Silicon Labs clock generator IC 514CCC000926AAG

Comments

• mikeologist Posts: 337

  2017-12-12 20:27

  MIchael_Michalski wrote: »

  It seems they made 50 of them, sent me my two and then put the other 48 up for sale at mouser. Makes it real easy to get more if it works out well.

  
  Do you have the mouser link?
  I'll pop on a couple to play with them. I'd love to replace my 555 generator on my bench.
• tonyp12 Posts: 1,950

  2017-12-12 20:27

  Pretty expensive though
  1: $12.28
  25: $8.84
  100: $7.02
  1,000: $6.23
  https://www.mouser.com/ProductDetail/Silicon-Labs/514BBC000932BAG/?qs=sGAEpiMZZMve4/bfQkoj%2bKNv6%2b%2bs0%2bSkk/XKLqyHHkk=
  
• jmg Posts: 15,148

  2017-12-12 21:45

  tonyp12 wrote: »

  Pretty expensive though
  1: $12.28
  25: $8.84
  100: $7.02
  1,000: $6.23
  https://www.mouser.com/ProductDetail/Silicon-Labs/514BBC000932BAG/?qs=sGAEpiMZZMve4/bfQkoj%2bKNv6%2b%2bs0%2bSkk/XKLqyHHkk=

  
  Yes, if you want a lower cost solution, the Si5351A is 96c/100, similar i2c Synth, with external crystal.
  
  Whilst the Si514 costs more, but does have it all in one package (Xtal included), and has a higher freq VCO, so gives lower jitter.
  
  Default ppm is not great on Si514, but you can calibrate from a GPS 1pps or similar.
  
  I like the combination of a GPS TCXO (0.5ppm) ($1.24/100) and a Si5351A.
  
• Phil Pilgrim (PhiPi) Posts: 23,514

  2017-12-12 22:39

  The 5 MHz external oscillator used as the FLiP's clock is only ~2mm square. You can't get a 5 MHz crystal close to being that small, so how do they do it? I've heard of silicon oscillators, but they're not renowned for tight PPMs.
  
  -Phil
• jmg Posts: 15,148

  2017-12-13 00:55

  Phil Pilgrim (PhiPi) wrote: »

  The 5 MHz external oscillator used as the FLiP's clock is only ~2mm square. You can't get a 5 MHz crystal close to being that small, so how do they do it? I've heard of silicon oscillators, but they're not renowned for tight PPMs.

  
  BOM says FLiP uses a MEMS Oscillator, SIT8918BE
  Those come from ±50ppm to as good as ±20ppm, I guess FLiP uses the cheapest ?
  
  - MEMS use an etched vibrating element, cheaper than Quartz, but not quite as high Q, and they do seem to be steadily improving.
  
  eg I find this one at ±5ppm ( note current has climbed from the 4.5mA of a SIT8918 )
  
  SIT5000AICGE-33N0-25.000000Y SiTIME OSC MEMS 25.000MHZ VCTCXO $1.65/1k LVCMOS 3.3V ±5ppm -40°C ~ 85°C 33mA
  
• Tracy Allen Posts: 6,656

  2017-12-13 17:31 edited 2017-12-13 17:35

  About the SIT8918 on the FLIP, a reminder from this thread that it was found to interfere with GPS lock due to a strong harmonic at at the L1 1575MHz frequency. A different version of the same chip, that is, SIT8918BEL13-33E-5.000000E (with "L" in place of "-" in the default) offers slower clock edges. I traded out on FLIP with samples that Silicon Labs provided that and found that the GPS interference went away, and Parallax was able to verify that also. I'm told that the next build of FLIPs will include that change. It is something to be aware of with the current batch if you are running sensitive RF receivers near the FLIP; there is a comb of odd harmonics of 5MHz.
  
  
• jmg Posts: 15,148

  2017-12-13 18:18 edited 2017-12-13 18:19

  Tracy Allen wrote: »

  About the SIT8918 on the FLIP, a reminder from this thread that it was found to interfere with GPS lock due to a strong harmonic at at the L1 1575MHz frequency. A different version of the same chip, that is, SIT8918BEL13-33E-5.000000E (with "L" in place of "-" in the default) offers slower clock edges. I traded out on FLIP with samples that Silicon Labs provided that and found that the GPS interference went away, and Parallax was able to verify that also. I'm told that the next build of FLIPs will include that change. It is something to be aware of with the current batch if you are running sensitive RF receivers near the FLIP; there is a comb of odd harmonics of 5MHz.
  

  Good to have a follow-up on that issue.
  Did Parallax also improve the ppm at the same time ? - which would be SIT8918BEL11-33E-5.000000E I think ? ( improves ±50ppm to ±20ppm)
  
  
• MIchael_Michalski Posts: 138

  2017-12-13 18:35

  The si514 claims to use a crystal. Its 30ppm accuracy is not particularly bad. (I ordered the most accurate/stable one of that series of parts) The 5mhz crystals parallax sells (and presumably the one in the quick start board that Im going to use this with) claims 30ppm initial accuracy and 50ppm frequency stability. The 514 claims 20ppm temperature drift and 30ppm overall accuracy including temperature drift,intial innaccuracy ,aging over a 10 year lifespan, supply and loading variations,mechanical shock,etc. I wonder how stable it could be if you were to put it in a temperature compensated oven,tightly control the supply voltages,and carefully buffer the output. It seems if you really wanted to all of those things could be eliminated. Of course, it might be easier to simply get a very accurate and stable fixed frequency oscillator, run both into a pair of counters and just adjust out any drift.
  
  My motive for getting these was actually for a USB driver Im writing. My plan is to use this to both run the propeller at a multiple of 6mhz for timing reasons and to possibly be able to adjust the timing as per the usb spec to match the transmitter when receiving.
• jmg Posts: 15,148

  2017-12-13 21:02 edited 2017-12-13 21:05

  MIchael_Michalski wrote: »

  The si514 claims to use a crystal. Its 30ppm accuracy is not particularly bad. (I ordered the most accurate/stable one of that series of parts)

  It does use a Crystal, but 30ppm these days, is not great precision.
  You can buy ±500ppb parts, for under $1/1k - but not in < 10MHz values.
  

  MIchael_Michalski wrote: »

  I wonder how stable it could be if you were to put it in a temperature compensated oven,tightly control the supply voltages,and carefully buffer the output. It seems if you really wanted to all of those things could be eliminated.

  Not quite eliminated, but those are called OCXO (Oven controlled Xtal Oscillators) and they can deliver low parts per billion (for a price...)
  
  

  MIchael_Michalski wrote: »

  Of course, it might be easier to simply get a very accurate and stable fixed frequency oscillator, run both into a pair of counters and just adjust out any drift.

  GPS modules these days usually have 1pps out, so that gives an excellent calibration reference to check drift.
  

  MIchael_Michalski wrote: »

  My motive for getting these was actually for a USB driver Im writing. My plan is to use this to both run the propeller at a multiple of 6mhz for timing reasons and to possibly be able to adjust the timing as per the usb spec to match the transmitter when receiving.

  
  For that, a Si5351A is probably good enough - Adafruit have a sub $7.95 breakout board with one
  
  Of course, if you want to "match the transmitter when receiving" you need to measure the actual transmitter value, which may not be exactly 6.0000MHz
  
  A single 80MHz clock in the 1ms USB frame time, is 12.5ppm, or 100 frames would measure to 0.125ppm LSB
  A one-half 12Mhz period movement, in 1ms, is ~ 41.7ppm, one quarter period is ~ 20ppm
  
  PC clocks may not be nearly as precise as we may hope.
  
  I wonder what the minimal setup needed to 'continually keep USB alive with 1ms frames' is ?
• MIchael_Michalski Posts: 138

  2017-12-13 22:52

  T
• MIchael_Michalski Posts: 138

  2017-12-13 23:40

  The USB specification suggests measuring the frequency of the bit transitions in the sync field of a data packet and then adjusting the clock to match. The 20ppm thermal drift is not relevant because it not going to happen over the time scale of a single packet. I may not bother though. The USB clock is 12MHz, so 6MHz is actually the value of the crystal you need if you want to simply go the fixed frequency route,to get a system clock of 96MHz, which gives 8 clocks or 2 instructions per core , per bit period.) The specification states that the data signaling tolerance is 2500ppm. The signaling period comes out to about 83-1/3 nanoseconds. (The period of the system clock is about 10.4ns) The allowable deviation in the transmit rate is then about 30khz. The receiver must also be able to handle a signal that is that far off,which is the more difficult part. Max transfer size for isochronous full speed packets is around roughly 1000 bytes, or 8000 bits. That comes out to about 666us. The timing error during that time period could be as much as 1666ns. At 30ppm, that would be about 20ns,or about a quarter bit period,which would be ok. Of course ,I dont HAVE to accept 1000 byte packets and that will make things considerably easier. The maximum size I HAVE to accept is 64 bytes,or 512 bits. That comes out to about 43 microseconds. With a timing error of 2500ppm, that would still be over 100ns. Still way too much. But at 30ppm,which is what my clock is rated at,that would be about 1.3ns,which isnt a problem at all. If I synchronize the clock at to the sync field,I should be able to get much better than that and easily be able to receive large packet sizes.
• SaucySoliton Posts: 488

  2017-12-16 07:09

  The Propeller 1 isn't very suitable for complying with USB specs. In addition to the timing tolerance, I think it suggests a differential receiver. Nevertheless, the tenacious folks on Parallax forums have gotten USB working anyway. BradC did the low speed device and Micah (scanlime) did the full speed host. Yours truly is responsible for the low speed host. I think full speed device can be done but it will be tricky due to the quick response times required by the spec.
  
  The Propeller full speed host uses a fixed sampling clock. With the same quarter bit tolerance it should handle 450ppm for a 64 byte packet. I have not tested that. In practice, no crystal controlled device should be that bad. Despite the massive non-compliance it "just works."