• State
  • Locked Locked
  • Replies 5 replies
  • Subscribers 17 subscribers
  • Views 336 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

TMUX646: TMUX646 vs TS5MP645 for lowest power loss between 1 - 250MHz.

TonyMotegi
Prodigy 130 points
Part Number: TMUX646
Other Parts Discussed in Thread: TS5MP645,

We are using multiple TMUX646 devices in a cascaded format to create a 4:1 multiplex of a 100 ohm diff pair. Unfortunately we are seeing a lot of losses in our system, which ramp up significantly above 100MHz. I note that this is the latest greatest of the TI high speed multiplexers, but I also notice it has a significantly higher Rds on than its previous generation part, the TS5MP645 ( 6 Ohm vs 2.45 Ohm).

Our implementation is actually attempting to drive an RJ45 cable and measure cable performance parameters such as crosstalk, insertion loss, return loss. It requires 3 MUXes cascaded. We require very low crosstalk in the MUX, such that we can measure crosstalk of the cable without nearing the noisefloor of the MUXes. The drive signal is sent through the cable under test and then looped back onto an adjacent pair, so the signal propagates through 6 MUXes before being terminated. The muxes are such that we can drive each cable pair one at a time. They are wired in such a way that if they are not being driven, they are being terminated at 100Ohm (the 110 Ohm on the schematic is in parallel with 1K on the signal measurement device).

We chose the TMUX646 over the TS5MP645 due to its better crosstalk performance. However, we are having problems above 100MHz where the losses are larger and affecting our measurements. My question is, do you think we would get reduced loss up to 250MHz if we chose the TS5MP645 ? Due to the nature of the package it's not easy to swap the devices by hand, but our PCB assembly house is confident they can swap them for us if you think there may be a reduction in power loss with the TS5MP645 ?

I would have thought that the compound loss of the 6 Ohm RDSon x6 (3 drops on the way out and 3 drops on the way back in) would have a serious effect, but the below image found in the 

TMUX646 S-Parameter Model

suggests there is negligible difference in gain below 1GHz. 

I attach an extract from our schematic that hopefully shows our implementation. Please let me know if you have any questions. 

  • 0
    TI__Mastermind 28578 points

    Hey Tony,

    What kind of losses are you seeing and how bad are the losses? 

    If you're looking at 100MHz, you'd be more interested in the bandwidth up to 500MHz since you typically want to incorporate around 5 harmonics for square waves. So you could be filtering out some of your higher harmonics so your square wave may be looking more sine wave like and by the 6th mux you really could be reducing quite a bit there and you may be seeing a combination of rounding out from filtering out harmonics and some signal loss from the insertion loss compiling up. TS5MP645 may be able to help with the second one but the first one would be worst on the 645 given the lower frequency. 

    Just to dive a little deeper, TS5MP645 may be help reduce insertion loss at DC levels but you may really be seeing the upper limits of the harmonics. So at 100MHz, through 1 switch you may be seeing close to about -0.75dB and as you add those up it may not bee too bad.

    But if you're looking at say 250MHz you may be seeing closer to 1.2GHz for the 5th harmonic you may be seeing closer to -3dB.
    By your 6th mux, your fundamental frequency itself may even be close to -3dB. 
    I guess naturally my next question here would be to confirm that you're using square waves and just some clarity on what expectations vs reality.  

    Thanks,
    Rami

  • 0 in reply to Rami Mooti1
    Prodigy 130 points

    Hi Rami, 

    Thanks for your response. Our test signal waveform is purely sinusoidal, so we shouldn't have to worry about higher order frequency components of the fundamental. Hence we are very much interested in getting the best performance between 1 - 250MHz, both in terms of minimal loss and lowest crosstalk. We are required to detect cable crosstalk in excess of 62dB at 1MHz, which is why I went for the TMUX646 in the first place. But, maybe the fact our signal is cascaded through multiple MUXes means we could get away with the TS5MP645. Maybe the cumulative attenuation of adjacent signals (crosstalk) will help us. I will get provide some further detail on the losses we're seeing, possibly via one of my colleagues. 

    Thanks

    Tony

  • 0 in reply to TonyMotegi
    TI__Mastermind 28578 points

    Hey Tony,

    Purely Sinusoidal I would expect it to be fine at <100MHz but after it the loss could still accumulate pretty quickly. 

    I have recently done some tests on cascading multiplexers (See this app note : https://www.ti.com/lit/an/scda047/scda047.pdf?ts=1714148423066) on a lower BW device and saw the trend to be typically that the first mux does decay your bandwidth to about half. After the 1st mux added the affects of attenuation do become less extreme. The roll off isn't as sharp as some of our other devices though so 250MHz still could be fine but going through 6 channels isn't so straightforward to determine what the loss would be at that point. 
    I ran some simulations in that app note I shared and the -3dB point drops to about 10% of the original by the 6th mux if that helps get any sort of estimation.

    I think getting more details on the losses from your colleagues can help and if it's at all possible, seeing the losses per cascading stage (ie see the output after 1mux, 2 mux, 3 mux ,etc.) can help see where that line for too much may be. 

    Bottom line though, if it's already sinusoidal and you can still capture the frequency component on the output but the signal amplitude is the limiting factor, The TS5MP645 would have a better chance to minimize the loss.

    Thanks,
    Rami

  • 0 in reply to Rami Mooti1
    Prodigy 20 points

    Hi Rami

    Here is the S21 insertion loss plot for our PCB. The measured chain consists of a wideband balun (single ended to differential) 3x MUX. RF relay, RJ45 connector, identical balun (differential to single ended). Unfortunately it is not possible to break into the signal chain and measure individual devices.

    Hope this helps

    Regards

    Geoff

  • 0 in reply to Geoff S
    TI__Mastermind 28578 points

    Hey Geoff,

    It looks like you still have losses close to DC as well though. I wouldn't expect those losses at frequencies that low as well. What are your trace lengths? 
    I'm assuming there's a good chance that we won't be able to do this but is there any way to short the signal paths and do an S21 measurement without the mux there? I'm just looking to see if there's a way we can verify the board itself isn't the limiting factor

    -Rami