f*****0
发帖数: 489 | 1 I just received 30 32bit mcus. those are 50mhz parts, 32bit risc mcus
running at 50mhz, 32k flash / 8k sram, one channel of ssi, one channel of
i2c, 8 channels of 10-bit adc running at 1 million samples per second.
60mips and 100mw in power dissipation.
for those chips, I paid more for shipping than for the chips themselves.
whoever is still programming 8/16 bit mcus must be concerned because their
days are numbered. |
w**i
发帖数: 136 | 2 I am not a professional in MCUs, but are you sure of your conclusion? How is
the cost to replace old systems designed with 8/16 bit MCUs, including
hardware redesign, software reprogram, system retest and perphaps FDA
approval reapply? If old systems function well, do you think designers and
managers will bother to use alternatives?
Think technically, but act practically.
【在 f*****0 的大作中提到】
: I just received 30 32bit mcus. those are 50mhz parts, 32bit risc mcus
: running at 50mhz, 32k flash / 8k sram, one channel of ssi, one channel of
: i2c, 8 channels of 10-bit adc running at 1 million samples per second.
: 60mips and 100mw in power dissipation.
: for those chips, I paid more for shipping than for the chips themselves.
: whoever is still programming 8/16 bit mcus must be concerned because their
: days are numbered.
|
p**********y
发帖数: 13 | 3
How is
including
and
so you think there is no market for mcu other than for replacement
purposes?
the annual sales volume for mcu is about 6 billion units and $15
billion. with about 100 - 200k people employed in their design,
engineering and manufacturing.
do you think they all sit around and work on how to replace those old
mcus?
Think rationally, but act strategically.
【在 w**i 的大作中提到】
: I am not a professional in MCUs, but are you sure of your conclusion? How is
: the cost to replace old systems designed with 8/16 bit MCUs, including
: hardware redesign, software reprogram, system retest and perphaps FDA
: approval reapply? If old systems function well, do you think designers and
: managers will bother to use alternatives?
: Think technically, but act practically.
|
w**i
发帖数: 136 | 4 I did not say there is no market for 32 bit MCUs. In my previous post, I
just gave an example which might possibly shake your conclusion. Take a
detailed example, say an Orman blood pressure monitor, which is quite
popular in Market. Do you think it is necessary to add fancy functions with
32 bits MCUS for the current products if you were the CEO of Orman? If there
is any improvement needed to be done in their system, modification is
enough. That is the reason I believe it is a long run before 8
【在 p**********y 的大作中提到】
:
: How is
: including
: and
: so you think there is no market for mcu other than for replacement
: purposes?
: the annual sales volume for mcu is about 6 billion units and $15
: billion. with about 100 - 200k people employed in their design,
: engineering and manufacturing.
: do you think they all sit around and work on how to replace those old
|
a****l
发帖数: 8211 | 5 why do you think programming/using 8/16 or 32bit mcu is so different that
those people now working on 8/16 mcu will have huge difficulty moving to
32bit mcu?
【在 f*****0 的大作中提到】
: I just received 30 32bit mcus. those are 50mhz parts, 32bit risc mcus
: running at 50mhz, 32k flash / 8k sram, one channel of ssi, one channel of
: i2c, 8 channels of 10-bit adc running at 1 million samples per second.
: 60mips and 100mw in power dissipation.
: for those chips, I paid more for shipping than for the chips themselves.
: whoever is still programming 8/16 bit mcus must be concerned because their
: days are numbered.
|
r*****d
发帖数: 44 | 6 绝大大多数字信号处理16位就够了.
对等情况下32位的功耗肯定高一截,未必是好的选择.
【在 f*****0 的大作中提到】
: I just received 30 32bit mcus. those are 50mhz parts, 32bit risc mcus
: running at 50mhz, 32k flash / 8k sram, one channel of ssi, one channel of
: i2c, 8 channels of 10-bit adc running at 1 million samples per second.
: 60mips and 100mw in power dissipation.
: for those chips, I paid more for shipping than for the chips themselves.
: whoever is still programming 8/16 bit mcus must be concerned because their
: days are numbered.
|
f*****0
发帖数: 489 | 7 "把part # 贴出来看看，1MSp/s，其他外设这个价位相当好。"
quite a few parts from luminary / ti do that. |
f*****0
发帖数: 489 | 8
so if a statement doesn't work for one example, the statement doesn't
work?
think rationally, and act intelligently.
【在 w**i 的大作中提到】
: I did not say there is no market for 32 bit MCUs. In my previous post, I
: just gave an example which might possibly shake your conclusion. Take a
: detailed example, say an Orman blood pressure monitor, which is quite
: popular in Market. Do you think it is necessary to add fancy functions with
: 32 bits MCUS for the current products if you were the CEO of Orman? If there
: is any improvement needed to be done in their system, modification is
: enough. That is the reason I believe it is a long run before 8
|
f*****0
发帖数: 489 | 9
that
why do you think that I think that way?
【在 a****l 的大作中提到】
: why do you think programming/using 8/16 or 32bit mcu is so different that
: those people now working on 8/16 mcu will have huge difficulty moving to
: 32bit mcu?
|
f*****0
发帖数: 489 | 10
I don't think performance is the reason for the proliferation of low-end
32-bit devices. obviously it helps. take a software-based vga display
for example. at 480 bit horizontal resolution, and 2 instructions per
second (to load up data), you need a mcu capable of 480*2/25us = 40mips.
that translates into a 160mhz PIC 8bit mcu - it doesn't exist. a 50mhz
32bit cortex-m3 chip (capable of 60mips+) can easily do that.
what will drive people to 32bit mcus is easier software development: the
ability
【在 r*****d 的大作中提到】
: 绝大大多数字信号处理16位就够了.
: 对等情况下32位的功耗肯定高一截,未必是好的选择.
|
|
|
a****l
发帖数: 8211 | 11 so what did you mean "whoever is still programming 8/16 bit mcus must be
concerned because their days are numbered."? My understanding is that they
might change to 32bit or even 64bit later at any time, when the application
requires such a switch.
【在 f*****0 的大作中提到】
:
: I don't think performance is the reason for the proliferation of low-end
: 32-bit devices. obviously it helps. take a software-based vga display
: for example. at 480 bit horizontal resolution, and 2 instructions per
: second (to load up data), you need a mcu capable of 480*2/25us = 40mips.
: that translates into a 160mhz PIC 8bit mcu - it doesn't exist. a 50mhz
: 32bit cortex-m3 chip (capable of 60mips+) can easily do that.
: what will drive people to 32bit mcus is easier software development: the
: ability
|
p******a
发帖数: 21 | 12 业界一般只会考虑适用性，提供50%的冗余就行了，太好的芯片可能需要新的平台，新
的开发工具，更复杂的调试设备。还有很重要的一点是老的8/16位的芯片可以用插座，
4层甚至双面板，这32位的至少得4面贴，甚至BGA，低功耗意味着多路电源供应，至少6
层板才能把信号走出去，这贴装测试的成本也不低。除非做独立的可替换控制模块，但
这模块的接口插座费用和可靠性都少不了。总之适用就好 |
r*****d
发帖数: 44 | 13
能做到多少个MIPS和bitwidth又没有直接关系, 市场上也有几百个MIPS的16bit
的DSP(可以用SIMD提高处理能力). 关键还是看需要, 有多少定点算法需要32位的精度?
功率就是和bitwidth和MHZ成正比的 (相同条件下, 不同处理器就不是
apple to apple比较了), 你还能举出反例出来?
据我所知Cortex虽然处理比arm7强一截,不过MIPS/mW 性能反而差了.
【在 f*****0 的大作中提到】
:
: I don't think performance is the reason for the proliferation of low-end
: 32-bit devices. obviously it helps. take a software-based vga display
: for example. at 480 bit horizontal resolution, and 2 instructions per
: second (to load up data), you need a mcu capable of 480*2/25us = 40mips.
: that translates into a 160mhz PIC 8bit mcu - it doesn't exist. a 50mhz
: 32bit cortex-m3 chip (capable of 60mips+) can easily do that.
: what will drive people to 32bit mcus is easier software development: the
: ability
|
f*****0
发帖数: 489 | 14
they
application
what I meant is exactly what I wrote: whoever is still programming 8/16
bit mcus must be concerned because their days are numbered.
【在 a****l 的大作中提到】
: so what did you mean "whoever is still programming 8/16 bit mcus must be
: concerned because their days are numbered."? My understanding is that they
: might change to 32bit or even 64bit later at any time, when the application
: requires such a switch.
|
f*****0
发帖数: 489 | 15
少6
through-holes are in general more expensive to manufacture than smd.
through-holes are great for experimenting though.
many of the newer chips have onboard regulators so they don't need
multiple power supplies - the ones requiring rtc may be an exception.
with 2ma or 4ma ports, you will have no problem running at 50mhz.
【在 p******a 的大作中提到】
: 业界一般只会考虑适用性，提供50%的冗余就行了，太好的芯片可能需要新的平台，新
: 的开发工具，更复杂的调试设备。还有很重要的一点是老的8/16位的芯片可以用插座，
: 4层甚至双面板，这32位的至少得4面贴，甚至BGA，低功耗意味着多路电源供应，至少6
: 层板才能把信号走出去，这贴装测试的成本也不低。除非做独立的可替换控制模块，但
: 这模块的接口插座费用和可靠性都少不了。总之适用就好
|
f*****0
发帖数: 489 | 16
度?
name a few 8/16 bit mcus capable of 几百个MIPS.
that's true because most of the power consumption in a cmos chip comes
from charging up and discharging gate capacitance.
agree. That's why i was quite puzzled when someone wrote the following.
绝大大多数字信号处理16位就够了.
对等情况下32位的功耗肯定高一截,未必是好的选择.
there are plenty of data on that so you don't have to guess.
But before you set out doing anything, I just wanted to point out for
you that your 所知 is wrong. but you can find facts if you want to see
for yourself.
【在 r*****d 的大作中提到】
:
: 能做到多少个MIPS和bitwidth又没有直接关系, 市场上也有几百个MIPS的16bit
: 的DSP(可以用SIMD提高处理能力). 关键还是看需要, 有多少定点算法需要32位的精度?
: 功率就是和bitwidth和MHZ成正比的 (相同条件下, 不同处理器就不是
: apple to apple比较了), 你还能举出反例出来?
: 据我所知Cortex虽然处理比arm7强一截,不过MIPS/mW 性能反而差了.
|
f*****0
发帖数: 489 | 17
true. they did a good job marketing the cortex-m3 architecture to compete
with the 8/16 bit mcus.
I am moving up from the pic world, and struggled with the arm7's way of
setting / clearing the gpio. and the bit banding on the cortex-m3 is a big
help for me.
they have a pretty good portfolio now and covers pretty much anything that
a typical 8/16bit mcu can do. |
