A
Anonymous
Guest
Hi all,
I have not checked the PI classroom in the last few weeks and it looks as if things have really taken off! I will try to get up to speed with you guys and maybe even contribute as time allows.
So one idea I had thought about a while ago, and one that Eric and maybe others have already messed with, is the use of a log amp in the front-end. The exponentially decaying pulse response presents a lot of dynamic range and a log amp seems to be a good approach in making the signal more palatable to either analog processing or to a high-speed sampling ADC for digital processing.
Analog Devices (with whom I design high-speed ADCs) makes several potential candidates. The AD8307 has 92dB of dynamic range and goes DC-500MHz. There are other similar parts such as the AD8310 and the AD606. I have not thoroughly dug through the data sheets to see what the differences are, but it appears to be mostly BW and dynamic range. I do know that the AD6xx parts are a little older, and the AD83xx parts are the latest.
One nice thing about the PI waveform is that it has very predictable behavior versus time: it always decays. ADI also makes variable gain amplifiers that have "linear-in-dB" gain control; that is, they have log-vs-time gain, basically a time gain amplifier (TGA). The AD603 is one such part, and the AD8307 data sheet has an apps note that shows the 603+8307 set up as a 120dB TGA. Also note that both parts have good noise performance, each about 1.5nV/rtHz.
To me, this seems to be the way to go, but I would like others to take a look at it and post their thoughts. Visit the ADI link below and look at the VGA section and the log amp section, see what's available, and what might be a good solution to try. Once we figure out the best approach, I will be glad to get a handful of samples and send some out to people who are seriously in a prototyping stage.
Finally, I would like to hear some thoughts on ADC requirements for doing DSP analysis. Without a logamp, I think the ADC would need a lot of dynamic range, maybe 14-16 bits. With the logamp front-end, an 8-bit converter should be plenty. But what about input BW and sample rate? BW appears to me to be rather low, an exponential decay in 20us does not present much frequency info beyond a few hundred kHz I think. And how many samples would you want in the decay window? Depends on the processing you want to do no doubt, so I'm looking for ideas.
As a datapoint on available ADCs, a design I worked on was released yesterday: 14 bits with 65MHz sample rate. The problem with it (for detectors) is power: 1.3 watts @ 5v! That's one reason why I think the logamp front-end is important, 8-bit ADCs are in the 100mW ballpark, and cheap.
I have not checked the PI classroom in the last few weeks and it looks as if things have really taken off! I will try to get up to speed with you guys and maybe even contribute as time allows.
So one idea I had thought about a while ago, and one that Eric and maybe others have already messed with, is the use of a log amp in the front-end. The exponentially decaying pulse response presents a lot of dynamic range and a log amp seems to be a good approach in making the signal more palatable to either analog processing or to a high-speed sampling ADC for digital processing.
Analog Devices (with whom I design high-speed ADCs) makes several potential candidates. The AD8307 has 92dB of dynamic range and goes DC-500MHz. There are other similar parts such as the AD8310 and the AD606. I have not thoroughly dug through the data sheets to see what the differences are, but it appears to be mostly BW and dynamic range. I do know that the AD6xx parts are a little older, and the AD83xx parts are the latest.
One nice thing about the PI waveform is that it has very predictable behavior versus time: it always decays. ADI also makes variable gain amplifiers that have "linear-in-dB" gain control; that is, they have log-vs-time gain, basically a time gain amplifier (TGA). The AD603 is one such part, and the AD8307 data sheet has an apps note that shows the 603+8307 set up as a 120dB TGA. Also note that both parts have good noise performance, each about 1.5nV/rtHz.
To me, this seems to be the way to go, but I would like others to take a look at it and post their thoughts. Visit the ADI link below and look at the VGA section and the log amp section, see what's available, and what might be a good solution to try. Once we figure out the best approach, I will be glad to get a handful of samples and send some out to people who are seriously in a prototyping stage.
Finally, I would like to hear some thoughts on ADC requirements for doing DSP analysis. Without a logamp, I think the ADC would need a lot of dynamic range, maybe 14-16 bits. With the logamp front-end, an 8-bit converter should be plenty. But what about input BW and sample rate? BW appears to me to be rather low, an exponential decay in 20us does not present much frequency info beyond a few hundred kHz I think. And how many samples would you want in the decay window? Depends on the processing you want to do no doubt, so I'm looking for ideas.
As a datapoint on available ADCs, a design I worked on was released yesterday: 14 bits with 65MHz sample rate. The problem with it (for detectors) is power: 1.3 watts @ 5v! That's one reason why I think the logamp front-end is important, 8-bit ADCs are in the 100mW ballpark, and cheap.