How to use the Wiznet io-module without Transformer

I tried to make a test jig that testing Wiznet modules(W5500-io, W5100S-io) like block diagram below

IP175G is Ethernet Switch IC and wiznet modules controlled by other board called control board.

Test jig board connected with control board using connector(PCC11-100D-2.54DS)

I used only one RJ45(with transformer) in test jig board.

After draw block diagram I searched Wiznet's Phy to Phy design guide but there were not any phy to phy design guide.

Same as me, In other community, many Wiznet Chip users looking for Phy to Phy design guides.

For those who are curious about Wiznet phy to phy application even though there is no design guide, I made a board and tested it.

After soldering, I connect W5500-io, W5100S-io but only W5500-io is worked

even that's connection was unstable.

So I check test jig board's schematic, and W5500-io/W5100S-io's schematic.

Then I figured W5100S-io were no dc bias in phy to phy connection cause no transformer.

See schematic below

when W5100S used with transformer, CT pin (T1 - 2pin, 7pin) is connect with 3V3A and this is DC vias for TX_P/N, RX_P/N

Now see W5100S-io's schematic below

J1 - 3pin, 4pin seems like CT pin so if you use W5100S-io with transformer

J1 should be connect like this

But in Phy to Phy transformerless connection, there are no CT pin.

For these reason, in phy to phy connection, W5100S-io's schematic should be edit like this for DC bias.

Then let's look at the case of the W5500.

As you see W5500 Ref schematic, there are only DC bias voltage in TCT.

RCT is AC to GND through capacitor.

These means only TXP/N should be connected 3V3 dc bias.

Fortunately, in W5500-io, TXP/N is already connected 3V3A through 49.9R

So W5500-io is no need to edit schematic for phy to phy connection.

However, the instability of the connection has not been resolved, so I looked at the test jig board schematic.

As a result of testing while changing the dc block capacitor value, the connectivity was the best at 2.2nF.

So I changed all dc block capacitor 100uF -> 2.2nF

(Uploaded schematics are unmodified)

also, RJ45's Link LED signal should be edit HUB1_LINK0 -> HUB1_LINK1.

(Uploaded schematics are unmodified)

Aftet do this all, I tested again.

I connected W5100S-io's SPI line with Raspberry pi Pico for loopback test.

you can find the code from this link(https://github.com/Wiznet/RP2040-HAT-C/blob/main/getting_started.md)

The result was very good. loopback code was working well, also connection was stable

The oscilloscope waveform was also compared to compare the waveforms of the phy to phy connection and the connection using a transformer.

I only measured the RXP/N of W5100S-io.

This is result.

Waveform between using transformer and transformerless is almost same.

The results of this experiment are as follows.

1. When you use W5100s, W5500 and want to transformerless connection, then use 2.2nF for DC block capacitor.(of course transformerless application must be use in same board and MDI line should be as short as possible)

2. In case of W5100S, connect 3V3 DC bias voltage with TXP/N, RXP/N through 49.9ohm

3. This is not Wiznet's official Phy to Phy design guide, so please just refer to it.

If you have any questions or incorrect information in this article, please leave a comment.

Thank you for reading and I hope this article will be helpful to many people.