TS5A623157: switch for 75mhz with AC signal

Hi Deepa,

Thanks for the quick reply.

Based on the information that you have provided these are our best bets:

TMUX136 (2 channel 2:1 switch, 4.6 Ohms (typ at VDD = 5V ), 1.4pF (typ) ,  Can pass 0V - VDD 

TMUX646 (10 channel 2:1 switch, 6 Ohms (typ at VDD = 5V), 1.5pF (typ), can pass 0V - VDD

TS5MP645 (10 channel 2:1 switch, 2.45 Ohms (typ at VDD = 5V), 1.5pF (typ), can pass 0V - VDD

These are the devices that we have in our portfolio that meet the requirements - however as a note - the mux will not see the overshoot and undershoots differently than a normal signal - but as long as these fall in the operational range of the part they should be passed. One benefit about the higher bandwidth of these parts is that more of the high frequency content will be passed through the switch.

As another note - unused channels can be left floating or terminated to ground via a 50 Ohm resistor to mitigate reflections. At 75MHz I don't think that should be a huge issue so floating unused pins shouldn't be an issue - but best practice is grounding with a 50 Ohm.

If you have any other questions please let me know!

Best,

Parker Dodson

Hi Deepa,

I wouldn't use prop delay as a measure of slew rate for two reasons.

1. Prop delay is calculated and not measured for this device

2. How we set up the prop delay calculation we assume that prop delay is the time between when the input reach 50% and when the output reach 50%.

         A. The time to reach 50% charge on a cap is much shorter than it takes to get to ~100% so  how fast output voltage rises is hard to pin                    down from prop delay.

         B. Slew rate is how fast the output can rise which is similar to what we are measuring for prop delay but not 1:1. 

The mux - at its core is an RC low pass filter when the switch is closed. The output slew rate is going to be roughly equivalent to the following equation, which can be used if the circuit can be modeled as the following:

This assumes the switch is essentially an RC with the R being on resistance and C being the on capacitance. It also assumes that the loading condition of the mux can be assumed to be a RC combination in parallel (which is equivalent to many IC's inputs). The circuit can be simplified by creating 1 RC circuit out of the two seen. To do this we can find the Thevenin eq. by adding the caps in parallel and treating that as the "load".

To find the Thevenin equivalent resistance remove the capacitors and short voltage sources to find the eq. input resistance. In the simple case above it is the (R_ON * RL)/(RL + R_ON) and if R_ON << RL than Rth ~ R_ON. 

Now we more or less understand the output circuit and since it is an RC - the output slew rate is going to be limited by the RC charging constant - and about 5RC = full charge on the output cap. so this leads to the slew rate being the following:

SR = (V_in / (5 * R_th * C_th_L) where R_th is the Thevenin resistance and C_th_L is CL+ C_ON) 

With the data we have an absolute worst case is when R_ON is max value - which is listed in the datasheet, since capacitance changes less using the typical value is how you can set up the approximation. For the min value you can use the Ron plots to determine the min value for the supply and operating temp that is being used.

This is approximating two cases.

1. It assumes the output of the multiplexer is going to be roughly an RC load - but this isn't always the case. Using a SPICE based software with the RLC stand ins during simulation would be the next step if the system setup cannot be simplified as described above. Please be aware that most IC's that have a high impedance input will have a very large RL (or none listed) and some parasitic capacitance - in these systems this type of approximation is a decent approach but if there is more RC networks or inductance that can't be neglected using a simplified spice circuit of the setup can help with  the approximations.

2. It is assuming (for worst and best case) that the resistance will be the same across input voltage - it isn't. This means that the worst case SR shouldn't' be reachable as the on resistance will not constantly be max across voltage and real SR should be faster than this worst case approximation calculates for. The same idea, but in reverse, applies to the Best case measurements. For a more accurate simulation - using voltage controlled resistor with data pulled from the plots will be more accurate - and the more points you include from the plots the more accurate the approximation will be. For quick boundary calculations the first approach is fine but if you want more granularity having a variable resistor is necessary. 

So in short it depends on the loading conditions, and the prop delay is at the end of the day measuring something different so it isn't a great comparison point. However using the above process should yield a educated approximation on what to expect for your system.

Best,

Parker Dodson