Pet IC Meeting Minutes written by Kathryn Ewell June 1, 1999 Attending: K.Ewell, M. Harrison, B.Krieger, E.Mandelli, G.Meddeler, O. Milgrome, B. Moses, M. Pedrali-Noy, V. Rosso Marzio: It looks like the WTA chip is now able to be tested. Valeria: In the MUX mode, the max frequency working and tested was 1MHz. If input is outside the range of [-200mV, -2.2V] output unable to follow. We did not test it higher than 1MHz. The goal is that It should pass peaking time without distortion. I will check with the spectrum analyzer. In the WTA mode, the output does not follow the sinusoidal input. As the sinusoidal input goes from the peak to the DC value, the output does not follow. The output instead remained at a steady value. We were able to reduce the step to go to the WTA value at the time that the sinusoid reaches its trough. At this early stage, we are happy just that the chip is testable. Marzio: The measurement Valeria ran were on Friday. I ran simulation where the setup had 14 channels running at DC 1.6V, 1 channel at DC 1.65 V, and 1 channel with DC sinusoidal with amplitude .5V and mean value 1.6V. There is an inital delay for the WTA output to switch over to the sinusoidal input and a longer, step-like time for the WTA output to switch back to the 1.65V output once the sinusoidal is descending. This flat period as the sinusoidal falls corresponds to the circuit not selecting any input channel as the winner. The tests since Friday were run with a 100 microsecond period, while my simulations were run with a 1 microsecond period. Possibly the outputs since Friday can be seen as a magnification and distortion of the simulated waveforms. Bill: I believe that in the previous versions of the WTA having no winner was a symptom of I_fight being too small. Marzio: We should try to adjust I_fight. The DC measurement shows a possible shift of the of p-channel devices threshold of several tenths of millivolt. Brad: In my FE-5 Chip All the biases came out, but I did not notice the threshold shifts. Valeria: The amplitude of the flat period is approximately 50mV for the on-off loss of the WTA signal Marzio: This is not really essential for the functionality of the chip. Bill : It was the rise time we focused on last time. It was around 30 ns with the larger amplitude signal. Valeria: I_fight and I_master were not changed at the same time during this measurement. Bill: We can use this "no winner" output condition to check the error bits. We should have an error bit that flags during this condition. Bill: (From Greg - Handout) This is in regards to the pre-amp and shaper amp. We have a few problems. 1) The Preamp had ringing. We changed Vp to eliminate ringing and we were successful. 2) Shaper amp also had ringing because the capacitive load was higher than expected. The capacitive load was 50-100pF (twisted pair). We added a 1k series resistor and successfully knocked down the ringing. 3) We want good noise performances. There are three major causes of noise, Johnson noise, which we'll term series noise, shot noise or leakage current, which we term parallel noise, and the constant noise proportional to 1/f. We have had problems with parallel noise in the past. During this series of tests it was so low, so close to the theoretical level that it made us uncomfortable. We then compared it to some real imaging measurements to see how narrow the peaks could be. It did quite well. We were still puzzled by some series noise. We tried both the Hammamatzu photodiode and Nadine's photodiode. Nadine's had worse series noise, but roughly the same parallel and 1/f noise. 4) We see consistent overshoot even with the 1k resistor nad the proper V_p. It could be the fault of the test board/layout design. This board was made for flexibility and has more capacitance and antenaa than desired. We will check the results of Brad's verification loaded to see how that effects the return to base. Oren: How is the set-up calibrated? Bill: We begin with a long fall time and measure 10-90% rise time. We then set a fast rise time and measure 10-90% fall time. We then set those particular rise and fall times, 1 microsecond and 3 microsecond respectively. They are usually accurate to within 10%. The longest reset time in the preamp it is stable to is 50 microseconds. WE usually have a 400ns minimum rise time, at 1-2 microseconds, the noise levels off to the 1/f noise and we measure out until 10-20 microseconds, enough time to get the whole curve. You have a hard time with a long measurement in seeing the parallel noise. The measurement is as good as previous amps, a good measurement. Eman: I ran a simulation on the FE-5 . I added some resistance for the power supply of the pre-amp and shaper. I did this simulation to determine the impact of metal wiring and crosstalk between the preamp and shaper. I did a simulation without using the digital selection for gain, there were good results with ringing (I did not move the Vp point). The falling time regulated well. The shaper output: the waveform is ok unless the gain is too small. The shaper is gaining by a facotr of 2, that is too low. When I tried to add gain, I ran into a problem. With more current in the shaper fold, I find strange results in the preamp output. I will try to isolate the problem and look at crosstalk. Bill: Usually we wait to change the digital selection for gain. I am comfortable with waiting five seconds after the gain is changed. Gerrett: I finished the functional description I changed one of the commands; It gives the same results with more information. You can still use it in the same mode. Bill: We should go over the functional description. Gerrett: Yes, to check for missing states. Monica: Who is working on debugging the non-working digital circuit right now? Bill: Gerrett is working on a new top-down model in VHDL. Oren: Monica can run the test-vectors on it when Gerrett is finished with the VHDL. Bill: There were some mistakes in the circuit that we can learn from. ********** The meeting is adjourned. ***************