Dear Sir,
Have you an explanation regarding huge propagation delay dispersion from 0.5 ns which is minimum up to 4.4 ns which is maximum? for SN74AVCH8T245 device.
Considering maximum propagation delay, this device doesn't meet 320 Mbps data rate, instead it can be around 55 Mbps.
Thank you,
Nicolae Zaharia.
Hi Nicolae,
The min/max prop delays are boundaries put in place by our systems team based on simulated and observed performance of the device. 0.5ns is most likely just put in to indicate that the minimum is not 0ns, and I wouldn't expect to see that in any real devices in production.
As for the relation between max prop delay and data rate -- propagation delay and data rate are somewhat interconnected, but you can't just directly calculate data rate from propagation delay.
As shown in this graphic, delay is how much the output shifts in time relative to the input, but you might get the exact same signal out that you put in even with large delays. This is highly dependent on the input transition rate, internal delay of the device, and the output transition rate.
What's most important in determining maximum data rate most of the time is output transition rate.
Faster transitions mean sharper edges and higher data rates possible, while slower transitions start to cut off your signal and you can lose data. The above waveform is showing how increasing load capacitance can impact transition rates and thus max operating frequency.
Hi Emrys,
Our concern is addressed to the maximum switching time (tPLH and tPHL).
For a 125 MHz CLK signal where T = 8ns, you have tPLH > 4 ns that is more than T/2.
What propagation delay we should consider from datasheet, when we choose a level shifter device for signal with T = 8 ns?
Because 2.5 ns could be reasonable but 4.4 ns is not.
Thank you,
Nicolae Zaharia.
Hello,
What you have highlighted as "t_PLH" is actually output rise time (transition time).
Please see the datasheet:
From what I know, there aren't any translators on the market that meet 100% of the very stringent timing requirements for RGMII. It seems to me that it was designed for only direct communication (ie no voltage translators being used).
I have seen customers use both AVC and AXC translators for RGMII and have it work, however we don't provide any guarantees. In our testing, we've found that these translators may not meet requirements for the 1.8V to 1.8V translation at 250 Mbps data rate. Our testing shows that they work well at 1.8V to 3.3V though.
The SN74AVCH8T245 can definitely be operated at 250 Mbps for 1.8V or greater supplies, however it may not always meet those strict timing specs I mentioned above. This will usually be system dependent, so you'd need to prototype and test.
I ran a quick IBIS simulation to show the expected performance of the SN74AVCH8T245 at 125 MHz, 5 pF load:
You could instead use the newer SN74AXCH8T245. If you don't need the bus-hold feature, I would recommend to instead use SN74AXC8T245. Bus-hold can sometimes cause problems - especially for systems that require any pull-up or pull-down resistors (default HIGH or default LOW states).
I just so happen to have the below eye diagram for the SN74AXC8T245 operating at 1.7V to 1.7V (which I would call the worst case for 1.8V to 1.8V):
This is typical data at 25C, but hopefully it shows you how the device is expected to perform.
I also ran the same simulation as above so you can directly compare AXC to AVC:
