set it up via serial and a gui. Once set, config written, it does the task, no worries. They treat it like a purpose built component. The tasks are simple. Any of us could do them.

    In this case, they employed a lit of these. 5 figures worth. To them, programming is more and not repeatable by future people, etc...

    Was very eye opening.

  • YanomaniYanomani Posts: 1,524
    edited 2018-10-01 03:52
    Hi Chip et al

    In the same line of thinking: one of those (several?) sub-$1.00, eight pin microcontrollers, selected among whichever family/line best suits the final target application, could complement that solution, without the need of any further new mask (no US$ 5,000.00 needed for this), nor respin, because there was no mention they will need something from that area of the circuit that was found to be faulty.

    And, IIRC, there are some of then (sub-$1.00s) that does readily have at least one ADC or DAC, or even both (here I'm unsure, someone could jump in and correct me, please).

    Thus a real 64 ADC/DAC combined solution is just at the corner, waiting to be called by someone with enough skills to program the eight-pin companion.

    Just a thought...

    Henrique
  • Certainly a custom ROM is an advantage especially if you need all 64 I/O although I would still opt for the "configuration device" being the SPI Flash leaving 60 I/O for those application specific functions since it is unlikely you would just sell the chip itself since it is not a simple DIP chip with 5V supply and a single decoupling cap, is it?
  • hi Peter Jakacki

    Unless you wanna sell some monster, 68000-alike, 64-pin dip, I'm totaly in favor of using a design just like your's; some P2D2-related ADDA!
  • Certainly a custom ROM is an advantage especially if you need all 64 I/O although I would still opt for the "configuration device" being the SPI Flash leaving 60 I/O for those application specific functions since it is unlikely you would just sell the chip itself since it is not a simple DIP chip with 5V supply and a single decoupling cap, is it?

    I imagine if Parallax can make it interesting enough (price-wise), there might be market for customers who want to get custom packages and custom bonding too. The Prop2 is so versatile, with all the DACs and ADCs and everything, someone might find it interesting to make a replica of another more expensive chip, or an old chip that's out of production, by putting a Prop2 in a DIP chip with a custom ROM and custom bonding.

    ===Jac
  • Cluso99Cluso99 Posts: 18,066
    edited 2018-10-01 04:10
    A number of chips for $30K+ custom programmed with security (as you can make it fully internal) could look quite feasible to a number of customers.
    I know such a person and he's following ;)

    I suspect Parallax would need to charge a customer 50K+. Now, what if the customer(s) could jointly fit their code into 16KB combined. Voila, they could share the chip together, and select the function by resistor(s). Two customers and its $25K+ each, three and its $17K+, four and $13K+.
  • jmgjmg Posts: 15,140
    Yanomani wrote: »
    Hi Chip et al

    In the same line of thinking: one of those (several?) sub-$1.00, eight pin microcontrollers, selected among whichever family/line best suits the final target application, could complement that solution, without the need of any further new mask (no US$ 5,000.00 needed for this), nor respin, because there was no mention they will need something from that area of the circuit that was found to be faulty.

    And, IIRC, there are some of then (sub-$1.00s) that does readily have at least one ADC or DAC, or even both (here I'm unsure, someone could jump in and correct me, please).

    Thus a real 64 ADC/DAC combined solution is just at the corner, waiting to be called by someone with enough skills to program the eight-pin companion.

    Yes, for ~ 25c these days you get 18k Flash (10kF parts ~ 18c) and 12b ADCs, UART, CalOSC, but no DAC. The ADC can select operating mode, via 2 resistors.
    That could boot consuming 2 pins ( I think the single-pin boot did not make it into the ROM ?)

    Of course, that's no longer quite a single-dedicated-mode part, with 'no programming', but maybe Parallax could sell pre-programmed small MCUs for common tasks, to gauge market demand ?
    If the customer hits critical mass, they can pay $5k and MOQ 2~5k ?

    If the customer profile for these does not include assembly, Parallax could sell a module (shrunk P2D2) with P2+Loaded PGM.

    Smallest MCUs are 3x3mm or, I see there are now 1.5mmx1.5mmx0.45 (!) 8-XFDFN SPI memories, in 64kB and 128kB, for 29/30c/3k, so that's going to look like a decoupling capacitor to most eyes :)
    With that, the P2 pin ADC can configure, and you lose a few more pins, but have more code space ?

  • YanomaniYanomani Posts: 1,524
    edited 2018-10-01 05:29
    Hi jmg

    There are so many options, and all just seem to perfectly aim to those new and still wandering targets.

    Despite P2 has emitted its first and loud cry way less than 72 hours ago, since it has been brought to life by Chip and the whole Parallax team (forum members included) the world has changed.

    There is a lot more fresh air to breath. Don't you feel something like this?
  • Why couldn't you make it with an OTP ROM that's programmed through a combination of the /TEST and RESn pins? It would just appear as a 16KB ROM chip. Then again why couldn't it have flash for the 16KB ROM section?

    Just asking stupid questions out loud (stupid because I'm "spending" someone else's money).
  • pedward wrote: »
    Why couldn't you make it with an OTP ROM that's programmed through a combination of the /TEST and RESn pins? It would just appear as a 16KB ROM chip. Then again why couldn't it have flash for the 16KB ROM section?

    Just asking stupid questions out loud (stupid because I'm "spending" someone else's money).
    I hope you're not being cynical.

    OTP was tried, but couldn't be guaranteed to correctly blow. 16kB of PROM inside a CPU would be a major accomplishment (what do you do with all the smoke?)
    Flash isn't feasible with the process technology used to produce the chips. Extra layers are more costly and affect yield. Internal Flash, when otherwise easily connected outside, is also a waste of silicon that could be actively used for other things.
  • cgracey wrote: »
    If it turns out that we can make unique ROM-variant versions for a $5,000 mask fee and a special wafer run (~$20k), we could do lots of interesting things.

    How many chips would a ~$20k wafer run yield?

  • whicker wrote: »
    pedward wrote: »
    Why couldn't you make it with an OTP ROM that's programmed through a combination of the /TEST and RESn pins? It would just appear as a 16KB ROM chip. Then again why couldn't it have flash for the 16KB ROM section?

    Just asking stupid questions out loud (stupid because I'm "spending" someone else's money).
    I hope you're not being cynical.

    OTP was tried, but couldn't be guaranteed to correctly blow. 16kB of PROM inside a CPU would be a major accomplishment (what do you do with all the smoke?)
    Flash isn't feasible with the process technology used to produce the chips. Extra layers are more costly and affect yield. Internal Flash, when otherwise easily connected outside, is also a waste of silicon that could be actively used for other things.

    It was the custom poly-fuse that Chip couldn't get working. OTP ROM has been around since 1969, there are plenty of micros with OTP ROM memory, so the tech certainly exists, no smoke involved.
  • jac_goudsmitjac_goudsmit Posts: 418
    edited 2018-10-01 19:33
    pedward wrote: »
    It was the custom poly-fuse that Chip couldn't get working. OTP ROM has been around since 1969, there are plenty of micros with OTP ROM memory, so the tech certainly exists, no smoke involved.

    But would one layer be enough for OTP? I'm thinking probably not. The Propeller 2 is just not deisgned for adding OTP ROM. It would probably require some peripheral hardware to enable and disable programming, to run relatively high currents to the matrix, and to stop high (programming) voltages from damaging the rest of the chip.

    Straight ROM is just a reticle (slide) with diodes drawn on it. I imagine it would be almost trivial to make one, and the yield probably doesn't change significantly for a different ROM mask either.

    ===Jac
  • For the ones that lost it, surrounded by (or better, buried under) the huge bunch of information that was furiously produced during the last months (years)...

    Flash does had the same fate of fuses; it has come, passed by and gone away.

    https://forums.parallax.com/discussion/comment/1432015/#Comment_1432015

    Knowing, now for sure, how much heat is being generated by P2 current silicon, I believe it was a wise decision to don't progress in that direction.

    Think about an interposer as being one more layer (three in fact, metal + silicon + metal, kind of a sandwich) of some heat conducting material that could not be understood as being a perfect heat spreader, stacked over P2 dice, with another smaller dice (flash memory) stacked over it.

    Now think on an egg, slowly cooking into some teflon-linned pan.

    Henrique
  • jmgjmg Posts: 15,140
    pedward wrote: »
    It was the custom poly-fuse that Chip couldn't get working. OTP ROM has been around since 1969, there are plenty of micros with OTP ROM memory, so the tech certainly exists, no smoke involved.
    Of course, but that's separate IP, and involves more complex powering, and redundancy, so Chip chose a simpler 16K serial ROM IP block from OnSemi.
    Future versions of P2 may well revisit the size and type of ROM.

  • potatoheadpotatohead Posts: 10,253
    edited 2018-10-01 20:06
    There is also new features vs existing ones.

    We have a ROM, and it has been suggested a mask change to update the ROM has a price.

    I will bet keeping a mask set around has another price, switching yet another one. That is having different ROMs available.

    OTP is a new feature / bigger price as synthesis and the other steps would come into play. Big price there. And we know this process is not up to doing those things, at least given the current design.

    Update the ROM basically means just that only, nothing else.

    And someone would have to ask about the price associated with keeping more than one mask, unless the ROM is simply changed, which is what Chip communicated.



Sign In or Register to comment.