Do I need the more expensive Parallax crystals for the Propeller?

cbmeeks

While I'm a huge fan of sending Parallax money and supporting the company, I need to buy a few dozen crystals for my various Propeller 1 designs.

I'm mostly interested in the 6.25 MHz crystals but I assume the same logic would work for 5.00 and 6.00 MHz as well (I mostly use Propellers for video and audio).

Anyway, Mouser has the following crystals:

CTS Version - 10 for $0.55/ea

and

Parallax Version - 10 for $1.35/ea

The specs look pretty close...well, close enough for my very amateur understanding.

Would you recommend one over the other? Should I use the Parallax version in production runs and cheaper CTS versions for prototyping?

I can't ever imagine me producing more than a few dozen circuits so I'm not talking mass quantities here. But still, every little bit helps....especially since I'm trying to recover from Christmas expenses. lol

Thanks.

jmg

Crystals are normally quite forgiving.
One spec to check is ESR, and if you want top precision, then the ppm numbers matter too.
The ESR is more an issue when going very small in size.

You can check how much margin you have, by adding a series resistor to the crystal. Above some value, the oscillator will fail to start reliably.

Seems 6.25MHz does not give a lot of choices - the Parallax one is lower profile.

Phil Pilgrim (PhiPi)

The Mouser crystal has the wrong capacitance spec to be used with the Prop. Look for a crystal that specs around 18-20 pF. The Mouser crystal will probably work okay -- just at the wrong frequency.

-Phil

cbmeeks

You know, that might explain some weirdness I've had with a few self-made Propeller circuits I've built. Where even simple LED blinking never seemed stable unless I didn't use a crystal at all and just used the internal oscillator.

I think I will stick with the better parts. Thanks for the info!

jmg

cbmeeks wrote: »

You know, that might explain some weirdness I've had with a few self-made Propeller circuits I've built. Where even simple LED blinking never seemed stable unless I didn't use a crystal at all and just used the internal oscillator.

That's unlikely to be CL issue, but more Decoupling or PLL problems.
CL simply shifts the operating point slightly.
Here is one representative example of ppm vs CL for Crystal:
(in this case ~+50% on CL moves ~50ppm down in f)

Peter Jakacki

Yes, your problems probably have more to do with layout or breadboarding and/or decoupling etc. Keep the crystal close to the chip and avoid any unnecessary track lengths such as you have when you breadboard. Decoupling caps (typically 0.1uf ceramic) should be close to the chip as well, and by close this also includes very short connections to the chip, not a scenic tour along the outer rim. The rest is very forgiving.

BTW, are you serious about the 8 bucks you could save by using full height crystals???

If you use TH crystals consider using the cylindrical types like this 6MHz one for 61 cents 10 off. What I like about them is besides the size and price is the fact that the pins are close together and so they can connect intimately with the Prop pins.

cbmeeks

Peter Jakacki wrote: »

BTW, are you serious about the 8 bucks you could save by using full height crystals???

If you use TH crystals consider using the cylindrical types like this 6MHz one for 61 cents 10 off. What I like about them is besides the size and price is the fact that the pins are close together and so they can connect intimately with the Prop pins.

Great recommendation. I didn't see those when I searched.

Thanks.

jmg

Peter Jakacki wrote: »

If you use TH crystals consider using the cylindrical types like this 6MHz one for 61 cents 10 off. What I like about them is besides the size and price is the fact that the pins are close together and so they can connect intimately with the Prop pins.

Those are cute - higher ESR, but still likely tolerable, and the Temperature curve is not bad, considering the package.
I see also 732-CA3016.000M-CPB
with better ppm

DigitalBob

I always buy them from digikey and never had any problems. I think I using the X1050-ND ecs-40-20

cbmeeks

I bought the 732-CA3016.000M-CPB and a few other things. :-)

Martin Hodge

I've used hundreds of these and never had a problem.


--Edit--

whoops, didn't catch the 6.25MHz part. Nevermind.

gis667en11

Say I wanted to run a 100MHz prop. Instead of buying an expensive 6.25MHz crystal, would would be the issue with getting a commonly found and lower cost 25MHz crystal and doing PLLx4?

Publison

Nope, 10Mhz input is about the max, although this has been pushed a little with the very correct board layout.

cbmeeks

So, what would happen if I used a 10 MHz crystal and some dummy (like me) set the PLLx16? How long until smoke comes out? lol

I *REALLY* wished we could use a 14.31818 MHz (4xNTSC) crystal. Not only would it provide 229 MHz (57 MIPS) but it would have less jitter with NTSC timings and the crystals are wildly available and cheap. :-D

Peter Jakacki

I use 10MHz all the time but the Prop won't run at all at PLLx16. I use Phil Pilgrim's clkset routine to automatically determine if the crystal is 5 or 10MHz. Of course my crystals are always 10MHz but the autoset feature is handy for all those 5MHz boards.

jmg

gis667en11 wrote: »

Say I wanted to run a 100MHz prop. Instead of buying an expensive 6.25MHz crystal, would would be the issue with getting a commonly found and lower cost 25MHz crystal and doing PLLx4?

The Xtal spec is given as 4~8MHz, and the key issue is there is no (Xtal * 4 PLL) mode, IIRC the PLL-VCO is hardwired to always be 16x Xtal, but you can divide the PLL out to the final SysCLK.
It is the VCO range that ultimately limits the crystal.

P2 will fix this oversight, P1 was designed so long ago, that smaller package crystals were not on the radar.

You can use 74HC6323 (Osc + Divider in SOP8) or a Si5351A to get into that valid range from a higher crystal.
Si5351A needs programming, but is small & highly flexible, and 74HC6323 is not as cheap or as small as it should be...

Cluso99

If you look at the xtal setting in the manual it has settings for higher xtals.

Phil & Peter,
Can you explain how the xtal detection works. I tried Phils code with various xtals 5, 6, 6.5, 7.3, 12, 13, 14MHz and I could get an average count returned (varies +/- 1) when using pllx16 for lower range xtals and pllx8 for upper range. I wondered if this varied between prop dice.
I know it is to do with the poll in the counters for video that is being tested for lock and it appears to loose lock the higher the base clock (xtal X pll setting)

jmg

Cluso99 wrote: »

If you look at the xtal setting in the manual it has settings for higher xtals.

Err, are you sure ?

["The PLL internally multiplies the XIN pin frequency by 16. OSCENA must be ‘1’ to propagate the XIN signal to the PLL. The PLL’s (VCO) internal frequency must be kept within 64 MHz to 128 MHz – this translates to an XIN frequency range of 4 MHz to 8 MHz."]

The other choices are post VCO dividers, of /1,/2,/4,/8,/16, but the manual is poorly worded there.
The counter terminology is clearer, with VCO/128 to VCO/1

Phil Pilgrim (PhiPi)

Cluso99 wrote:

If you look at the xtal setting in the manual it has settings for higher xtals.

True, but the higher frequency crystals cannot be used with the PLL.

Can you explain how the xtal detection works. I tried Phils code with various xtals 5, 6, 6.5, 7.3, 12, 13, 14MHz and I could get an average count returned (varies +/- 1) when using pllx16 for lower range xtals and pllx8 for upper range. I wondered if this varied between prop dice.
I know it is to do with the poll in the counters for video that is being tested for lock and it appears to loose lock the higher the base clock (xtal X pll setting)

Basically, it attempts to overdrive one of the counter's PLLs if the system clock frequency is too high (i.e. 10 MHz crystal at PLLx16). It uses the video generator to monitor the counter PLL and compares that with the system clock. If there's a mismatch, it knows that the frequency is too high, and sets the PLL to x8.

-Phil

Cluso99

Phil Pilgrim (PhiPi) wrote: »

Cluso99 wrote:

If you look at the xtal setting in the manual it has settings for higher xtals.

True, but the higher frequency crystals cannot be used with the PLL.

Below is the part of the manual which contradicts what we thought...

Can you explain how the xtal detection works. I tried Phils code with various xtals 5, 6, 6.5, 7.3, 12, 13, 14MHz and I could get an average count returned (varies +/- 1) when using pllx16 for lower range xtals and pllx8 for upper range. I wondered if this varied between prop dice.
I know it is to do with the poll in the counters for video that is being tested for lock and it appears to loose lock the higher the base clock (xtal X pll setting)

Basically, it attempts to overdrive one of the counter's PLLs if the system clock frequency is too high (i.e. 10 MHz crystal at PLLx16). It uses the video generator to monitor the counter PLL and compares that with the system clock. If there's a mismatch, it knows that the frequency is too high, and sets the PLL to x8.

-Phil

This is interesting because I can get these results from the counters...

-------    PLLx8 ---- PLLx16
5.00MHz    64         88-90
6.00MHz    64         106-107
6.50MHz    64         115-116
7.37MHz    ?          eratic!
12.0MHz    113-114
13.5MHz    119-121
14.0MHz    124

So this looks like I may be able to determine the xtal frequency.

Phil Pilgrim (PhiPi)

Cluso99 wrote:

Below is the part of the manual which contradicts what we thought...

I don't see the contradiction. There's nothing there to suggest that a 60 MHz crystal will work with the PLL.

-Phil

Cluso99

It says you can use a crystal or resonator using
XTAL1 for 4-16MHz
XTAL2 for 8-32MHz
XTAL3 for 20-60MHz

Alternately you can use
XINPUT for DC-80MHz oscillator input.

My understanding is that when a XTAL is used, then the PLL will automatically multiply by 16, and the PLLxX setting will divide up the result by 1 for PLLx16, 2 for PLLx8, 4 for PLLx4, 8 for PLLx2, 16 for PLLx1. Is this different to your understanding?

When an Oscillator is used, the PLLxX setting is not available, and the Osc frequency is used without the x16 multiplier.

Now, I recall that Chip said the PLL tops out at just above 14.3MHz x16 = 228.8MHz.
Sapieha and I found that a 15MHz xtal would work on the DIP prop, but only about 14.318MHz would work on the QFP prop, both being used with PLLx16.

jmg

Cluso99 wrote: »

My understanding is that when a XTAL is used, then the PLL will automatically multiply by 16, and the PLLxX setting will divide up the result by 1 for PLLx16, 2 for PLLx8, 4 for PLLx4, 8 for PLLx2, 16 for PLLx1..

Yes, it is always x16, followed by a Divider,
The manual poorly words, this, it should use VCO/1, VCO/2, VCO/4, VCO/8, VCO/16, instead of XIN x 8

Because the divider comes after the VCO, the VCO spec trumps all others in determining the valid crystal range.

{"The PLL internally multiplies the XIN pin frequency by 16. OSCENA must be ‘1’ to propagate the XIN signal to the PLL. The PLL’s (VCO) internal frequency must be kept within 64 MHz to 128 MHz – this translates to an XIN frequency range of 4 MHz to 8 MHz."}

What users then want to over-clock by, outside the spec, is up to them.