TXB0108: Output signal at high side (VCCb) is garbled

Hi Benoit,

Do you have the scopeshot and the schematic of the system? Are all channels behaving this way? What is the interface? Any external pullups or cap loading?
The shark fin shape i imagine is because of the large external loading on the output of the TXB device.
The TXB is a weak driver with only 20uA limited drive strength and is intended for interfacing with light loads (high input impedance).
It could also be worsened by scope probes with high loading( passive probes).

I referred to one of the similiar queries which had this issue:
e2e.ti.com/.../2655544

Further to my message above, I should have noted that outputs pin 12 and 13 out of the TXB0108 are actually two clock signals that are sent to a TLV320AIC3204 CODEC (yours!) to pins 32 and 8 respectively:

TXB0108 pin 12 to TLV320 pin 32 (BCLK 2.048MHz)
TXB0108 pin 13 to TLV320 pin 8 (WCLK 8kHz)

On a previous circuit where I didn't require a level translator, all worked fine. But the signal source is a new device that outputs 1.8V therefore had to use your level translator.

So basically:
Signals OUT of source INTO pins 12 and 13 of TXB0108 are fine.
Signals OUT of TXB0108 INTO pins 32 and 8 on TLV320 are garbled as explained in my initial post.

Thanks again,

Benoit

Hi,

Thanks for the super fast reply. I'll see how I can get a snapshot and a schematic section when I get back to work tomorrow.

However, in the meantime, not sure if you saw my second message where I specify that the TXB0108 goes to Texas Instruments TLV320AIC3204 CODEC.

Given that both parts are from TI and that the problem is between both of these TI parts, can you maybe provide a quick indication of how I should make the connections?

My "other" part generating the signals is a SIM7600 modem by SIMCOM. The signal output is 2V max (from what I see on my scope).

Can the TLV320 take-in the 2V signals directly, bypassing the TXB?

Thanks,

Benoit

Hi Benoit,

It depends on the logic Vih and Vil levels of the TLV320. An expert on that device may be able to let you know more about this.
also, closer to the threshold voltage, the higher is the current consumption due to the shoot through current.

What is the drive strength of the SIM7600's output pins?

Are there any pull-up/-down resistors on these clock lines? (A direct connection from the TXB to the AIC, without any capacitances or resistances, should work fine.)

I will submit a new question to the TLV320 forum page and let you know the outcome here.

Thanks,

Benoit

I will send an email to the modem manufacturer (SIMCOM) and post the answers here.

Benoit

Thanks Benoit.
Let me know once you have more information

Hi again,

Here's an update hoping it helps... I've done some more tests and I think I can rule-out the CODEC because I cut the physical trace on the PCB going from the TXB to the CODEC to make sure that the TXB outputs are isolated (physically disconnected) and I still have the garbled signals.

I've attached images from my oscilloscope screen.

FYI, in case it helps, my oscilloscope is a Tektronix TDS220 and my probes are Circuit Test OP-100A with the following specs (in case you wonder if my probes have an effect on the signal when I monitor it):

Input Resistance: x1 1M Ohm / x10 10M Ohm

Input Capacitance: x1 56pF / x10 13pF

Compensation Range: 10-30pF

Cable Length: 120cm

Operating Temperature: 0 to 50°C

The pictures from my oscilloscope screen for both signals for you to see.

Unfortunately I am not authorized to share the schematics but what I can say is that these two connections in particular are very straightforward.

The other I/O signals on that same TXB0108 work just fine but note that they are simple I/O lines getting toggled once in a while as opposed to the "problematic" ones being clock signals.

On either side of the TXB0108 on the PCM lines, there aren't any capacitors or resistors;

So to re-cap:

Modem out perfect 2.048MHz 50% duty (top-right image) > TXB0108 in 1.8V > TXB0108 out 3.5V > Lower-amplitude 2.048MHz triangular waveform with DC offset (top-left image)

Modem out perfect 8kHz 0.4% duty (bottom right image) > TXB0108 in 1.8V > TXB0108 out 3.5V > 8kHz 0.4% duty shark-fin waveform (bottom left image)

On this same circuit, I have a second TXB0108 with unused lines. So after cutting the trace to the modem signals, I soldered a mini piece if wire from the modem's PCM BCLK signal to the unused 1.8V input on the secondary TXB. Its output is not tied to anything. When I put my probe on the output, I get the exact same result. So I've ruled-out the original TXB0108 IC because the secondary behaves exactly the same.

Could it just be the oscilloscope probe corrupting the signal when I touch the line but when I remove it, the signal is fine but I just don't know? Am I beating a dead horse and I don't know it?

Any idea or suggestions?

Thanks again,

Benoit

Are you using ×1 or ×10?

The 3.5 V waveforms look as if the load on the TXB outputs is too high.
And the 1.8 V waveforms are not exactly perfect, either.

How long are the traces? Are there any cables?

The probes are set to x1.

That's the thing - there's NO load on the TXB 3.5V output - it's just the pad sitting there on an unconnected copper pad.

The 1.8V waveforms don't look perfect because of the granularity of the LCD. This is a *very old* scope so the pixels have a staircase effect. When I make the waveform wider, the staircase effect dissipates over much more pixels.

For the trace lengths, if you're referring to the 3.5V side, it's 0 as indicated above - there's no trace. The pin is just sitting on its copper pad.

Thanks!

Benoit

×1 probes are a nontrivial load; always use ×10, if possible.

Not sure if you saw in my post with the pics, I put the specs for the probes (I guess they're passive? Never looked into that kind of stuff so I don't know what an active probe it quite frankly):

Circuit Test OP-100A

Input Resistance: x1 1M Ohm / x10 10M Ohm

Input Capacitance: x1 56pF / x10 13pF

Compensation Range: 10-30pF

Cable Length: 120cm

Operating Temperature: 0 to 50°C

So I cut the traces going from the modem to the original TXB 1.8V inputs and re-wired the trace from the modem using tiny wires to go from the original modem trace to the second TXB's 1.8V input. The wire is like gauge 30 and about 6mm long.

The outputs cannot have any excessive loading, *unless* it's caused by the probe itself because the 3.5V output is sitting there on its own copper pad which does not have a trace tied to it. These were unused I/O's I had on that second TXB and since I didn't have signals for them, I didn't connect them until just now to do a quick test by re-routing the modem PCM clock signals to the 1.8V input side resulting in the same thing as what you see in my previous screenshot on the 3.5V output side.

Is there anything I can put on the scope probe to see if it's the probe causing the issue?

Wow..... all this time. I switched to x10 as you suggested and now the signal seems nice and clean. Quite honestly I feel stupid for not trying it before. I guess I was too preoccupied by the shape of the signals and didn't even think to change the probe setting!

Leaving for home now but I will re-test tomorrow and confirm.

THANK YOU SIR! Have a wonderful evening!

Benoit