l***g
发帖数: 1035 | 1 hope i can pick up some brains here. so i have 100 batteries, each one sampl
es a temperature and a voltage. i want to send all info to a center bms to p
rocess. after all, that is what a bms does.
most existing schemes are using CAN or Serial, throughput is pretty low as 5
Hz or worse. Can you think of a way of sending these data faster to somethin
g like 20Hz? keep in mind that some overhead information has to be included
like the ID of each data point and timestamp.
constraints: communication has to be bidirectional. | g******u
发帖数: 3060 | 2 have you heard fast CAN, can be implemented using the same tranceivers.
or you can group some local batteries to a local LIN or CAN and send the
data pack as a power good signal or sth. | l***g
发帖数: 1035 | 3 mathe problem with CAN is the many many wires.. and they fight for the same
hardware resource, a concern when all modules are sending data at a fast
pace.
【在 g******u 的大作中提到】
: have you heard fast CAN, can be implemented using the same tranceivers.
: or you can group some local batteries to a local LIN or CAN and send the
: data pack as a power good signal or sth.
| g****t
发帖数: 31659 | 4 My two cents:
我非常怀疑1/20秒内的平均温度有没有任何意义.如果你的散热设计正常的话.
温度数据里面,频率高的那部分,多半都是仪器噪声吧.
电压,你找个带A/D和内存的嵌入式板子,每1/5秒在板子上记录之前1/5秒内的四个采样
的数据,然后把这1/5秒的全部4个数据每1/5秒发一次CAN,不就行了?
数据有1/5秒的延迟.但是是包含之前快速采样的所有数据的.
按此原理,你可以自己专门弄个快速缓存.
hope i can pick up some brains here. so i have 100 batteries, each one sampl
es a temperature and a voltage. i want to send all info to a center bms to p
rocess. after all, that is what a bms does.
most existing schemes are using CAN or Serial, throughput is pretty low as 5
Hz or worse. Can you think of a way of sending these data faster to somethin
g like 20Hz? keep in mind that some overhead information has to be included
like the ID of each data point and timestamp.
constraints: communication has to be bidirectional.
【在 l***g 的大作中提到】
: hope i can pick up some brains here. so i have 100 batteries, each one sampl
: es a temperature and a voltage. i want to send all info to a center bms to p
: rocess. after all, that is what a bms does.
: most existing schemes are using CAN or Serial, throughput is pretty low as 5
: Hz or worse. Can you think of a way of sending these data faster to somethin
: g like 20Hz? keep in mind that some overhead information has to be included
: like the ID of each data point and timestamp.
: constraints: communication has to be bidirectional.
| l***g
发帖数: 1035 | 5 i take the points on the temp measuments. it's there if all data packs are c
onstant length.
for the voltage, no, the problem is on the delay, which reduces the reaction
time. the measurements is part of the closed loop controller, has to be dis
crete but periodic data.
thanks for the input.
sampl
p
【在 g****t 的大作中提到】
: My two cents:
: 我非常怀疑1/20秒内的平均温度有没有任何意义.如果你的散热设计正常的话.
: 温度数据里面,频率高的那部分,多半都是仪器噪声吧.
: 电压,你找个带A/D和内存的嵌入式板子,每1/5秒在板子上记录之前1/5秒内的四个采样
: 的数据,然后把这1/5秒的全部4个数据每1/5秒发一次CAN,不就行了?
: 数据有1/5秒的延迟.但是是包含之前快速采样的所有数据的.
: 按此原理,你可以自己专门弄个快速缓存.
:
: hope i can pick up some brains here. so i have 100 batteries, each one sampl
: es a temperature and a voltage. i want to send all info to a center bms to p
| c*******o
发帖数: 62 | 6 你的设计不可能100个brick就是一整块电池吧，通常要划分成若干module
譬如说每个module用一块pcba采集10-20个brick的电压和温度,pack成1-2条msg通过CAN
发送。另一端bms接收~处理~打包~整合~再发送，再加上其他乱七八糟的,整个过程20Hz
一点问题没有
每个brick单独做肯定不可能的，总共两百条msg才能发送一轮，就算你的bus其他什么
事都不干也来不及阿，什么CAN bus都顶不住 | l***g
发帖数: 1035 | 7 it's 100 modules, each module already contains some cells/bricks.
i know its difficult, that's why i'm brain storming here... | g****t
发帖数: 31659 | 8 什么控制会不滤波就使用1/20秒的温度?
既然后面还是要滤波,那还不如直接发1/5秒钟1次的温度.这样就省了一半的数据量.
对于电压,如果你的控制器承受不了1/5秒钟的delay,那压根就不能上车用.
因为即便你实现了20Hz的数据通讯,你必须考虑通讯出错的可能,对延迟的
tolerance是必须有的.
c
reaction
dis
【在 l***g 的大作中提到】
: i take the points on the temp measuments. it's there if all data packs are c
: onstant length.
: for the voltage, no, the problem is on the delay, which reduces the reaction
: time. the measurements is part of the closed loop controller, has to be dis
: crete but periodic data.
: thanks for the input.
:
: sampl
: p
| a****l
发帖数: 8211 | 9 Assume your sampling resolution is 32bit(which is insanely high), you have
4B*20Hz*100=8KB/S=64bkps. adding 29bit header (which is the longest) for
each message, you need approximately 128kbps. Assuming you are using 250k
CAN (which is a mid-range speed one), doesn't seems to have much problem
since the load is only around 50%.
sampl
p
5
somethin
included
【在 l***g 的大作中提到】
: hope i can pick up some brains here. so i have 100 batteries, each one sampl
: es a temperature and a voltage. i want to send all info to a center bms to p
: rocess. after all, that is what a bms does.
: most existing schemes are using CAN or Serial, throughput is pretty low as 5
: Hz or worse. Can you think of a way of sending these data faster to somethin
: g like 20Hz? keep in mind that some overhead information has to be included
: like the ID of each data point and timestamp.
: constraints: communication has to be bidirectional.
| g******u
发帖数: 3060 | 10 temperature is ok, even if there is short somewhere, it can't burn sth out
within 1/5 or 1/2 sec.
voltage is better be sensed faster. | | | g****t
发帖数: 31659 | 11 按我写的作,效果就是电压测量频率是足够快的,但是发送延迟了1/5秒.
就是搞个二级缓存.
temperature is ok, even if there is short somewhere, it can't burn sth out
within 1/5 or 1/2 sec.
电池倒是有可能温度急剧上升,然后爆炸.但那个需要保险丝或者
PTC来保险.控制loop用的温度,我认为1/20秒完全不make sense.
又不是核电站.
voltage is better be sensed faster.
【在 g******u 的大作中提到】
: temperature is ok, even if there is short somewhere, it can't burn sth out
: within 1/5 or 1/2 sec.
: voltage is better be sensed faster.
| l***g
发帖数: 1035 | 12 the purpose is to detect over-under voltage on the cell to shutdown control
fast. with the delay, that purpose is defeated, and the 4 samples is not
important if they are delayed samples.
【在 g****t 的大作中提到】
: 按我写的作,效果就是电压测量频率是足够快的,但是发送延迟了1/5秒.
: 就是搞个二级缓存.
:
: temperature is ok, even if there is short somewhere, it can't burn sth out
: within 1/5 or 1/2 sec.
: 电池倒是有可能温度急剧上升,然后爆炸.但那个需要保险丝或者
: PTC来保险.控制loop用的温度,我认为1/20秒完全不make sense.
: 又不是核电站.
: voltage is better be sensed faster.
| p***e
发帖数: 472 | 13 separate the transmission of temperature and voltage to two completely
separate communication channels. this way you may be able to double the
rate of each channel to 10Hz.
sampl
p
5
somethin
included
【在 l***g 的大作中提到】
: hope i can pick up some brains here. so i have 100 batteries, each one sampl
: es a temperature and a voltage. i want to send all info to a center bms to p
: rocess. after all, that is what a bms does.
: most existing schemes are using CAN or Serial, throughput is pretty low as 5
: Hz or worse. Can you think of a way of sending these data faster to somethin
: g like 20Hz? keep in mind that some overhead information has to be included
: like the ID of each data point and timestamp.
: constraints: communication has to be bidirectional.
| l***g
发帖数: 1035 | 14 separating current transmission into 2 channels will have a 4 fold improveme
nt. yes that is one way doing it but a bit more expensive.
【在 p***e 的大作中提到】
: separate the transmission of temperature and voltage to two completely
: separate communication channels. this way you may be able to double the
: rate of each channel to 10Hz.
:
: sampl
: p
: 5
: somethin
: included
| p***e
发帖数: 472 | 15 agree, especially the cost and connection harness with wires...
i think the sampling rate of temperature can be sacrificed a bit, so maybe
you don't have to transmit temperature every reading, don't know if you
local DAQs have this functionality.
for voltage measurement, you said before to shut down the cells/modules when
it's read too low. are the cells/modules in series or parallel? maybe you
can group some of the voltage readings together based on the circuit
configuration. like in series, once you open one, the entire circuit will
shut down anyway.
【在 p***e 的大作中提到】
: separate the transmission of temperature and voltage to two completely
: separate communication channels. this way you may be able to double the
: rate of each channel to 10Hz.
:
: sampl
: p
: 5
: somethin
: included
| l***g
发帖数: 1035 | 16 thanks for the insight.
when
you
【在 p***e 的大作中提到】
: agree, especially the cost and connection harness with wires...
: i think the sampling rate of temperature can be sacrificed a bit, so maybe
: you don't have to transmit temperature every reading, don't know if you
: local DAQs have this functionality.
: for voltage measurement, you said before to shut down the cells/modules when
: it's read too low. are the cells/modules in series or parallel? maybe you
: can group some of the voltage readings together based on the circuit
: configuration. like in series, once you open one, the entire circuit will
: shut down anyway.
|
|
