Surface mount crystal
bnikkel
Posts: 104
in Propeller 1
im designing a .8"x1" board and i have everything on it i need, but i have the same crystal as on the prop mini and was wondering if there is surface mount version that is compatible with the prop? has anyone used one or know of one? the one used on the flip wont work because its too small, im hand soldering, im pretty good but not that good haha.
Comments
-Phil
It is hard to find 5MHz xtals in small SMT packages. The smallest I have is 5x7mm. On the other hand, 10MHz xtals in much smaller SMT packages are easy and work fine with the Prop (up to PLLx8).
As above, the Xtal element is physically too large to fit into tiny packages, so you can buy 5MHz Xtals but not that much smaller
If you are ok with reduced precision, there are SMD ceramic resonators as small as (4.50mm x 2.00mm)
Otherwise, you need powered Oscillators, and those can come quite small.
eg As small as (1.60mm x 1.20mm) for DSC6001HI1A-005.0000T ±50ppm, or even smaller to 1.5x0.8 BGA parts, for SiT8021 (±100ppm only)
If you have a need for real precision (single digit ppm?) and 5MHz, that's tougher, but I do spot a non-stocked part code for a
ASTX-H12-5.000MHZ-I25-T 2.5ppm 5MHz 2.5 mm x 2 mm
Best precision stocked looks to be ±15ppm
ECS-3225S33-050-EN-TR Miniature Crystal Oscillator HCMOS Tight Stability 3.3V 5MHz ±15ppm -40 to 85°C 3.2x2.5x0.9mm SMD T/R
and that supplier and family, does have down to this, as a valid part code, if your volumes are good
ECS-2520S33-050-FN-TR Miniature Crystal Oscillator HCMOS Tight Stability 3.3V 5MHz ±10ppm -40 to 85°C 2.5x2x0.9mm SMD T/R
re-checking into that FLiP osc, gives BOM was SIT8918BE-13-33E-5.000000E (2520), which now shows as 'Not For New Designs', so they must have an alternative ?
There is also this older thread, where the lower slew, lower RFI oscillator SIT8918BEL13-33E-5.000000E was tested in FLiP
https://forums.parallax.com/discussion/comment/1414277/#Comment_1414277
Maybe parallax should qualify the physically compatible, alternative of
ECS-2520S33-050-FN-TR 3.3V 5MHz ±10ppm -40 to 85°C 2.5x2x0.9mm
a few more cents for a good boost in precision ?
SIT8918BEL13-33E-5.000000G 3.3V ±50ppm -40°C ~ 105°C 4.7mA 2.50mm x 2.00mm $0.96288/1000
Hehe, look, it is the lower slew variant, & the same part code Tracy Allen confirmed with FLiP+GPS
- which make sense to be a default for a 5MHz Osc.
hmm... I wonder if a sample request, like Tracy did, triggers a build of 200 pcs and those go onto Digikey's shelves ?
Good to see stock of a lower-RFI choice, should any FLiP users need it
Comparing with Avnet
ECS-2520S33-050-FN-TR 3.3V 5MHz ±10ppm -40 to 85°C 2.5x2x0.9mm $0.7889/1k
so you need to order 1000, but the 5x better spec of ±10ppm, is cheaper than Digikey's price for ±50ppm SIT8918
If you are ok with thru-hole, another choice could be 'watch crystal' cylinder packages, you can get 8mm x 3mm dia 5MHz Xtals.
Smallest SMD 5MHz looks to be ~ (8.00mm x 4.50mm)
±50ppm will likely be just fine, but if there is a low cost (or sometimes even lower cost, as above) better ppm choice, then take it.
The very low ppm are usually used for Test & measurement systems, but it does no harm to have better precision on a general board
Larry, do yourself a favor and don't even bother with those huge and horrible SMD HC49 rocks. If you are going to use HC49 then use through-hole because at least they don't need huge pcb smd pads out to the side which was the industry's feeble attempt at SMD at the time, and somehow it stuck. However there are plenty of 6MHz and 10MHz rocks available in all sorts of shapes and sizes and cheap.
Here's a 2 pad 10MHz 60 cent SMD 5x3mm that's easy to solder even with an iron, just make the pcb pads protrude out a little. I've used similar ones in commercial products for many years and my startup routine automatically checks for 5 or 10MHz crystal to end up with 80MHz but it is just as easy to specify that in your code.
CSTNR5M00GH5C000R0 4.5x2.0mm 5.0MHz +/-0.07% +/-0.11% max. +/-0.07% max. 60Ω 39pF -20℃ to 85℃ Consumer/Industrial 26c/3k
CSTNR5M00GH5L000R0 4.5x2.0mm 5.0MHz +/-0.10% +/-0.13% max. +/-0.07% max. 60Ω 39pF -40℃ to 125℃ Automotive 26c/3k
As long as I can solder them with a fine tip iron they are worth a try. I'll include the footprint in my next pcb.
Yes, they are easy but I find a good flat tip much more usable than a fine needle tip since they have problems conducting enough heat to the joint. The flat tip also allows you to solder QFP's easily.
SIT8918BEL13-33E-5.000000G.
For some time now, no interference with nearby GPS.
However, one has to keep in mind that clock oscillators, despite their tiny tiny size, are not a lowest power solutions. They keep running at ~4mA even when the Prop is switched to RCslow. I suppose some trickery might be conceived to put it into µA standby mode, using its OE pin. ("E" in the part designation.)
According to SiTime's Part Number Decoder:
https://sitime.com/part-number-decoder
the part that is being used at Flip (SIT8918BEL13-33E-5.000000G) has only the Output Enable function, available when Pin 1 is pulled Low, as per SiT8198B datasheet (Tables 1, 2).
By driving Pin 1 to a Low state, the present device would drop its I_od comsumption to 4.5 mA, wich is no big savings at all.
Keeping the other specs, for it to alternativelly have the Standby function available its part number would be: (SIT8918BEL13-33S-5.000000G). Then I_std (Table 1) would rule, dropping power consumption to mere 8.5 uA max. (@3.3 V).
Hope it helps a bit.
Henrique
A move to ±20ppm from ±50ppm would have been nice too, but that does add a few cents.
Fair point. When running, the ~4mA is not the total change, as there is a slight saving from the Xtal-Amplifier not being used linear ( 0.3mA @ 5MHz in the data )
There are lower power oscillators, but that comes at the cost of precision, & worse jitter, which drops to ±100ppm
eg
SiT8021 160 μA(MAX) f = 6.144 MHz, Vdd = 2.25V to 3.63V, no load ±100ppm (1.54mm x 0.84mm) BGA4
That one may give less total power than using a Crystal.
It is a moot for controlling power on the FLIP, which uses the "E" version, and has pin 1 hard wired high and 100% enabled anyway.
Someone designing their own tiny board and in need of shutdown could probably come up with a scheme with the "S" version, to shutdown the clock oscillator using an auxiliary Prop output pin. The Prop can be set to boot up in RCfast, then to switch over to the more accurate oscillator when needed.
SitTime claims to have a USB field programmer for these things, the "time machine 2", so, churn out whatever variant in frequency and features you want.
JMG suggested this one
CSTNR5M00GH5C000R0 4.5x2.0mm 5.0MHz +/-0.07% +/-0.11% max. +/-0.07% max. 60Ω 39pF -20℃ to 85℃ Consumer/Industrial 26c/3k
Take my advice and just use a 10MHz SMD since I have commercial designs running with these for close on 10 years without any problems.
Change
to
The Prop data sheet formally says this (PLL is always XTAL * 16)
"The PLL internally multiplies the XIN pin frequency by 16. The PLL’s internal frequency must be kept within 64 MHz to 128 MHz – this translates to an XIN frequency range of 4 MHz to 8 MHz."
- but as Peter mentions, many 'get away with' using 10MHz Xtals and the 160MHz PLL.VCO that then operates at.
If your application is not commercial, or your temperatures are not extreme, it's easy enough to trial 10MHz.
Parallax chose the 5MHz oscillator in FliP, so they can stay inside the data sheet specs.
Anything where that 0.07% precision is not good enough
eg UARTs are fine, and even USB is ok, most industrial control and servos will be fine, but TV colour burst would be out, High End Digital Audio is probably going to be sniffed at, and RF carrier work or Instrumentation / Timing would usually choose lower ppms.
The one I quoted was not 5000ppm, but you are right, you can buy lower spec resonators, and at lower prices.
eg
STCR5M00G53-R0 Murata 5.0000MHZ $0.19290/3k ±0.2% ±0.5% Built in Capacitor 15pF 60 Ohms -20°C ~ 80°C (4.50mm x 2.00mm)
Just to clarify for those that might find this confusing, the VCO in the PLL locks onto a frequency that's 16 times the crystal frequency. IOW, it always runs at the maximum frequency, regardless of the _clkmode setting. When this setting specifies, for example, pllx8, nothing internally is multiplied by 8 but, rather, the VCO frequency is divided by 2 to get the system clock frequency. Actually, the divider chain is always active down to divide-by-16. The pllx- just selects the tap.
-Phil
About 14.3MHz has been the experienced top range although Sapieha got the DIP40 to 15MHz reliably.