Couple questions re: Si5338

Discussions related to schematic capture, PCB layout, signal integrity, and RF development
Post Reply
KE5FX
Posts: 7
Joined: Fri Aug 19, 2016 2:24 am

Couple questions re: Si5338

Post by KE5FX »

I have one of the boards from the original Kickstarter campaign, which I've just now started to play with, and ran into a couple of questions with regard to how the Si5338 interacts with the SMB CLK input/output jack.

The schematic shows the SMB jack tied to pins 4 (IN4/FDBK) and 10 (CLK3A) of the Si5338. That makes sense in a master-slave setup, where there would need to be a feedback input to give the master access to its own output clock. Trouble is, I have an Si5338A on this board, but it seems that the Si5338A can't use pin 4 as an input. This functionality apparently requires the Si5338N. According to the data sheet, pin 4 of the Si5338A is labeled I2C_LSB, and is used to set the low bit of the part's I2C address.

Possibly as a result of this, I can't seem to use any of the set smb_mode commands in bladerf-cli, including the output command. It will send 38.4 MHz to the CLK jack as expected, but then the board locks up, presumably because it doesn't see its own loopback through pin 4.

I happened to have an Si5338N, so I replaced the Si5338A on the bladeRF with it. This didn't work at all. The clock outputs are all stuck at 1.4V or so, and the FPGA never gets recognized. Where bladerf-cli would normally say "Read FPGA version: 0.5.0", it times out with "Transfered (sic) failed" and "Failed to read FPGA version[0]: File or device I/O failure."

It's not impossible that something else went wrong with the chip swap, but I thought I'd ask a few basic questions before going too much farther.

1) What Si5338 variant should be present? Specifically, what's up with pin 4?

2) Is the Si5338 factory-programmed or otherwise OTP'ed? Should it power up with output clocks active? If not, how does NIOS bootstrap the Si5338 without a valid c4_clock signal?

3) Are the earliest boards 100% compatible with the current ones, in general? There are definitely a few differences -- for instance, I don't have C248 or C242 as far as I can tell.

TIA for any input.
bpadalino
Posts: 303
Joined: Mon Mar 04, 2013 4:53 pm

Re: Couple questions re: Si5338

Post by bpadalino »

To answer your questions:

1) Si5338N/P/Q are fine. We should be using Si5338Q.
2) It is OTP'd. I will discuss NIOS bootstrapping a little later.
3) Is your board green or black? If it's green, then no. If it's black - then it should be a good board. Surprised you don't have some of the components on there?

The Si5338 has an OTP to it that allows the 38.4MHz reference to go through to the C4 as the C4 clock. Absent this, you can still feed a clock into the NIOS from the internal C4 oscillator if you want to try. We don't bootstrap the whole Si5338 through the library, though, since the device should have had the OTP.

Let me know if you want more info on the internal C4 oscillator to feed the NIOS so you don't get transfer timeouts. Also let me know if you want more info on what a default config is for the Si5338.

Brian
KE5FX
Posts: 7
Joined: Fri Aug 19, 2016 2:24 am

Re: Couple questions re: Si5338

Post by KE5FX »

bpadalino wrote:To answer your questions:

1) Si5338N/P/Q are fine. We should be using Si5338Q.
2) It is OTP'd. I will discuss NIOS bootstrapping a little later.
3) Is your board green or black? If it's green, then no. If it's black - then it should be a good board. Surprised you don't have some of the components on there?

The Si5338 has an OTP to it that allows the 38.4MHz reference to go through to the C4 as the C4 clock. Absent this, you can still feed a clock into the NIOS from the internal C4 oscillator if you want to try. We don't bootstrap the whole Si5338 through the library, though, since the device should have had the OTP.

Let me know if you want more info on the internal C4 oscillator to feed the NIOS so you don't get transfer timeouts. Also let me know if you want more info on what a default config is for the Si5338.

Brian
It's black, but again, a really early one.

Can you email the .nvm file you use for programming your Si5338Qs? (john (at) miles.io) It would be good to get this board up to date, or as up-to-date as possible.
robert.ghilduta
Posts: 156
Joined: Thu Feb 28, 2013 11:14 pm

Re: Couple questions re: Si5338

Post by robert.ghilduta »

Hi KE5FX,

I believe you have a preproduction unit. You have been emailed.

Regards,
Rob
turan304
Posts: 1
Joined: Tue Aug 07, 2018 1:06 am

Re: Couple questions re: Si5338

Post by turan304 »

KE5FX wrote: Mon Apr 10, 2017 5:56 pm
bpadalino wrote:To answer your questions:

1) Si5338N/P/Q are fine. We should be using Si5338Q.
2) It is OTP'd. I will discuss NIOS bootstrapping a little later.
3) Is your board green or black? If it's green, then no. If it's black - then it should be a good board. Surprised you don't have some of the components on there?

The Si5338 has an OTP to it that allows the 38.4MHz reference to go through to the C4 as the C4 clock. Absent this, you can still feed a clock into the NIOS from the internal C4 oscillator if you want to try. We don't bootstrap the whole Si5338 through the library, though, since the device should have had the OTP.

Let me know if you want more info on the internal C4 oscillator to feed the NIOS so you don't get transfer timeouts. Also let me know if you want more info on what a default config is for the Si5338.

Brian
It's black, but again, a really early one.

Can you email the .nvm file you use for programming your Si5338Qs? (john (at) miles.io) It would be good to get this board up to date, or as up-to-date as possible.
Hi
I couldn't create nvm file for Si5338 via Clock Builder Desktop . How can I create this ? Could you send me any nvm file ?
Post Reply