Reputable source for Si5351A clock PLL?

[ebastler]

I tend to think that people cry "fake IC!" too quickly -- for me personally, when one of my circuits does not work, the problem is usually me.   But now the fake IC problem seems to have caught up with me:

I had ordered five Si5351A (MSOP-10) from AliExpress, since they were out of stock at the major distributors. And I could not get these ICs to work properly: The output, approx. 5 MHz on a single output channel, jitters massively and has regular periods where it drops out entirely for approx. 700µs. Meanwhile the 25 MHz input clock looks nice and stable. I was wondering whether the ICs might be pre-configured with some unusual settings and have changed my configuration software to write the full complement of 43 registers calculated by SiLab's Clock Builder -- but no improvement.

Finally I swapped another of the newly bought chips with the Si5351A on a small no-name prototype module I had bought last year. And presto: The old chip works nicely in my new PCB layout, and the old prototype module now exhibits the jitter and dropout problem. So I conclude that the AliExpress ICs are indeed bad.

Which raises the question: Is anyone aware of a reputable source for the Si5351A -- in small quantities, and with a delivery date before Christmas?

Alternatively: Any idea in which way my batch of AliExpress chips could be "different" from the regular ones, short of being fake or factory rejects? Anything else I could try in the software configuration?

Thank you for any ideas!

[quadtech]

From the ubitx forum: The MS5351M is a Chinese version of the Si5351a that is apparently
used in the Chinese space program so availability may be a bit better than the SiLabs 5351a.

https://groups.io/g/BITX20/message/87189?p=,,,20,0,0,0:

[ebastler]

Thank you, I was not aware of that clone. Available from LCSC, which should be fine.

The units I got via AliExpress are hopefully not rebadged MS5351M, since they definitely do not work properly. Soldered into a know-good commercial evaluation board, they show massive jitter and periodic drop-outs in their default startup configuration, and can't be convinced to operate properly by reconfiguring them with known-good settings. Let's hope that these were factory rejects or such, and that the MS5351M works just like the proper Si Labs part. 

[jensrenner]

Exactly the same has happend to me / our company.
I am using that chip for years now in our own SDR. But for the current production batch, we couldn't get hold on any Si5351A-B-GT (MSOP-10) anymore. So I was checking AliExpress as a last resort and ordered some samples from two sellers.

Before taking a closer look at them I simply soldered the first sample on a working SDR board (i.e. replacing the original SI5351A), just to find that the PLL would never lock. I used a scope to check the outputs and could observe exactly the same behaviour as described by @ebastler
Unstable, jitttering signal and periodical drop-outs. So I took the other sample from the second seller, same behaviour. Just to make sure I did not damage anything on the board, I took the original (desoldered) Si5351A and put it back in. Everything was ok again. Stable PLL reference and  ADC sample clock.

So I decided to take a closer look. The top marking on both China ICs appeared to be the same as on the original one. But on the first China IC, the font was different. On the second chip, the font looked pretty much the same compared to the original, but the date code was reading 2xx (xx = two digits). 2 would mean either 2012 or 2022 and that doesn't seem to be correct.

Bottom line is, I am conviced that this is a China clone (not even scrapped genuine parts) and it's not worth the effort or the risk to make it work on our board.

During my research I found that there is an "official" China clone named MS5351M in an MSOP-10 package, that's how I discovered this thread. That clone is probably different from the Si5351 sold on AliExpress. I am not going to try it though. Instead, our company decided to buy "new old stock" from a reliable broker for 10 x the original price.

[ebastler]

In the meantime, I have received another set of five Si5351A from AliExpress, which I had ordered shortly after the first lot. Those have a different manufacturing and date code than the earlier units, and the new chips turned out to be good.

Reassuring to know that those "gray" batches are not all bad, and I am set for now. But of course taking this kind of gamble is not an option for professional use.

The marking on the bad chips reads "5351 AUBI 248", by the way. The new, good ones are "5351 BYOO 016". As mentioned by Jens, the "2xx" in the date code is suspicious...

[kronos]

I bought a SI5351A breakout board from ebay more than 1 year ago, for use as a programmable oscillator. It did not work properly, and took me a time (I am newbie, I thought it was a SW problem) to discover that the chip was probably defective: it showed a terrible jitter and periodical drop-outs.
After a lot of playing with different libraries, I discovered that keeping the PLL frequency below a threshold (835 MHz in my case) eliminated these problems. It seems to be a defective PLL. I also have a NanoVNA V2 which exhibits the same behaviour: periodical dropouts. It seems it uses these defective chips.

I recently bought another breakout board (AliExpress) and this one works perfectly. And I discovered that not only the PLL has a problem in the old chip: the output levels were half of the ones in the new chip, when loaded with 50 ohm. In all drive strengths, 2 to 8 mA. So not only the PLL was defective.

[Bassman59]

I bought a SI5351A breakout board from ebay more than 1 year ago, for use as a programmable oscillator. It did not work properly, and took me a time (I am newbie, I thought it was a SW problem) to discover that the chip was probably defective: it showed a terrible jitter and periodical drop-outs.
After a lot of playing with different libraries, I discovered that keeping the PLL frequency below a threshold (835 MHz in my case) eliminated these problems. It seems to be a defective PLL. I also have a NanoVNA V2 which exhibits the same behaviour: periodical dropouts. It seems it uses these defective chips.

The 5351A maximum output frequency is 200 MHz and as far as I can tell the max VCO frequency is 800 MHz.

Their ClockBuilder software won't let you create a register set which exceeds 200 MHz output or VCO about 800 MHz.

If you're using someone's library, that library may not respect -- or even care about -- those limits.

[rcbuck]

Just a note on the MS5351M part. Hans of QRP Labs has tested the part and says it performs slightly better than the Si5351A. Link to the tests here:
https://www.qrp-labs.com/synth/ms5351m.html

I needed to order some toroids from Toroid  King a couple of weeks ago. They are stocking the MS5351M part so based on Hans' report I added 2 pieces to my order. I have a custom Si5351A board that I designed for my ham projects. I placed the parts on two of the boards. They perform exactly like the Si5351A parts do. Looking at the output on my spectrum analyzer I do not see any jitter or dropout. The spur level is equivalent to the Si5351A.

I think you will be fine if you use the MS5351M part.

[kronos]

The 5351A maximum output frequency is 200 MHz and as far as I can tell the max VCO frequency is 800 MHz.

According to the docs, the VCO frequency is 600-900 MHz. It seems to me some chips are either fake or defective and a symptom is that they do not reach those 900 MHz.

[MisterHeadache]

Just a suggestion - confirm the spread spectrum mode isn't turned on by default on your chips.  I had that issue on a recent Si5351 project.  Several of the public code libraries do not bother to turn that mode off on start up.

[Southerner]

The Adafruit breakout board/modules have spread spectrum turned on by default but that should not impact the upper limit.

[Dave_g8]

Hi All,
For information, I received a couple of boards some time ago from eBay which appeared to have Revision A devices fitted.
The device marking was “5351 AJXO 204” where the second line “A” indicates Revision A.
The REVID register = 0, both boards had low VCO range and limited output drive.

I recently replaced the SI5351A on both of the boards with Revision B devices, device marking “5351 B8D6 228”.
These were purchased from RS components in the UK for £1.34 each.
Both boards now have a REVID register = 1, operate over the full VCO range (actually better than specification), and the output drive is good.

Regards, Dave