04-27-2012 02:31 PM
Is it possible to extract a regular, high-frequency clock from the NI 4462?
I need to synchronise with another card (a PulseBlaster) which requires a 100 MHz clock: either card can be the main source, but they must stay perfectly in sync.
(Both cards have onboard 100 MHz crystal oscillator modules; the PulseBlaster's one is socketed, so would be easy to connect an external source intead)
I'm using the PCI version, so the 10 MHz PXI star clock isn't available. The 10 MHz or 20 MHz clocks that DAQmx usually exports don't seem to be available for routing on this device.
I can't use the Sampleclock, because that is variable (~ 25.6 MHz, changing with sample-freq) and because it is only active when the 4462 is actually making measurements.
Also, this clock needs to be either 100 MHz, or a simple integer fraction of it (eg 10 MHz, so that a PLL can re-generate the 100 MHz).
At the moment, the best option looks to be to solder onto the 4462's crystal oscillator output pin - but that's a really ugly way to do it....has anyone had any experience of making the relevant modifications?
Is there a better place to connect to? I'd prefer to use the RTSI bus connector, but despite the tantalising hints in the documentation, it doesn't seem possible to get the 10/20 MHz master timebase out: there is
a pin called RTSI7/RTSI_OSC, but this is permanently low.
Thanks for your help.
Solved! Go to Solution.
05-01-2012 06:11 AM
It sounds like you want to share a reference clock that you can use for synchronisation. Unfortunately, I think as you realise this is only possible with PXI.
You could export the sample clock but you have already mentioned the downsides of this.
Also, you can't route the reference timebase. It seems that as you said you need to modify your board, I am not sure who would have tried this as most would go with the PXI option.
Let me know if I can help you anymore - it seems like I have only confirmed what you already knew!
05-08-2012 01:51 PM
Thanks very much for your help. As you say, this confirms what I already knew.
But we really needed to keep everything in sync. So, here's what I did, which should hopefully help anyone else in the same position.
The circuit is very simple, using a FIN1001 LVDS line-driver. This is an SOT-23 package. Connections are as follows:
* Din connects to the 100 MHz oscillator output (TC0-2104, pin 4).
* Gnd connects to ground (C210, lower)
* +3v3 connects to the most accessible point, which is TCO-2104, pin 6. [The 3V3 regulator is on the reverse of the board]
* The FIN1001 power-supply is decoupled with an adjacent 0.1uf capacitor.
* Dout+ and Dout- are connected to a short (30cm length) of unshielded twisted pair, made from 0.7mm PTFE wire. (Z ~ 110 ohm)
Photograph is attached. The SOT-23 package is physically secured with a very tiny drop of cyanoacrylate glue, onto C210, and mounted on its back.
The receiver uses a FIN1002 receiver, terminates the twisted pair with 100 ohms, and uses a 74VHC04 to convert 3.3V to 5V TTL.
It's assembled on a 14-pin DIL socket, so as to plug directly into the socket on the other PCI card which normally has its own oscillator module.
Result: everything works, and I have a 100 MHz clock available to drive the other timing card.