CLKIN electrical characteristics

We are using W5500 with external 25MHz clock. Clock input XI/CLKIN is connected to OSC_OUT of STM32F412 (ARM Cortex M4). Signal is stable, sinusoidal, voltage level about 400mVpp.

Everything seems to be working fine, but we need to confirm, is such connection within the electrical specifications of the W5500? What would be the minimum signal level expected on XI/CLKIN input?

Datasheet says:

Crystal input / External Clock input
External 25MHz Crystal Input.
This pin can also be connected to single-ended TTL oscillator (CLKIN). 3.3V clock should be applied for the External Clock input. If this method is implemented, XO should be left unconnected. Refer to the ‘Crystal reference schematic’ (Figure.3) for details.

What’s this? Does it stand for 400 millivolts peak-to-peak?

1 Like

Thanks for the answer. Yes, we are aware of datasheet, it gives the recommended level (3.3V), but doesn’t specify the minimum.

Today we tested with a different attenuation resistors. W5500 works until about 200mVpp (millivolts peak-to-peak), so I’d say our 400mVpp is pretty much on the safe side. To be more precise, between a few samples, the actual oscillator was 420-480mVpp.

Still, we’d like to have the official data, it would be nice of Wiznet to specify the minimum level, within the other electrical characteristics.

I think datasheet is pretty clear on it. It says that oscillator input must be fed with single ended TTL, and it means that voltage must be in range of maximum 0-3.3V, and minimum 0.8-2.0V, and this minimum implies 2-0.8=1.2V peak-to-peak of the square wave with offset of +0.8V.

Sorry, I’m not following, where did you get those numbers?

Data sheet page 9 define XI/CLKIN as “AI” (analog input), so DC characteristics of digital i/o don’t apply.

Check by yourself, with a crystal connected to XO/XI (figure 3), neither of two signals are square. The circuitry inside is not digital, it is (most probably) a Pierce oscillator, with a linear amp of finite gain. The non-linearity of this amp is another (not unrelated) story, it produces more harmonics than expected.

I go here, then click “datasheet” to the right mid of the page, and go to page 10/68 of the datasheet.

I recall we already had this discussion before, and someone was already noticing it. Here’s the related post.

We concluded that external oscillator must be powered by the 3V3A (analog 3.3V), and this is why it has AI/AO label in the table.

I understand your reasoning, and partly agree with the conclusion, in some cases it is a better choice to use analog power supply for the external oscillator. However I disagree on this as a general rule, it’s easy to think of example where digital 3.3V is a better choice.

Whatever the reason Wiznet had for marking XI/CLKIN as analog, it surely doesn’t imply that circutry for XI/XO is digital, or that DC characteristics for digital i/o may be applied to XI/CLKIN.