buck, 24V input, 12V output. PIC18F458, 40KHz, 采样周期50us,即每两个PWM周期采样并更新duty cycle.PID控制,基于查表法,用了各种办法,最短的PID计算时间为30us左右,没能小于25us.有比较小的cycle limit(2003 jan IEEE power electronic 有一篇文章专门提到了这个问题).变负载和输入电压改变,输出均能稳定在12V,下一步看看能不能改进算法.
看了很多关于关于数控电源的paper,收获很大.看看这次能不能搞出点名堂.
5月份参加了TI的一个seminar, 专门有一个section就是数控电源.但我听了之后,发现其实这个数控电源并不是真正意义上的纯数控电源.单片机仅仅monitor 和通讯而已,控制部分还是模拟的type II.这是TI最新的seminar,由此可看出工业界的数控电源的现状.
个人感觉,纯数控电源完全取决于硬件.用DSP虽然快,但成本显然不能接受.AUBURN UNIVERSITY 的paper用TIF240DSP控制buck,显然实际应用是步可能的.18F458很简单好用,但10位A/D最快要19us,显然又限制了电源的频率.
microchip的dsPIC, 2010系列,普通的28pinDIP封装,16位PWM, 10位A/D只用2us,可能会是很好的选择,下一步准备用,看看能把频率提高到多少.
我的第一个纯数控电源今天成功了
全部回复(52)
正序查看
倒序查看
现在还没有回复呢，说说你的想法
@一个好人
wang兄,恭喜,恭喜!!!能不能传一张图上来,先谢了,代表大家先谢了,谢谢
惭愧惭愧,跟各位DX相比,我只是刚刚入门不久.硬件本身很简单,buck加上反馈和控制输出.简单得不好意思把电路贴出来.这几天希望能改进改进算法,提高暂态性能.比如,提高到50khz, 这样每40us更新一次.改进PID的参数.更正一下,需要30us计算新的duty cycle, include 19us A/D,18us PID table lookup for 3 parameters. one of parameter is obtained during A/D convertion. so totally, it takes about 30-34us to get a new duty cycle.if the direct multiplication is used to get new duty cycle, like u(k)=Kp*error(K)+Ki*sum(error)+Kd*(error(k)-error(k-1)),it takes 8*25us to get a new duty.Obviously, it is bad, too long delay to update the duty cycle.this is why I employ look up table, and the price is relatively bad transient reponse.
In UC power electronic center, they made a digital controller to buck converter at 1MHz, but they made their specific hardware, include high frequency DPWM chip, high speed A/D chip and DSP.
as for my case, I use very popular PIC18F458, running at 50KHz.
In my opinion, the improvement of algorithm doesn't make any sense to improve the transient reponse.The small difference disappears then the floating numbers are changed to interger DPWM. if a higher swtiching frequency, higher sampling frequency and shorter PWM gerneration, can be achieved, it improves the transient reponse greatly.
In UC power electronic center, they made a digital controller to buck converter at 1MHz, but they made their specific hardware, include high frequency DPWM chip, high speed A/D chip and DSP.
as for my case, I use very popular PIC18F458, running at 50KHz.
In my opinion, the improvement of algorithm doesn't make any sense to improve the transient reponse.The small difference disappears then the floating numbers are changed to interger DPWM. if a higher swtiching frequency, higher sampling frequency and shorter PWM gerneration, can be achieved, it improves the transient reponse greatly.
0
回复
提示
@maychang
恭喜恭喜!前几天我的modem被雷击坏,一个星期不能上网,来迟了.TI也还是用模拟的typeII做控制,看来纯数控电源确实不容易.microchip的dsPIC,PWM的频率是多少?一般PWM不可能超过单片机时钟,那么该芯片的时钟最大是多少?请Wang兄介绍介绍.
thanks, man.
I jusr ordered some dsPIC samples and learning how to program with C.I think it should be fine for high frequency switch , above 100khz.
it seems 8 bit PWM for 150kHz for PIC18F458 at 4MHz, look, the impossible for PIC18F458 to finish the A/D and calculation at that short time even it can provide 150Khz, 8bit PWM.
but I think dsPIC can do it. we will see..
I jusr ordered some dsPIC samples and learning how to program with C.I think it should be fine for high frequency switch , above 100khz.
it seems 8 bit PWM for 150kHz for PIC18F458 at 4MHz, look, the impossible for PIC18F458 to finish the A/D and calculation at that short time even it can provide 150Khz, 8bit PWM.
but I think dsPIC can do it. we will see..
0
回复
提示
@gunking
thanks,man.IjusrorderedsomedsPICsamplesandlearninghowtoprogramwithC.Ithinkitshouldbefineforhighfrequencyswitch,above100khz.itseems8bitPWMfor150kHzforPIC18F458at4MHz,look,theimpossibleforPIC18F458tofinishtheA/Dandcalculationatthatshorttimeevenitcanprovide150Khz,8bitPWM.butIthinkdsPICcandoit.wewillsee..
Thanks.
We are waiting for you...
We are waiting for you...
0
回复
提示
@cmg
从你的描述中可看出单片机控制是有很多问题存在的,响应速度和保护问题就很难解决,我也很关注数字电源,拒我了解现在情况是这样:用单片机实现智能保护和通信功能的比较多,但不能算真正的数字电源,TI有一款DSP是做纯数字电源的,全部数字化,型号忘记了,IWATT有一款数字控制IC,它把类似CPU的东东集成到IC里面去了,反映速度超快,这是目前唯一的一款真正的实际可用的数字电源.
cmg兄也来了!
我也感觉纯数字电源很难,原因之一就是上面这位说的程序跑飞.如果跑飞,必定烧功率管.用单片机作显示、通讯等等都没有问题,唯独单片机PWM直接控制不好解决.
希望cmg兄以后多多介绍芯片和电路动态.
我也感觉纯数字电源很难,原因之一就是上面这位说的程序跑飞.如果跑飞,必定烧功率管.用单片机作显示、通讯等等都没有问题,唯独单片机PWM直接控制不好解决.
希望cmg兄以后多多介绍芯片和电路动态.
0
回复
提示
