4 Replies Latest reply on Feb 15, 2015 7:25 PM by userc_41800

    Advice on New Design

      Hi, I'm designing a usb device around the USB FX3. I have a few questions I hope contributors might be able to help me with:


      1) In my application, data transfer speed is is very important. I require at minimum 200 MB/s. From what I've read, the FX3 is more than capable of this, but performance is very host-controller dependent. As I need to acquire a new PC for development, is there any chipset (or even better, laptop from e.g. Dell/Lenovo ) that somebody could recommend?


      2) I'm somewhat concerned about PHY errors and the abrupt termination of superspeed transfers (see the related thread). The error rate apparently seems to be heavily dependent on noise from the GPIF switching, cable length, and clock quality. With respect to noise from the GPIF, beyond the obvious preference of 1.8 V over 3.3, is there any specific PCB layout that could mitigate switching noise? I could do some electromagnetic simulations, but if anybody's got any tried and true methods, I 'd be happy to hear about them. Regarding clock quality, I have at my disposal an extremely low jitter LVDS clock that I can program to any frequency (a nice side effect of having ADCs in the design). I would imagine this is my best option (along with very clean clock power). The question then is frequency (19.2, 26, 38.4, or 52 MHz). As I don't know the FX3's internal structure, I can only guess that 52 MHz might possibly result in the lowest jitter in the device's PLL. Is this correct?


      Any advice would be very much appreciated!



        • 1. Re: Advice on New Design



          We have seen best performance numbers using Dell vostro 3460 laptop.


          Coming to your second question, we would recommend you to go with 19.2 MHz crystal.




          Sai Krishna.

          • 2. Re: Advice on New Design



            I experimented with 19.2MHz oscillator (ASDMB-19.200MHZ-LY) instead of  crystal on FX3 DVK. See attached picture. Kit behavies with oscillator much better than with its default crystal. These are results of my noise test:


            1) Oscillator, 3.3V GPIF , 1.5m USB cable, Renesas add-on adapter


            01:39:36 Read/Write=200.8/106.7MB/s Errors PHY/LNK=0/0


            2a) Crystal, 3.3V GPIF, same cable/adapter


            00:00:01 Read/Write=194.2/104.5MB/s Errors PHY/LNK=14/0
            Read from device failed (GetOverlappedResult error code=31)
            A device attached to the system is not functioning.


            2b) Crystal, 1.8V GPIF, same cable/adapter 


            00:01:47 Read/Write=201.9/103.3MB/s Errors PHY/LNK=96/1





            • 3. Re: Advice on New Design

              Update to my previous post. It was actually 2m (and not 1.5m) USB cable I experimented with.

              • 4. Re: Advice on New Design

                It's almost two years on, but thanks very much for your replies -- they were quite helpful!