• State
  • Locked Locked
  • Replies 3 replies
  • Subscribers 18 subscribers
  • Views 730 views
  • Users 0 members are here
Support feedback
Options
Options
Related

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.

Comparing Bus-Switch Families for 5V to 3.3V Level Translation (PCI etc)

AP
Intellectual 390 points

Just a brief summary regarding current TI bus-switch solutions for level translation (legacy PCI, etc.).

Application:
5V to 3.3V level translation.
For either direction (auto-detection).
No need for I/O Off feature (both sides are powered-up at the same time).
Bus Switch is always on.
Bus Switch should pass digital low levels.
Bus Switch should pass digital high levels up to 3.3V.
Bus Switch should pass digital high levels in the range from 3.4V to 5.5V as 3.3V high level.
Ideally Bus Switch should not pass undershoots (e.g. -1.5V undershoots should be passed as -0.3V ... 0V).

We did some measurements with CBT-A, CBT-S and CBT-C.
VCC was 4.25V.
For CBT-A and CBT-C (without internal Schottky diodes) the VSS pin level was raised to approx 0.4V above system ground (to help cutting undershoots properly).

In short:

Best performer was CBT-S (with internal Schottky diodes).
Second best performer was CBT-A (with 0.4V on VSS).

The CBT-C did not really qualify for us.
The CBT-C showed a strange behaviour.
Our test environment was a PCI card (with the bus-switches on it) in a PCI system.
When the system board drove towards the PCI card, everything seemed to be fine on the PCI card side of the bus-switch.
When the PCI card drove a 3.3V high level towards the system board,
the measurement on the system board showed a step to approx 4V
and then a really slow (R-C typical) discharging until the 3.3V (or 3.1V) level was finally reached (took approx 2us).
Right now, we don't have any explaination.

Summary:

We could live with the CBT-S and CBT-A performance.
We will not use the CBT-C for this application (we do not want to pass a 4V level step when we are driving a 3.3V step).
Hopefully the older CBT-S and CBT-A families wills stay for some time (especially in the 0.4 pitch TVSOP package), because for our application CBT-C does not seem to match.

Comments are welcome.

  • 0
    TI__Mastermind 46480 points

    The C-rev devices set the I/O's to high impedace when the input goes below gnd. 

    The other 2 type parts use a clamp diode to control the negative signal.

    The purpose for the C-rev is that when you take an input or output  1Vt  below gnd the switch will turn on.

    The Crev will keep the I/O's High impedance when the signal goes negative.

  • 0 in reply to Chris Cockrill
    TI__Mastermind 46480 points

    Here is another part to try.  CB3T is a 5V to 3.3V translator. with Vcc at 3.3V the inputs up to 5V will be translated down to Vcc.

    The input must be 1V above Vcc for the part to function correctly. When going back the 3.3V to 5V direction your output will need to be pulled up to 5V.

     

  • 0 in reply to Chris Cockrill
    Intellectual 390 points

    Thanks for the input.

    It seems we will go with the older (plain feature-less) parts CBT-A or CBTS.

    As said, CBT-C showed some unexpected (and unwanted) behaviour compared to CBT-A and CBTS (in the same setup). Also, in our application, the switch is always active on (never off), so no need for the CBT-C features.

    We already had tested the CB3T, but for us it failed too. Right the opposite problem as the CBT-C. The CBT-C showed an unexpected high-level of up to 4V when driven with a 3.3V step (seems to depend on the attached load, leakage currents, routing, capaitance etc.). The CB3T was working fine for 5V downto 3.3V translation, but failed in a way for plain 3.3V level passing (to high RON, to much depending on the existing leakage current). The step to 3.3V was there, but then the signal level fell quickly to about 2.1V (to less margin for a safe high-level).

    So for us, the CBT-C would be no replacement for the older CBT-A.

    We are hoping that CBT-A and CBTS will stay active for some time (TVSOP package).