2.2.2 源端端接的上升时间
从线路上任何一点向源端看去,驱动阻抗都是Z0;当驱动容性负载的时候,得到的响应看起来像个RC滤波器。时间常数t=Z0C;上升时间Tr=2.2倍时间常数=2.2Z0C。对比末端端接时的上升时间Tr=2.2t=1.1CZ0,源端端接的上升时间是末端端接电路的2倍。
2.2.3 源端端接电阻的位置[15]
源端端接的电阻离驱动器越近越好,这样可以减小驱动器和端接电阻之间连线的影响。但是很多情况,特别是BGA的情况,电阻必须放在封装外边比较远的地方,这时,距离就构成了一个问题,我们希望知道,在多大的尺寸之内,源端端接还是有效的。
我们假设驱动器和端接电阻之间的stub长度小于1/3上升时间的长度。The connection stub, because it is connected at one end to a low-impedance driver, acts like a little inductor L_STUB.
在频率f,电感的电抗是(j*2*pi*f)L_STUB,因此,端接变成不准确的了。这个电感和端接电阻串联得到R+ Z(L_STUB)。可以预测这样一个阻抗产生的反射——1/2(VSTEP)(LSTUB/Z0)(1/TR), where VSTEP is the step amplitude, Z0 is the transmission line impedance and TR is the 10-90% risetime of the step waveform.可以参考[1],公式4.74.
That’s the theory, except for these embellishments:
1. The stub inductance may be calculated as LSTUB = DLY*Z0, where DLY is the delay of the stub in seconds and Z0 is the stub impedance.
2. Add to the stub inductance the parasitic series inductance of the driver package, LPACKAGE.
3. The stub affects the risetime of the first incident waveform by a tiny amount. Keep the stub delay less than 1/3 of the risetime and you will hardly see this effect. (Thanks to Tom Giovannini and Joe Cahill for reminding me to mention this).
Example: BGA package, LPACKAGE = 6000 pH, with an ideal 70-ohm series terminator located 1/2 inch (microstrip trace) from the driver package. Assume we have a 3.3-v driver with a 1-ns risetime. In this case LSTUB= 1/2(145 ps/in)(70ohm ) = 5075 pH. The total inductance L TOTAL = LPACKAGE + LSTUB = 11075 nH. The reflected signal: Refl. = 1/2(3.3) [(11075/70) / 1000 ] = 261mV.
If you want 20dB or more of reflected-wave attenuation, use a stub delay of no more than 1/6 the risetime, a very good low-inductance package, and an accurate carbon-composition or low-inductance, non-etched metal film resistor.
所以,保证驱动器和端接电阻之间的距离<1/6(Ltr),端接引起的反射就会很小。实际上,这是一个很宽松的条件,按1ns的上升时间计算,1/6上升沿长度将达到1in,实际上,无论如何我们也不会把电阻放在那么远的位置上。
2.3 末端端接的交流偏置(AC termination)
由于末端端接和split端接有一个到地的直流通路,因此增加了功耗。经常在末端端接使用一个电容来减少静态功耗。
2.3.1 功耗计算
如果驱动电路处于HI和LO的状态的时间基本相当(直流平衡电路),电容器C1上的电压的平均值将是HI和LO的平均值,那么,电阻R1的功耗是:(V/2)^2/Z0=V2/4Z0。与分离端接相比较,电阻两端的压降是V,功耗是V2/2Z0。浪费的功耗源于Vcc到地的电流。
V是逻辑高和逻辑低的差值。其实,驱动电路的功耗是相同的,只是端接电阻的功耗不同。
如果信号是直流不平衡的,比如,电路处于HI的状态的时间很长,那么,电容就会被充电到V;当电路跳变到LO时,V将全部加在R的两端,驱动电流是V/R,这是直流平衡时的2倍。所以我们就很矛盾,如果驱动器能够提供这么大的电流,为什么不干脆端接到地或者Vcc呢;如果不能提供这么大的电路,为了让容性端接正常工作,我们就需要直流平衡的信号!
[4]中讲到了AC端接时电阻和电容的选择:The resistor R should be equal to the characteristic impedance of the transmission line, and the capacitor C should be chosen such that the RC time constant at the load is approximately equal to one or two rise times. It is advised that simulations be performed to choose the optimum capacitor value for the specific design。
这样端接的优点是减少了直流功耗,缺点是,电容会减慢上升或下降时间,从而增加信号的延时。AC端接的一个潜台词就是希望能找到一个大的电容,完成良好的端接;同时,希望电容足够小,以便从源端不吸收太多的电流[10]。
对于长线,(line delay)/(signal risetime)is greater than 1,无法找到这样的电容,
对于短线,(line delay)/(signal risetime)is 1/3or 1/2,AC端接才会有效。
With DC-balanced data you are free to pick any capacitor BIG ENOUGH to hold its charge throughout the bit period。
文章评论(0条评论)
登录后参与讨论