Analog to digital converter confustion

[EricVanWyk]

For ADCs, I tend to just go to analog's webpage directly (www.analog.com). They have a rather decent interface for selecting an ADC, and they tend to have decent prices. Even better, they'll sample you a handful of just about anything!

Could you post your specs for us?

How many inputs?
How many bits of resolution?
How many samples per second?
What type of digital interface?
Voltage requirements?

[Adam Y.]

Originally Posted by EricVanWyk

For ADCs, I tend to just go to analog's webpage directly (www.analog.com). They have a rather decent interface for selecting an ADC, and they tend to have decent prices. Even better, they'll sample you a handful of just about anything!

Could you post your specs for us?

How many inputs?
How many bits of resolution?
How many samples per second?
What type of digital interface?
Voltage requirements?

The ability to measure 5 volts and 20MSPS are the only requirments I really care about. I'm pretty sure that Im botching something up or looking at the wrong specification because I can find the 20MSPS but I can't find the five volts. Im assuming that 2Vp-p means 2 volts peak to peak. Im even apt to using a DSP.

[Kevin Sevcik]

Originally Posted by Adam Y.

The ability to measure 5 volts and 20MSPS are the only requirments I really care about. I'm pretty sure that Im botching something up or looking at the wrong specification because I can find the 20MSPS but I can't find the five volts. Im assuming that 2Vp-p means 2 volts peak to peak. Im even apt to using a DSP.

Unfortunately, that's a pretty extreme specification and I don't think you're likely to find anything that will meet both of those at the same time. You'll probably have to relax one of those requirements to get a chip you can work with. If you really absolutely have to measure signals in the 10MHz range for some reason, then I think your best bet is to run your signal through a high bandwidth op-amp first with a gain of .5 or some such so you can look for ADCs with a smaller input range. I'm pretty certain that even switching to a DSP won't help, as the faster DSPs tend to need 3.33V supplies or less.

[Adam Y.]

Originally Posted by Kevin Sevcik

Unfortunately, that's a pretty extreme specification and I don't think you're likely to find anything that will meet both of those at the same time. You'll probably have to relax one of those requirements to get a chip you can work with. If you really absolutely have to measure signals in the 10MHz range for some reason, then I think your best bet is to run your signal through a high bandwidth op-amp first with a gain of .5 or some such so you can look for ADCs with a smaller input range. I'm pretty certain that even switching to a DSP won't help, as the faster DSPs tend to need 3.33V supplies or less.

Never mind. I just talked to my boss and he basically told me that anything about 10MHz, 2Vp-p, and 16 bits. Fortunatly, this actually means I can use a DSP.

[EricVanWyk]

Originally Posted by Kevin Sevcik

Unfortunately, that's a pretty extreme specification and I don't think you're likely to find anything that will meet both of those at the same time. You'll probably have to relax one of those requirements to get a chip you can work with. If you really absolutely have to measure signals in the 10MHz range for some reason, then I think your best bet is to run your signal through a high bandwidth op-amp first with a gain of .5 or some such so you can look for ADCs with a smaller input range. I'm pretty certain that even switching to a DSP won't help, as the faster DSPs tend to need 3.33V supplies or less.

I have to respectfully disagree with you here Kevin. I have found the following 20MSPS 5V parts in 15 minutes of searching.

AD9280 8bit 32MSPS 2.7V-5.5V $2.15
AD9200 10bit 20MSPS 2.7V-5.5V $2.50
AD876 10bit 20MSPS 4.5V-5.25V $3.31
Analog

TI
has a few more.

As does Linear
Bits Speed ADCs S/H SINAD Power Supply Voltage Range Packages
LTC1406 8 20 1 1 48.5 150 5V SSOP-24
LTC1741 12 65 1 1 72 1,275 5V TSSOP-48
LTC1742 14 65 1 1 76.2 1,275 5V TSSOP-48
LTC1743 12 50 1 1 72.2 1,000 5V TSSOP-48
LTC1744 14 50 1 1 76 1,200 5V TSSOP-48
LTC1745 12 25 1 1 72.5 380 5V TSSOP-48
LTC1746 14 25 1 1 77.5 390 5V TSSOP-48
LTC1747 12 80 1 1 72 1,400 5V TSSOP-48
LTC1748 14 80 1 1 76 1,400 5V TSSOP-48
LTC1749 12 80 1 1 71.7 1,450 5V TSSOP-48
LTC1750 14 80 1 1 75.2 1,450 5V TSSOP-48

[Manoel]

Eric,

It's not the supply range that's critical (as you pointed out, there are tons of 5 V parts), it's the analog input. With those specifications, it is really hard to find a suitable A/D.

[Kevin Sevcik]

Originally Posted by EricVanWyk

I have to respectfully disagree with you here Kevin. I have found the following 20MSPS 5V parts in 15 minutes of searching.

Originally Posted by Manoel

Erik,

It's not the supply range that's critical (as you pointed out, there are tons of 5 V parts), it's the analog input. With those specifications, it is really hard to find a suitable A/D.

Ditto Manoel. He was originally implying he wanted 5V p-p input range and said himself he could only find 2V p-p devices. No one makes them since few people want, need or try to directly sample RF signals without any buffer or filter circuitry. Thus all the devices out there expect a 2-3V p-p signal so the device designers don't have to trouble themselves compensating for voltage levels near the rails, making a difficult task even more so.

[Adam Y.]

Originally Posted by Kevin Sevcik

Ditto Manoel. He was originally implying he wanted 5V p-p input range and said himself he could only find 2V p-p devices. No one makes them since few people want, need or try to directly sample RF signals without any buffer or filter circuitry. Thus all the devices out there expect a 2-3V p-p signal so the device designers don't have to trouble themselves compensating for voltage levels near the rails, making a difficult task even more so.

You are right. Of course in my infinite wisdom I end up picking a sigma delta converter. Anyone ever use these before or should I go back to the drawing board?
Edit:
The signal is the percentage of ones compared to zeros in sample.

[steveg]

Adam, what resolution do you need?