Hi,
I am using TMUXHS4412 for a MIPI design, I found in the S parameter there are 8ps difference between P/N ports.
And in the datasheet the typical propagation delay is 4ps. So is 8ps the worst case? Is it always p port longer than N port?
Xiaole:
I did see difference between A0P/A0N to B0P/B0N with our new data
But looks ok between A1P/B1N to B1P/B1N.
So it's not always P is longer than N. Do you need new S-paramter data for your simulation?
Regards
Brian
Hi Brain,
One more question for the TDR, why this switch's TDR is so weird?
We are using this switch IC for different projects now to switch MIPI signals.
Hi Xiaole,
Our systems engineer will return to the office next week (1/8). We can discuss this behavior with him at that time.
In the meantime, I have a few questions to get a better idea of your simulation.
1. The propagation delay in the 4412 datasheet is typically 50ps at 1GHz. You mentioned that with 3 cascading MUXes you are seeing 30ps of propagation delay correct?
2. The skew between channels is typically 4ps at 1GHz, yet you see 18ps of skew with 3 cascading 4412s correct?
3. What level of skew do you need? If 18ps is too high, how much would be reasonable for your application?
4. Are you using the 4412 s-parameter file that Brian sent for your simulation?
Best,
Shane
Hi Xiaole,
After reviewing with systems, the skew in the datasheet (4ps) is an average of all 4 channels on multiple devices, while the S-parameter skew (6ps) is modeled after one or two channels on one device in particular. This could explain the 2ps discrepancy between the datasheet and the S-parameter.
We suspect the difference in length between pins on the package may cause the P path on channel 1 to be slightly more delayed than the N path. This would not be consistent across all channels, as the length between input/output pins changes. Here is the MUX with input/output pins circled to help visualize the length change.
When you consider that all pins connect to an IC die within the package, the length from D0P to DA0P may be longer than the length from D0N to DA0N. As you go down, the opposite may be true. For example, DB3N may be slightly longer than DB3P because DB3N extends further from the IC die within the package.
Best,
Shane
Hi Xiaole,
I understand the concern. If you need lower skew than what the TMUXHS4412 can provide, consider using the TS3DV642.
The TS3DV642 is a 12-channel 2:1 MUX with inter pair skew of 2ps.
Would you be interested in this device?
Could you elaborate on why you need to cascade three MUXes? If you need more channels the TS3DV642 could be a better solution.
Best,
Shane
Hi Xiaole,
I drew a simplified diagram of what I believe you are going for. Each TX/RX would have two lanes in this depiction. Does this look correct?
Essentially this would be a 6:1 MUX for MIPI signals over 12 channels (9TX + 3RX) to 2 channels.
I am looking into alternatives and want to be sure I have the right idea.
Best,
Shane
