Professional Documents
Culture Documents
∗ Corresponding author at: School of Electronics and Information, Hangzhou Dianzi University, Hangzhou, 310018, PR China
E-mail address: yuzhang1978@163.com (Y. Zhang).
https://doi.org/10.1016/j.nima.2018.07.041
Received 18 December 2017; Received in revised form 18 May 2018; Accepted 16 July 2018
Available online 21 July 2018
0168-9002/© 2018 Elsevier B.V. All rights reserved.
Y. Zhang et al. Nuclear Inst. and Methods in Physics Research, A 904 (2018) 171–178
2. Proposed model for prediction of number of transistors a basic priority encoder block is shown in Fig. 2. The definitions of all
signals in basic priority encoder block are shown in Table 1.
The model for predicting the number of transistors in AERD circuit The relationship between 𝑖 and 𝑛 is: 2𝑛 = 𝑖. The structures of standard
comprises two major parts: predicting the number of transistors in intra- CMOS gate are shown in Fig. 3.
layer and inter-layer circuits, respectively. The intra-layer circuit is the The output signals of the basic priority encoder block include three
circuit in every layer, while the inter-layer circuit is the circuit between parts: valid, reset, and address. Three part of circuits produced by these
adjacent layers. three output signals and corresponding numbers of transistor are shown
as follows:
2.1. Model for predicting the number of transistors in a basic priority encoder (1) valid circuit
block Because valid=state[0]+state[1]+⋯state[𝑖 − 1], according to the
structure of a standard CMOS gate, as shown in Fig. 3, one bit input
Basic priority encoder block is the main component of intra-layer variable needs a corresponding pair of P-MOS and N-MOS transistors in
circuit of AERD circuit. In the same layer, the bit-width of the input the NOR logic gate, and there are two transistors in one Inverter logic
states of every basic priority encoder block is the same. The structure of gate. As a result, the NOR logic gate with 𝑖 bit input variables needs 2 ∗ 𝑖
172
Y. Zhang et al. Nuclear Inst. and Methods in Physics Research, A 904 (2018) 171–178
Table 1
The definitions of all signals in basic priority encoder block.
Signal name Direction Bit-width Function
1 state Input 𝑖 It is high when there is a hit in the corresponding pixel.
2 valid Output 1 It is active if at least one pixel is hit.
3 reset Output 𝑖 It resets the pixel state register on falling edge.
4 addr Output 𝑛 It represents the address of the hit pixel that is being reset.
5 select Input 1 It enables the basic priority block to transport the hit pixel.
Fig. 4. Truth table of a 4–2 priority encoder. Fig. 6 is the truth table of an 8–3 priority encoder, and X means the
arbitrary value.
By using the similar Karnaugh map simplification in Fig. 5, the logic
relationship between the input state and address can be obtained by (2).
transistors, and it needs one Inverter gate to realize OR logic, then the
number of transistors of the valid circuit is 2 ∗ 𝑖 + 2. 𝑎𝑑𝑑𝑟[0] = 𝑠𝑡𝑎𝑡𝑒[1] + 𝑠𝑡𝑎𝑡𝑒[0] + 𝑠𝑡𝑎𝑡𝑒[3] + 𝑠𝑡𝑎𝑡𝑒[2] + 𝑠𝑡𝑎𝑡𝑒[1] + 𝑠𝑡𝑎𝑡𝑒[0]
(2) Reset decoder circuit +𝑠𝑡𝑎𝑡𝑒[5] + 𝑠𝑡𝑎𝑡𝑒[4] + 𝑠𝑡𝑎𝑡𝑒[3] + 𝑠𝑡𝑎𝑡𝑒[2] + 𝑠𝑡𝑎𝑡𝑒[1] + 𝑠𝑡𝑎𝑡𝑒[0]
Since, +𝑠𝑡𝑎𝑡𝑒[7] + 𝑠𝑡𝑎𝑡𝑒[6] + 𝑠𝑡𝑎𝑡𝑒[5] + 𝑠𝑡𝑎𝑡𝑒[4] + 𝑠𝑡𝑎𝑡𝑒[3] + 𝑠𝑡𝑎𝑡𝑒[2] + 𝑠𝑡𝑎𝑡𝑒[1] + 𝑠𝑡𝑎𝑡𝑒[0]
𝑎𝑑𝑑𝑟[1] = (𝑠𝑡𝑎𝑡𝑒[2] + 𝑠𝑡𝑎𝑡𝑒[3]) 𝑠𝑡𝑎𝑡𝑒[1] + 𝑠𝑡𝑎𝑡𝑒[0]
reset[0] = select&state[0] = select + state[0]
+𝑠𝑡𝑎𝑡𝑒[5] + 𝑠𝑡𝑎𝑡𝑒[4] + 𝑠𝑡𝑎𝑡𝑒[3] + 𝑠𝑡𝑎𝑡𝑒[2] + 𝑠𝑡𝑎𝑡𝑒[1] + 𝑠𝑡𝑎𝑡𝑒[0]
reset[1] = select&state[1]&state[0] = select + state[1] + state[0]
reset[2] = select&state[2]&state[1]&state[0] 𝑎𝑑𝑑𝑟[2] = 𝑠𝑡𝑎𝑡𝑒[3] + 𝑠𝑡𝑎𝑡𝑒[2] + 𝑠𝑡𝑎𝑡𝑒[1] + 𝑠𝑡𝑎𝑡𝑒[0] (2)
= select + state[2] + state[1] + state[0] The number of the red square in Fig. 5 means the number of items in
reset[3] = select&state[3]&state[2]&state[1]&state[0] the logic expression of addr in Fig. 4, and the corresponding positions of
these items are shown as solid red circles in Fig. 4. The similar regulation
= select + state[3] + state[2] + state[1] + state[0]
can be seen in Fig. 6. According to the Karnaugh map simplification, the
……
conclusion that can be drawn from Fig. 4 to Fig. 6 is that there are 𝑖∕2𝑛+1
reset[𝑖 − 1] = select&state[𝑖 − 1]&state[𝑖 − 2]& ⋯ &state[0] items in the logic expression of addr[𝑛]. As a result, in the light of the
= select + state[𝑖 − 1] + state[𝑖 − 2] + ⋯ + state[0] above linear relationship between the input variable and the number of
transistor in CMOS standard gate circuit, the number of transistors of
According to the structure of the standard CMOS gate, the NOR logic
address[𝑛 − 1:0] can be obtained as follows:
gate with 𝑖 bit input variables needs 2 ∗ 𝑖 transistors, and there are
Firstly, the number of transistors of addr[0]:
two transistors in one Inverter logic gate. As a result, the number of
transistors of the reset signal can be calculated by: 1 + 2𝑛 + 𝑖 𝑖 𝑖 𝑖
× ×2+2× +2× +2
2 2𝑛+1 2𝑛+1 2𝑛+1 (3)
1
3 ∗ 2 + 2 ∗ 2 + 3 ∗ 2 + 2 ∗ 2 + 4 ∗ 2 + 2 ∗ 2 + 5 ∗ 2 + ⋯ 2 ∗ 2 + 2 ∗ (𝑖 + 1) 𝑛 = 0 × 𝑖2 + 3 × 𝑖 + 2
( ) 2
(𝑖 + 1 + 3) [ ]
= 6 + 4 ∗ (𝑖 − 1) + 2∗ ∗ (𝑖 − 1) = 𝑖2 + 7𝑖 − 2 Secondly, the number of transistors of addr[𝑛], 𝑛 ∈ 1, log2 (𝑖) − 2 :
2
( )
(3) Address encoder circuit 𝑖 𝑖 𝑖 𝑖2 + 8𝑖
𝑖× ×2+ −1 ×6+2× +2= −4 (4)
2 𝑛+2 2𝑛+1 2𝑛+1 2𝑛+1
In order to deduce the logic relationship between the input variables
∑log (𝑖)−2
and address of a basic priority encoder, a 4–2 and 8–3 priority encoder It can be deduced from (4), the number of transistors of 𝑛=12
are employed to set as the example used for analysis. Fig. 4 is the truth 𝑎𝑑𝑑𝑟[𝑛] can be obtained by (5).
( )
table of the 4–2 priority encoder, and X means the arbitrary value. The (2 ) 1 1 1 ( )
Karnaugh map of addr[0] and addr[1] in Fig. 4 are shown in Fig. 5. 𝑖 + 8𝑖 + +⋯+ − 4 × log2 (𝑖) − 2
4 8 log
2 2 (𝑖)−1
173
Y. Zhang et al. Nuclear Inst. and Methods in Physics Research, A 904 (2018) 171–178
174
Y. Zhang et al. Nuclear Inst. and Methods in Physics Research, A 904 (2018) 171–178
(𝑘−1 )
( ) ∑
num_bk_l (𝑘) × 2 × 2addr_wid_l(𝑘) + 2 × addr_wid_l (𝑗) (11)
𝑗=1
sum_tran_t =
max _layer ( (
∑ ( )2
num_bk_l (𝑘) × 2 × 2addr_wid_l(𝑘)
𝑘=1 ( ) ))
+14 × 2addr_wid_l(𝑘) − 4 × addr_wid_l (𝑘) − 6
( ( 𝑘−1 ))
∑
max _layer
( ) ∑
+ num_bk_l (𝑘) × 2 × 2addr_wid_l(𝑘) + 2 × addr_wid_l (𝑗)
𝑘=1 𝑗=1
(12)
175
Y. Zhang et al. Nuclear Inst. and Methods in Physics Research, A 904 (2018) 171–178
Fig. 10. Simulation result when the number of input pixels is 484.
3.3. Comparison result with FPGA implementation Artix-7 FPGA of Xilinx, as shown in Fig. 11. The results in Fig. 11 indi-
cate that the variation trend of prediction results of the proposed model
is similar to that of the post-layout simulation results implemented
The proposed method is also assessed by conducting a comparison by Xilinx FPGA. The optimal solutions with the minimum number
with the FPGA implementation, which are simulated in Vivado 2016 and
176
Y. Zhang et al. Nuclear Inst. and Methods in Physics Research, A 904 (2018) 171–178
Fig. 11. Comparisons with Cadence post-layout simulation and FPGA implementation.
Table 2
Number of transistors of Cadence post-layout simulation of an AERD circuit with 1024 input pixels.
Circuit components Count Number of transistors in component
Nr23d4 1 1*14=14
Bufbd7 5 5*16=80
Invod1 84 84*2=168
Bufbd3 10 10*8=80
Buffd7 24 24*13=312
Nr02d1 80 80*4=320
An03d1 768 768*8=6144
Buffd3 32 32*6=192
An12d4 1 1*12=12
Or02d1 339 339*6=2034
Bufbda 41 41*20=820
Bufbd4 1 1*9=9
Nd23d4 1 1*14=14
Aor31d4 2 2*14=28
Buffda 1 1*19=19
An02d4 7 7*10=70
An03d2 168 168*10=1680
An12d1 340 340*8=2720
Nd12d2 80 80*10=800
Nr02d7 1 1*16=16
An02d1 424 424*6=2544
Or04d7 224 224*20=4480
Inv0d0 256 256*2=512
Aor31d2 82 82*12=984
Nr02d4 4 4*12=48
An02d2 586 586*8=4688
An03d4 256 256*12=3072
Nr02d0 256 256*4=1024
Or02da 1 1*19=19
Or02d4 2 2*10=20
Ora211d4 256 256*14=3584
Nd12d4 260 260*14=3640
Total number of transistor 40174 − 1512(delay buff)=38635
transistors under proposed estimation and FPGA implementation are circuit. This prediction model can output all the different probable
same, and are 2_2_2 structure. According to the literature [12], one LUT structures of an AERD circuit, and highlight the structure that costs the
in FPGA equals approximately 21 transistors. As a result, it can be seen minimum number of transistors. The experimental results showed that
from Fig. 11 that the number of transistors estimated by the proposed the number of transistors obtained by the proposed prediction model
prediction model is of the same order of magnitude as the approximate approaches the Cadence post-layout simulation result and FPGA im-
number of transistors in the post-layout results of FPGA implementation. plementation. The proposed prediction model can supply the candidate
reference for an ASIC designer and high-energy physicist to select the
4. Conclusion appropriate structure for a digital AERD circuit in MAPS to obtain the
minimum implementation area, according to the number of input pixels
of different high-energy particle physics experiments.
This paper proposes a model for predicting the number of transistors
of an asymmetrical priority address-encoder and reset-decoder readout
177
Y. Zhang et al. Nuclear Inst. and Methods in Physics Research, A 904 (2018) 171–178
178