NVT4858 - Signals on Host Side not being pulled up to full VCCA ?

[benoitlemieux]

Hi NXP and community,

I am using an NVT4858 level-shifter for SD3.0-compliant circuit.  Everything is working fine but analyzing the waveforms I can see that the signals on the host-side data lines are not being pulled up completely to VCCA. 

I'm seeing the data line being idle at full VCCA level (1.8V in my case) but during communication with the SD card it's dropping to 1.0V.

Analyzing the block diagram of the IC, I see that the one-shot circuit driving the output stage should have no effect on the output, other than speeding up the edge.  Is there a reason why my data lines are not being pulled up to full VCCA?

Thanks for the help.

[diazmarin09]

Hello,

I hope all is great with you. Thank you for using the NXP communities.

Could you please share your schematic? Have you considered the internal resistor of the SD card?

Regards,

David

[benoitlemieux]

Hi David,

Here is a simplified block diagram of my schematic:

The signals on the SD card side are fine, all at 3.3V (using an SD 2.0 compliant card).  The problem is on the processor side.  The CLK and CMD lines are all properly being pulled up to 1.8V.  The data lines are idling at 1.8V, but when transactions are occurring they are only coming up to 1.0V instead of 1.8V.  See example waveform below:

[diazmarin09]

Hello Benoit,

I am pleased to contact you again.

It is not a normal behavior. The signal should be up at VCCA.

When the A side is switched from low to high, the one-shot will attempt to pull the signal to VCCA via its Ron which is less than 70 ohm. Once the one shot is timeout and turns itself off, the internal pull up will keep the signal at VCCA.

Is there something pulling down the signal on that side? I mean between host and device?

What is the data rate? Make sure the data rate is slower than the one-shot on time.

[benoitlemieux]

Hi,

Thanks for confirming this is anormal behaviour. 

When you're talking about Ron, you mean the 70k pullup resistor on VCCA?  

There is no passive component pulling the signal low, but I will try to add a stronger pull-up resistor on the data input pins to see if this helps behaviour.  Maybe the problem is something that is invisible to me, in my host processor.

As for the data-rate, my clock frequency is around 16.666MHz.  I don't see a documented ON-time for the one-shot circuit on the datasheet, but this is an SD2.0 datarate for now so I'm fairly certain that the time is being respected, given that the NVT4858 should work at SD3.0 data rates.

Thanks for the help, I will see if adding a strong external pull-up helps.