Reply
Highlighted
Posts: 5
Registered: ‎01-04-2017
Accepted Solution

SI5351 MSOP10 package XA used as clock input frequency

Hello,

 

I'm looking at clock builder, and the register map.  I cannot see the reason why Clock Builder insists that XA (crystal) must be between 25MHz and 27MHz.  BTW, the datasheet says -ether- 25MHz or 27MHz.  Is there any reason for this range specification?  Is it related (for instance) to internal startup?

 

Specifically, I need to use a 10MHz OCXO (sinewave) input to XA.  Is it a question of the PLL not locking on power up, simply corrected once the registers are set correctly through I2C, or is there some other issue I should be aware of.

 

D.

Posts: 69
Registered: ‎09-10-2014

Re: SI5351 MSOP10 package XA used as clock input frequency

Hi,

 

The XA/XB pin is connected to an internal crystal oscillator which operates ONLY in the 25-27 MHz range.

 

If you want to use a OCXO, the best choice is to use a Si5351C but that part is not available in the 10 pin MSOP package but available in the 20-pin QFN package. The 20-pin package is actually about the "same area" on the PCB. Given this, will you be able to consider the Si5351C variant instead?

 

Regards,

Hari

Posts: 69
Registered: ‎09-10-2014

Re: SI5351 MSOP10 package XA used as clock input frequency

Hi,

 

The XA/XB pin is connected to an internal crystal oscillator which operates ONLY in the 25-27 MHz range.

 

If you want to use a OCXO, the best choice is to use a Si5351C but that part is not available in the 10 pin MSOP package but available in the 20-pin QFN package. The 20-pin package is actually about the "same area" on the PCB. Given this, will you be able to consider the Si5351C variant instead?

 

Regards,

Hari

Posts: 5
Registered: ‎01-04-2017

Re: SI5351 MSOP10 package XA used as clock input frequency

"The XA/XB pin is connected to an internal crystal oscillator which operates ONLY in the 25-27 MHz range."

 

Really?  Hard to imagine how, since I would assume that the input is a simple Pierce oscillator, with the frequency determining components (crystal resonator) outside.  I suppose it is possible that the internal (perhaps clever and synthetic?) load capacitors somehow introduce unexpected frequency dependent effects...

 

I will set the internal multiplexor to connect the crystal input directly through to an output (bypassing the PLL) and measure input amplitude sensitivity vs frequency (when I get time) and post the results.

Posts: 69
Registered: ‎09-10-2014

Re: SI5351 MSOP10 package XA used as clock input frequency

That is fine. The issue is not that for a "given" part if the internal oscillator operates for any given frequency range. The issue is if we (Silicon Labs) can guarantee the spec over process , voltage and temperature and based on that, we have specified it as such.

 

There is always an alternate of using just the CLKIN pin which just has an input buffer and it is your choice to pursue either option.

 

 

<a href="http://community.silabs.com/t5/Welcome-and-Announcements/Community-Ranking-System-and-Recognition-Program/m-p/140490#U140490"><font color="#000000"><font size="2">Hero</font></font> </a> jmg
Posts: 1,185
Registered: ‎04-27-2004

Re: SI5351 MSOP10 package XA used as clock input frequency


I will set the internal multiplexor to connect the crystal input directly through to an output (bypassing the PLL) and measure input amplitude sensitivity vs frequency (when I get time) and post the results.


Be interested in your results.

 

You can AC couple a Clipped Sine into the Xtal Buffer, and there is a spec somewhere about an upper limit on the swing. 

 

 Can you get a 20MHz or 24MHz OCXO ? - that may be closer to the test range SiLabs use.

 

 

 I have asked SiLabs to fix the software limitations on their Clock Builder, but the wheels turn slowly.

I was told it was coming. I also suggested they test and qualify parts for Clipped Sine drive, but it takes  a while for the SiLabs gears to turn, and grasp there really is a market for that use.

 Clipped Sine VCTCXOs are driven down in price by GPS volumes, and can now come for 62c in modest volumes.

 

Other vendors specify the valid range of PFD and VCO, which are what really limits any PLL.

Everything else is digital dividers.

Posts: 5
Registered: ‎01-04-2017

Re: SI5351 MSOP10 package XA used as clock input frequency

Finally got around to doing this testing.  Capacitively coupled OCXO with 50ohm termination (develops about 350mv RMS) into the XA pin Xtal buffer input.  The results are as one would expect.

 

Almost certainly XA and XB are just the pins of Pierce oscillator, and essentially it is just a gain block inside the chip.  Disable the on-chip capacitors, of course, and it amplifies when you put in XA.  10MHz sinusoid from an OCXO works right out of the box just fine.

 

The 25-27MHz limitation seems only applicable to when the oscillator is used with an external crystal resonator.  SiLabs say they didn't test that (see up thread) but this was the expected result, that driving it externally works just fine at 10MHz.  It is also likely at any frequency the C version CLKIN pin can accept, but this I did not test (not that it's hard to do so).

 

It will be interesting to see the jitter and phase noise achievable with this approach.

<a href="http://community.silabs.com/t5/Welcome-and-Announcements/Community-Ranking-System-and-Recognition-Program/m-p/140490#U140490"><font color="#000000"><font size="2">Hero</font></font> </a> jmg
Posts: 1,185
Registered: ‎04-27-2004

Re: SI5351 MSOP10 package XA used as clock input frequency


Finally got around to doing this testing.  Capacitively coupled OCXO with 50ohm termination (develops about 350mv RMS) into the XA pin Xtal buffer input.  The results are as one would expect.


Good to hear.


Almost certainly XA and XB are just the pins of Pierce oscillator, and essentially it is just a gain block inside the chip.  


Yes, I think a  N_FET, bias resistor G-D, current source fed for a Class-A Stage. This is why the give a 1v p-p spec for XA drive, AC coupled.


  It is also likely at any frequency the C version CLKIN pin can accept, but this I did not test (not that it's hard to do so).


Certainly I'd expect the 10MHz MIN they spec to apply. The 100MHz Max, may depend on how 'good' that Class A Crystal amplifier is at 100MHz...

Still, most TCXOs are well below 100MHz, so even a MAX of 40MHz would include most.

Posts: 66
Registered: ‎11-04-2016

Re: SI5351 MSOP10 package XA used as clock input frequency

Hi ! Can you give me your regs settings ? Mine setup did not working with 10M OCXO Robot Sad 

<a href="http://community.silabs.com/t5/Welcome-and-Announcements/Community-Ranking-System-and-Recognition-Program/m-p/140490#U140490"><font color="#000000"><font size="2">Hero</font></font> </a> jmg
Posts: 1,185
Registered: ‎04-27-2004

Re: SI5351 MSOP10 package XA used as clock input frequency


 Mine setup did not working with 10M OCXO Robot Sad 


You could check the Xtal-out pin with a scope, and config for XTAL/N -> Pin (no VCO.PLL) to check the signal is getting thru ?  What is your OCXO output amplitude ?

Posts: 66
Registered: ‎11-04-2016

Re: SI5351 MSOP10 package XA used as clock input frequency

Addad HMOS buffer (for sine wafe 2 hmos  convert) - now it is works! Thanks! Sine was about 2 v p-p

<a href="http://community.silabs.com/t5/Welcome-and-Announcements/Community-Ranking-System-and-Recognition-Program/m-p/140490#U140490"><font color="#000000"><font size="2">Hero</font></font> </a> jmg
Posts: 1,185
Registered: ‎04-27-2004

Re: SI5351 MSOP10 package XA used as clock input frequency

[ Edited ]

Addad HMOS buffer (for sine wafe 2 hmos  convert) - now it is works! Thanks! Sine was about 2 v p-p


Good that it works, - Curious, did you AC couple (via a small series CAP to Xin), in Xtal mode ?

eg Post #7 has less sine amplitude than yours, and reports ok, 10MHz,  as AC Sine coupled.

IIRC the data suggests 1v p-p is ok ?

 

Posts: 66
Registered: ‎11-04-2016

Re: SI5351 MSOP10 package XA used as clock input frequency

I tried AC coupling, I tried 50 Ohms termination with sine OCXO it was not helped. I think Si5531 wants high slew rate on XA pin.
PS I think Silabs shoked what we done with their chip Robot Happy of course I will use this setup for hobby only, because not guarantee from Silabs
<a href="http://community.silabs.com/t5/Welcome-and-Announcements/Community-Ranking-System-and-Recognition-Program/m-p/140490#U140490"><font color="#000000"><font size="2">Hero</font></font> </a> jmg
Posts: 1,185
Registered: ‎04-27-2004

Re: SI5351 MSOP10 package XA used as clock input frequency


I tried AC coupling, I tried 50 Ohms termination with sine OCXO it was not helped. I think Si5531 wants high slew rate on XA pin.

Interesting, did you try termination/padding down to 1v p-p from 2v, to see if that was ok ?

Too large an AC swing could  start to DC-restore on the Xin clamp diode - one check for that , would be to multimeter thru 100k  XA & XB  pins, first with no signal, and then with the attached sine.

Posts: 66
Registered: ‎11-04-2016

Re: SI5351 MSOP10 package XA used as clock input frequency

Will try tonight, now only 8 AM Moscow. What do you think about resulting output signal quality in this mode?
<a href="http://community.silabs.com/t5/Welcome-and-Announcements/Community-Ranking-System-and-Recognition-Program/m-p/140490#U140490"><font color="#000000"><font size="2">Hero</font></font> </a> jmg
Posts: 1,185
Registered: ‎04-27-2004

Re: SI5351 MSOP10 package XA used as clock input frequency


Will try tonight, now only 8 AM Moscow. What do you think about resulting output signal quality in this mode?

If you have this working with a HCMOS buffer, it is somewhat academic.

If the distance from OXCO to Si5351A is larger, you might prefer to use a more CMOS signal, but if they are physically very close, and you plan volume production, it seems AC couple should be workable, as per #7.

Posts: 66
Registered: ‎11-04-2016

Re: SI5351 MSOP10 package XA used as clock input frequency

It is 74HC125 buffer. No, it is hobby project, but it if i will get good results I will use Si5351C.. (but maybe I will found what cheap 26 MHz TCXO is enough for me). But now no time to wait PCB for 5351C

Posts: 5
Registered: ‎01-04-2017

Re: SI5351 MSOP10 package XA used as clock input frequency

My suspicion is that this is a DC bias problem also.  I don't think the 2v p-p is a problem, I have driven that amplitude into the XA input.

 

What is really critical is to make sure that the XA pin can and does self bias.  That also means you need to make sure there are no other DC paths to ground... your 50r OCXO termination needs to be on the OCXO side of the coupling cap.  I don't remember if there is a bit in the register map to turn the oscillator on, but I suspect if there is I have that set to ensure the bias point is correct (I will dig out the code that does the register setup and post the settings a little later).  The suggestion to check the DC voltage through a resistor (isolate the circuit from the AC capacitive loading of the meter) is a good idea... I will try to do that test and post the results as well.  You can also quickly check what you see at XB with a scope to see if the internal amp is producing output, or if it's output is slammed to rail.

 

As for the HCMOS buffer, that is surely contributing horrible input jitter.  The PLLs will attenuate that, but by how much you would have to measure.  The reason using the XA input is actually better than using a dedicated (digital) clock pin IMHO, is because the XA XB amp is already designed to accept a clean sinusoid at XA (because the nominal 25MHz crystal resonator is very narrow bandwidth).  The appropriate gain to convert that to a clean square is internal to the chip.  In the normal use (25MHz crystal) the slew rate at XA will therefore only be 2.5x what your OCXO is presenting to the pin...

 

Posts: 5
Registered: ‎01-04-2017

Re: SI5351 MSOP10 package XA used as clock input frequency

Here is the minimal register settings I used when initially testing this.  My notes say the PLL is set to 600MHz.  5MHz is output, for the purposes of looking at the jitter between the OCXO input and the output.

 

reg 26 = 0xff = 11111111

reg 27 = 0xff = 11111111

reg 28 = 0x00 = 00000000

reg 29 = 0x2b = 00101011

reg 30 = 0x00 = 00000000

reg 31 = 0xf0 = 11110000

reg 32 = 0x00 = 00000000

reg 33 = 0x00 = 00000000

reg 42 = 0x00 = 00000000

reg 43 = 0x01 = 00000001

reg 44 = 0x00 = 00000000

reg 45 = 0x58 = 01011000

reg 46 = 0x00 = 00000000

reg 47 = 0x00 = 00000000

reg 48 = 0x00 = 00000000

reg 49 = 0x00 = 00000000

reg 177 = 0xa0 = 10100000

reg 16 = 0x4f = 01001111

 

There is nothing very interesting in there, actually... specifically the registers for crystal capacitive loading etc are not written.

 

This is what that looked like on the scope.

Ch1 is the pin output,driving a 50r cable and terminated at the scope

Ch2 is the OCXO which has an external 50r termination.

 

si5351_scope.jpeg