Galvanic isolated L-Link ??

[lilltroll]

Is it possible to use GMR isolators on L-Links ?

For an example Si846x https://www.silabs.com/Support%20Docume ... Si846x.pdf

Assume you need a device with galvanic isolated ADC ,DAC and isolated figital IO.

To isolate all pins between a XCORE and a multi-channel DAC is expensive due to the amount of pins needed including I2C for setup, but if you instead can isolate the L-LINKS between XCORE chips a +100 Mbps "2+2 way" chip would cover it.

[Allein]

Isolating a 133MHz XLink (7.5ns bit period) with such a 150MHz GMR may work... or not...
The issue is the signal distorsion (1.5ns according to DS) to be added to the signal skew (you know, the eye diagram). So, without a serious SI analysis or practical validation, and based on my experience of such near-limit cases, I wouldn't assert that it will be reliable for sure.
Another way, you may have looked at, would be to use a serial SPI ADC and to isolate the SPI itself: there are two or three SPI GMRs at least in the market.

PS: should you need a contractor for such developments, feel free to contact me through my web site (see profile)

[lilltroll]

I guess I will order a GMR isolator and try what Link speed that can be achieved between 2 pc XK-1

to be cont.

[lilltroll]

PicoDump1.png

First I manage to run the L_link a lot slower. With delay=200,200 in the xn file it looks like this
sending 0xAAAAAAAA on a streaming channel.

[lilltroll]

A good start

Code: Select all

Available XMOS Devices
----------------------

  ID    Name                    Adapter ID      Devices
  --    ----                    ----------      -------
  0     XMOS XTAG-2             4afae0aa        L1[0..1]

[lilltroll]

It works, and the signal looks cleaner on the output on the GMR isolator than on the input.

At delay=0,1 it takes 1.6 us (Seen on the scope) to send 8 int32 or 32 bytes (2w streaming channel) = 160 MBit/s

Using a 2w packetized link with 8 int32 it takes 2.6us (Seen on the scope) = 98 Mbit/s

[bearcat]

The SI parts work great to isolate an XLINK. I have been running a 2W Delay 6,6 XLINK with no issues using the 150MHz parts. Those speeds met my criteria, and seemed reasonable on the scope. I initially performed a solid 2 day test passing known values between the cores, and had no errors. As you crank up the speed of the XLINK, you will notice the voltage levels will start to reduce, with the eye getting smaller. This starts to happen well before 150MHz.

Using a channel in streaming mode, has some limitations with XC that need to be realized.

[Folknology]

Nice work Lilltroll, this is a great way to provide galvanic isolation with good throughput and latency, whilst simultaneously not having to jump through programming hoops (transparent use of channels). This may be useful with some of our robotics stuff, will need to play with this later in the year.

regards
Al