The first time I played with FPGA applications, I played with the TSL1401 linear array sensor

littleshrimp

The first time I played with FPGA applications, I played with the TSL1401 linear array sensor [Copy link]

TSL1401 is a 128-point linear array sensor with a main frequency of up to 8MHz. Some time ago, I drew a board and planned to use FPGA to make it.

Almost all the information that can be found about TSL1401 is in Chinese, most of which are used for tracking cars and are driven by microcontrollers, but there are very few FPGAs.

Because of the limitation of AD sampling rate of single chip microcomputer, the clock frequency will not be too high.

I want to get a linear array with a higher clock frequency, but some professional ones are too expensive. In addition to the sensor, they also need a large professional lens. The whole thing costs less than ten thousand, which is unaffordable.

There is an ELIS-1024 that looks pretty good in terms of parameters, but unfortunately I couldn't buy it after contacting many places. In the end I decided to practice with this TSL1401.

In order to run at 8MHz clock, I started to draw the AD807E 12-bit 53MSPS ADC board. Maybe there was a problem with the LDO. It smoked when powered on and burned 2 ADCs.

Later, I changed to AD9245-65, a 14-bit 65MSPS ADC. I bought EBAZ4205 some time ago, which I think is very good and very suitable for those of us who want to get started with FPGA.

After working with FPGA for a few days, I was finally able to read the data of TSL1401, but the effect was still far from what I expected.

At the beginning, I just used FPGA to generate a control sequence to make TSL1401 work at 8MHz clock, and then used an oscilloscope to measure the analog output.

From the oscilloscope, the output noise is very large. Because under normal circumstances, the AO output will have a jump every time the rising edge of the clock transfers pixels, so using ADC to collect signals on the falling edge should seem to be more effective.

Then I spent a long time to read the data from AD9245-65 and directly displayed the data through ILA. At 8MHz, it looked better than the oscilloscope, but it was basically useless. The saw teeth were still one by one when sawing wood. Later, the effect was much better when the clock was reduced to 2.5MHz.

The "compile" speed of VIVADO is really slow. It takes more than ten minutes to compile a little bit of code. The CPU basically runs at about 15%. I don't know where the time goes.

It may be that the integration time is too short at 8MHz, and the noise is also amplified after being amplified by the amplifier. Although the waveform of the 2.5MHz clock looks good, the frequency is too low. The frame rate is only 19.5K when running at 128 pixels, which is not enough to do anything.

w494143467

Boss, this place has everything. I'm so jealous.

freebsder

What is so expensive?

littleshrimp

freebsder posted on 2020-11-17 21:34 What is so expensive?

A second-hand line scan camera with a lens from a certain fish brand costs several thousand yuan, and a Roderick's lens costs one or two thousand yuan.

freebsder

littleshrimp posted on 2020-11-17 22:33 A second-hand line scan camera with a lens from a certain fish costs several thousand yuan, and a Rod Gang lens costs one or two thousand yuan.

Boss, is this for your job or are you doing it for yourself? Damn, I can't afford this price...