GTL2003 Problems with totem-pole drivers

[Electr0nicus]

Hi everyone!

I have a problem using a GTL2003 bidirectional Level shifter. http://www.nxp.com/documents/data_sheet/GTL2003.pdf

The background:

I want to level shift a parallel display data bus from 3.3V to 5V. The fequencies are pretty high, (pixel- clock has approx. 26MHz). I'm using ac- termination on the end of the databus. This just works fine, the signal is really clean compared to the look of the signal when unterminated (@ no load). All signals are measured with a ground sping on the oscilloscope probe instead of the usual ground clip.
Here is the signal without termination. It is really horrible as you can see. This is because of the long ribbon cable (approx 20cm).


After adding a parallel AC- termination using a 470pF cap and a 47ohm resistor, the signal looks really clean.


Those terminated signals are fed into the GTL2003 device.  The device is practically a array of 8 pass FETs and one reference transistor. According to the datasheet, for shifting the signals up to 5V there will be a pullup transistor required on the output of the GTL2003. The problem I now have is that the signal isn't shifted up to 5V at all. It pretty much stays absolutely the same.
Whatever pullup resistor I use on the output (tried 10k 2.2k 1k 470R 100R) the voltage level of the signal stays the same.



Also the voltagedrop= current across the pullup- resistor (1k) is very interresting



There is 2mA DC- current flowing into the output of the GTL2003 the whole time, generating a voltage drop of 2V across the resistor, of course overlayed by the signal's waveform. According to the datasheet http://www.nxp.com/documents/data_sheet/GTL2003.pdf on page 4 chapter "Function selection" under "Function selection, LOW-to-HIGH translation" when there is a high level on the input then the pass transistor is nearly off. But there is nowere definded what "nearly off" means. So I would think this would explain the static DC- current flowing from the 5V output side trough the pass transistor down into the 3.3V logic driver (totem pole) and up into the 3.3V rail.

So my question is simply: Is there a connection scheme of those GTL devices, which can be used with totem pole output drivers on the lower side. Because as the situation is now, I could lower the pullup resistance even more, until the pass transistor or the totem pole output driver on the lower side gives way. But as long as the resistance of the pass transistor in "nearly off" state is low enough and the low side driver is strong enough to consume the current fed back from the 5V side to its input, the output will never be nearly 5V. 

Thanks Gregor