Connecting a high-speed comparator to a PC

[llamoure]

Dear All,

I need to store the output of a stand alone comparator (Hittite HMC875LC3C) in computer RAM drive. Currently, I'm using an 8bit 2GHz ADC (an Agilent PCIe - Acquiris U1071A) to bridge between the comparator and the RAM drive. This solution is redundant, slow, and very expensive (the Agilent ADC is almost $10k).

I would enjoy a better solution where no ADC is present and the comparator connects directly to the PC. Perhaps through a USB port?

Ideally, I would like the comparator to operate at around 1GHz (it supports operations up to 20 Gbps). I would suspect the maximal throughput essentially depends on the computer's processor and interface i/o characteristics.

The comparator outputs pulses are about 60ps long and the Output Voltage High Level is -10mV while the Output Voltage Low Level is about -420 mV. The comparator is emmbeded on a PCB with SMA connectors.

Question: Is it possible to connect a high speed (1GHz) comparator into a PC without having to use an ADC? If so would some form of synchronization be required?

Thanks

Louis

 

[Skogsgurra]

USB will not work. Orders of magnitude too slow.

It depends a lot.

Are you interested in the relation between signals and not necessary to deal with each transition immediately?

Or is it necessary that the PC reacts immediately when a signal changes from lo/hi or hi/lo?

In the latter case, I see no way except a buss connected board. Not even the old parallel (Centronics) port would be able to react fast enough.

Then, will the cycle time in the PC CPU be fast enough to handle your signals? CPU clock is one thing - CPU cycle time is something very different.

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[IRstuff]

USB 3.0 might be potentially usable, since it has a maximum raw datarate of 4 Gbps. But, even at that, the serial interface requires a protocol that limits your ability to use the interface in that fashion. Your short pulsewidth and voltages cannot be directly interfaced to USB anyway.

However, that really isn't the bottleneck; cnosider the fact that the fastest PC only has barely 1.5 GB/s memory bandwidth.

What exactly do you want your PC to do with the data? As Gunnar indicates, any real-time processing of this data with a PC may be unrealistic.



TTFN

FAQ731-376
Chinese prisoner wins Nobel Peace Prize

 

[llamoure]

Thanks for your replies. No real-time processing.

In the end, I would like to send these bits to remote clients (who request them) via the PCs network card. The RAM is used to store the bits while they await a request. If client demand is high, RAM storing is not really necessary (I think). If there is no client demand, the RAM is full and new incoming bits from the comparator are simply discarded.

In my setup, I can operate the comparator up to 1.5GHz. This upper bound is reached when the outside client requests total that bandwidth. It boils down to finding the most efficient method to transfer the bits of the comparator to the network card.

 

[IRstuff]

Well, something that can handle 1.5 Gbps datarate isn't going to be cheap. An A/D is obviously gross overkill, but the sample rate is high enough requires commensurate I/O hardware.

"new incoming bits from the comparator are simply discarded"
why would you not keep the most recent data and dump the oldest?

TTFN

FAQ731-376
Chinese prisoner wins Nobel Peace Prize

 

[IRstuff]

You might try to look for some sort of GPIO board, or a Virtex 4+ FPGA board with direct input into the FPGA

TTFN

FAQ731-376
Chinese prisoner wins Nobel Peace Prize

 

[llamoure]

Thanks IRstuff. I will definitely consider the FPGA approach.

Louis