SN3257-Q1: Maximum value of Inter-channel skew and device-to-device skew for SOT-23 (DYY) and TSSOP (PW)

Thanks for the reply. There are some numbers that is a bit confusing. Let me try to double check. So from the datasheet, I can see for DYY: inter-channel skew 10ps, propagation delay 78ps, for PW: 18ps and 95ps accordingly. All NOM. 1. seems PW has wider spread and higher values than DYY, which conflicts with what you said. 2. the propagation delay numbers are also much higher. From the datasheet, I only find the Ron 2(NOM)-5(MAX), Con 8pF (NOM), but cannot see the Con MAX. do you have another datasheet than what I have (revised in October 2022)?. 3. It is fine that you have only small sample units, some general rough estimation is good enough for us. You mentioned +/-3pF, is that the skew between different units? So if we say the expected skew between channels within and among different channels, that's going to be 10ps (Tsk)+3ps = 13ps? 4. I thought the propogation delay between difference channels are the contributor of skew. Am I correct?

In addition, I've downloaded the IBIS model of this switch and simulated the insertion loss, the 3dB bandwidth is up to 880MHz not 2GHz as stated in datasheet. Could you help to double confirm that? Thanks. 

Hello,

So I fully follow what you're saying on the skew on the PW vs DYY but I'm simply looking at char data and that's what it shows for RON and CON of the device and using that to try and draw an estimation since we don't have skew specific data. From the data I would expect these to be backwards honestly. 
All that being said, sometimes our math and approximations isn't as accurate as real life test data and I did some more digging and was actually able to find skew data that was directly measured! 
So from our characterization data which was only 3 device and all channels measured, I'm seeing the DYY having worst case a delta of device to device skew being about 15ps and inter-channel skew the devices vary by a max of 4ps.

For PW unfortunately we only have the 1 char device saved in this internal file I found. The inter-channel on this one is also 4ps. The PW package would in theory have a slightly larger delay since it would be larger and have more conductor space though. I would anticipate the device to device skew being very close to the DYY though still since the packaging don't make a substantially large difference. 

How tight is your window here for acceptable device to device skew?

I typically don't see IBIS used for higher frequency applications so it wouldn't surprise me if there are some BW limitations in the model. Can you share your schematic? I'm trying to run it as you did with the s-parameter model but it fails. I don't have much experiencing with running AC tests like BW in IBIS (usually I do DC and some timing tests) so I may be setting up the S-parameter simulation incorrectly. 

Thanks,
Rami

Hi Rami,

Thanks for the answer. So then the skew is roughly 15ps, +/-7.5ps. That's ok then. Our spec is +/-25ps. There are other things in our channel also contribute to the skew of course. 

I'm using Ansys tool, and we need to follow a special way to simulate the Switch IBIS behavior as attached. Then run Linear Frequency Sweep

Glad to hear that the spec is within acceptable value!

On the IBIS bandwidth, I'm running into some problems simulating it as it looks like the license I have for ADS only allows transient simulations which is a bit of a problem for something like this which requires an AC sweep. I'm still looking into how i can confirm the model having that limitation. I will not on your setup, it looks like you're missing a 50ohm output impedance that I'll usually like to add in simulations to represent the VNA's impedance. But I don't think that would account for the differences in what you're seeing. These models aren't always accurate in every spec. Typically for IBIS we focus on DC. 

Is PSPICE an acceptable tool for simulation for you? The TMUX1574 is the non-automotive version of this device and does have a pspice model available on ti.com. Parametrically it's the same, the device just doesn't have the automotive qualifications.

Thanks,

Rami 

That's a bit strange that ADS only allows transient sim. In my setup, the ports added are already with 50ohm, so no external resistors needed. 

If you provide pspice model of an identical part, that's totally fine for us. We can use that too.

Hey Fish,

Yea it's a bit frustrating but I believe it's just a limitation on the license I have active. Trying to sort through it now but certainly I would expect some AC capabilities given that I have AC simulation options

I've attached the PSPICE model here for 1574. The lib file and OLB file will be in there.
The model itself has a disclaimer that says the BW is higher than the datasheet value and only the 5V supply case was modeled here. That being said, I ran a test myself and didn't quite see the same results but could be how the tests were done when this device was released (Anything we've released in the last 2-3 years has a different PSPICE flow that will typically cover all the supply cases in the datasheet and will typically have a little more accuracy as we get better and better at creating these models).

My setup was fairly straightforward. I know some people like to do just use 20*log10(V(OUT)/V(IN)) for BW which here gives about 1.7GHz.
In theory in simulation in a setup like this where you have 50ohm input with 50ohm load, you really should be doing 20*log10(V(OUT)/100m), with the 100m just coming from the 200m input and the expected 100m you would see at that voltage divider if the mux wasn't there. Using that I actually saw closer to 800MHz 3dB in this model. So not quite to where the datasheet shows. Are you working with a frequency where the BW limitations will be an issue?

TMUX1574.zip

Thanks,
Rami

Hi Rami,

Our application is only up to 200Mbps, but our signal is with fast edge, like 200ps or smaller. We want to simulate the behavior and see how the signal would look like and be impacted with the switch. 

Hey FIsh,

Ah I see. if it's square waves with those sharp edges I typically recommend having about 5-7x the operating frequency as the bandwidth so you're probably looking for about 1.4GHz atleast since you're really wanting to preserve those sharp edges. 

The model doesn't quite maintain as much.

That being said, for higher frequencies we have a s-parameter file as well that may be useful? 

scdm204 (1).zip

-Rami

Hi Rami,

Thanks for the answer. Yes, the s-parameter model seems a good choice for our simulation!