TS3A27518E: Skew rate

Danilo,

Are you referring to the output skew? Comparing the input to the output at 50%? We don't have this information on this device. 

You can attempt to simulate this as models are available on the product page. You can also model it using passive components as i've shown below. I will note that the skew will be especially dependent on the load. The package type will also make a difference. I've added no load here which isn't an entirely realistic scenario. The device itself will have low On-Resistance, low On-Capacitance and the propagation delay on these muxes is extremely low as they're effectively just a FET that's turned on. So I wouldn't expect the Skew to be that high. The model shows 0.1ns but I'm willing to bet this is more of a limitation of the model precision than the device.
What level of skew/precision is the customer looking for? We have higher Bandwidth 2:1 devices that have the skew measured since they're more applicable for the use case. For example the TMUX1575 is a 4 channel that has the skew spec'ed at 6ps and the TMUX1574 has similar results. Both of these are good selections for high speed applications.

Thanks,
Rami

Thank you for the thorough response. We are interested in channel-to-channel skew (i.e. skew between the outputs). We are currently using two of the TS3A27518E. Each TS3A27518E is switching a QSPI interface to a primary or backup flash memory device. Per vendor requirements, the skew on each signal group needs to be within 50 ps. When considering just the trace skew, the signals vary between 29 and 48 ps of skew, however we do have some level of flexibility with the routing. We could certainly work with 6 ps of skew, but not 0.1 ns. If the TMUX1574 and TMUX1575 are considered higher bandwidth, then I would imagine, without the need of a simulation, that the TS3A27518E skew would be north of those numbers and problematic for us. Do you concur, or do you believe low single digit picosecond skews are achievable with the TS3A27518E? What's nice about the TS3A27518E is that we can switch 6 inputs at a time which is a perfect marriage to the 6 signal QSPI interface. Using either the TMUX1574 or TMUX1575 may be an issue since we would need to split each interface into two devices.  I did notice that the TMUX1574 and TMUX1575 appear to be significantly smaller in package dimensions (5 mm*mm vs 16 mm*mm) which may mitigate the impact of this since the devices can be closer together.

Another characteristic we're interested in is propagation delay. I believe that may be what you're referencing above. There is a MINIMUM trace length (and hence propagation delay) that we need to meet, and understanding the delay of the device would be beneficial as it can minimize the amount of copper we need to route.

Thank you!

Shane

Hey Shane,

I would agree with you that the 27518E can be expected to have higher skew than the TMUX1574 and TMUX1574. To what degree though, is where my uncertainty would lie. However, with the TS3A27518E the R-On matching (how much the R-ON varies between channels) is very low and the R-On is very flat. These are really just FETs and from a timing perspective, the Ron and Con would have the biggest impact and seeing how well the device matches the channels, I would feel confident that the skew between the channels wouldn't be that great. Comparable to other devices i would expect it to peak into the 50ps range but once again the caveat here is there isn't data available on this.

If you're looking to keep the signals all on one switch, minimize the propagation delay, and channel skew take a look at the TS5MP645. It's a 10channel 2:1 MIPI switch. Provides powered-off protection as well. Same 1ku price and a smaller package too. This may be a good fit, unless you have some reservations in which, i'll be happy to sort through any thoughts with you!



Thanks,
Rami

Does an IBIS model for TS5MP645 exist? Unfortunately our simulations with TS3A27518E show only a maximum operating frequency of 56 MHz is achievable. Also the rise/fall times of the mux exceed the maximum of the flash device we're using.

Shane,

What's the frequency bandwidth that your project requires?
We only have a s-parameter models available. But you could also use passive components as I showed above using the R-ON of the TS5MP635 (typ 2.65 max 6.5) and the C-ON (1.5pF). The rise and fall time will be dependent on the load seen from the driver so I expect it to get better with the TS5MP635 since your load will be smaller, depending on what the rest of the load looks like in your system here.

Thanks,
Rami Mooti

The hope is to operate at 100 MHz through the device.

We decided to go back to the original TS3A27518E and found that the datasheet and IBIS models do not match.

From the datasheet, the typical load capacitance is 21.5 pf but according to the IBIS model, it is quite smaller for the RTW package:

[Component]            TS3A27518_RTW

[Manufacturer]         Texas Instruments, Inc.

[Package]

|                                           typ         min         max

R_pkg                                       3.700e-02   2.500e-02   5.300e-02

L_pkg                                       8.060e-10   5.310e-10   1.139e-09

C_pkg                                       1.880e-14   5.700e-15   3.100e-14

|

|******************************************************************************

|

[Pin]    signal_name   model_name           R_pin       L_pin       C_pin

|

1           COM1       TS3A27518B_COM       4.700e-02   1.028e-09   3.100e-14

C_pkg_max is 3.1e-14

C_COM1_max is 3.1e-14

That is 6.2 e-2 pf, which is 0.062pf. 

 
If the true load is 0.062 pf then I believe we can use this mux in our application. Thank you

Hey Shane,

What you've pointed at are the capacitances for the package and pins/pads themselves. This is additional capacitance relative to the device die itself. 0.062pF will not be your true load.
Since these aren't actually buffered devices, which is what IBIS is typically used for, the work around is to to use IV curves to represent the passthrough on the signal pins though which is what the IBIS model does.
Another question from something you previously mentioned. You spoke on the rise/fall time of the mux. Could you elaborate more on this? The mux doesn't drive signals so i'm not sure what you mean by this. Unless you're meaning the turn on/off time perhaps? If you mean the mux is impact your rise/fall time too much, then this would be a load problem either way and if the rest of your system is set in place and you can't reduce load by altering something down stream, I would recommend finding a device that would have a smaller load (i.e. TS5MP645) as you couldn't avoid that problem in that particular system with that particular setup with the TS3A27518E. 

Thanks,
Rami