I have to agree that if Parallax sells things using PLL at 160 MHz, then the datasheet should say that is OK.
Equivalently, stating the contraposition of that assertion, "If the datasheet does not say it's okay, then Parallax should not sell things using PLL at 160 MHz."
Now suppose the Propeller was another company's product, and Parallax had no control over the datasheet. Would you then absolutely proscribe them from using a 10 MHz crystal, even though a significant experience factor showed that it was okay?
I have to agree that if Parallax sells things using PLL at 160 MHz, then the datasheet should say that is OK...
Or, the product should have a disclaimer...
Equivalently, stating the contraposition of that assertion, "If the datasheet does not say it's okay, then Parallax should not sell things using PLL at 160 MHz."
Now suppose the Propeller was another company's product, and Parallax had no control over the datasheet. Would you then absolutely proscribe them from using a 10 MHz crystal, even though a significant experience factor showed that it was okay?
-Phil
Sounds like the old debate on 4D's uOLED-96-Prop display, They used both 8MHz and 10MHz xtals. 8Mhz was first, 8Mhz x PLL16 (128MHz) was considered norm by 4D Systems, but when they switched to the 10 MHz XTAL, PLL x 8 (80 MHz) was advised IIRC.
Heck, sometimes, even following the data sheet can get you into trouble, but then that is clearly the vendor's fault.
What matters is whether something works or not -- not whose fault it is if it doesn't. But even more to the point, regardless of what the datasheet says, if it doesn't work, the onus is on the designer -- not the datasheet originator -- to fix the problem. In any event, an experienced designer should be able to spot errors in the datasheet -- at least the digital ones. Errors in analog characterization are more difficult -- not only to spot, but to prove once inferred.
The 4D thing is exactly what I'm thinking about right now...
I think they were an early adopter and had the idea that the Prop should work fine at 10 MHz * 16 = 160 Mhz.
And, it did really. But, if you tried to start a second cog, then it froze up...
I think 4D was a bit disgruntled by this actually...
But, it works just fine at 80 MHz anyway, even with the PLL at 160 MHz.
Sounds like the old debate on 4D's uOLED-96-Prop display, They used both 8MHz and 10MHz xtals. 8Mhz was first, 8Mhz x PLL16 (128MHz) was considered norm by 4D Systems, but when they switched to the 10 MHz XTAL, PLL x 8 (80 MHz) was advised IIRC.
The other thing to remember is that the datasheet may specify operating conditions for a wider or narrower range of temperature / voltage / etc. than the product being made using the part. It's the product's designer and manufacturer that are responsible for the product's behavior under narrower or wider conditions than the datasheet describes.
The other thing to remember is that the datasheet may specify operating conditions for a wider or narrower range of temperature / voltage / etc. than the product being made using the part.
Where a restriction is needed 'to meet a better spec', that is common to see that in Data sheets.
I have many Micro data sheets here, that specify Multiple MHz values :
Some give x1 and x2 Clock specs, and others elevate the MHz for a higher/tighter Vcc spread.
Others specify a de-rating for industrial (wider) temperature ranges.
I've also seen ripple specs for Vcc, on devices with VCOs.
It may be that Parallax needs to do some of that, in order to spec the device to guarantee the 10MHz / 160MHz VCO operation.
It might need 3.0V Vcc Min, and/or a lowering of the 125'C spec, or they may be happy to guarantee it at 2.7V and 55'C - 125'C
Notwithstanding specsmanship -- that discussion could go on forever -- let's get back to the OP's initial concern:
1. He was successfully using the Spin Stamp, which Parallax had ostensibly discontinued.
2. Ken stated that the reason for the product being in limbo was the crystal being discontinued by the manufacturer.
3. I stated that there was a suitable drop-in replacement for said crystal, and provided a link.
As far as I'm concerned, this issue has been resolved, and Parallax can continue manufacturing the Spin Stamp. All other discussion is superfluous and off-topic.
It may be that Parallax needs to do some of that, in order to spec the device to guarantee the 10MHz / 160MHz VCO operation.
It might need 3.0V Vcc Min, and/or a lowering of the 125'C spec, or they may be happy to guarantee it at 2.7V and 55'C - 125'C
I don't think anyone is advocating running the Prop at 160MHz. It has just been shown that production boards, (Hydra, Backpack, uOLED-96-PROP), all have run with a 10 MHz XTAL. But they all run at PLL x 8. Within spec.
Maybe Parallax could relax the XTAL frequency a bit to in include 10 MHz.
EDIT: Phil , didn't see your post until i posted mine. Your correct. Case Closed.
I don't think anyone is advocating running the Prop at 160MHz. It has just been shown that production boards, (Hydra, Backpack, uOLED-96-PROP), all have run with a 10 MHz XTAL. But they all run at PLL x 8. Within spec.
Maybe Parallax could relax the XTAL frequency a bit to in include 10 MHz.
Notwithstanding specsmanship -- that discussion could go on forever -- let's get back to the OP's initial concern:
1. He was successfully using the Spin Stamp, which Parallax had ostensibly discontinued.
2. Ken stated that the reason for the product being in limbo was the crystal being discontinued by the manufacturer.
3. I stated that there was a suitable drop-in replacement for said crystal, and provided a link.
As far as I'm concerned, this issue has been resolved, and Parallax can continue manufacturing the Spin Stamp. All other discussion is superfluous and off-topic.
-Phil
Now who is pontificating ? - you 'forgot' to mention
4) the OP is using the Spin Stamp in industrial applications, in good volumes
5) The Spin Stamp that Parallax actually ship, is operating outside the Guaranteed Data sheet values.
Only to the 'she'll be right' camp, can the combination of 4) and 5) be considered merely 'superfluous'
The issue is Parallax Specifications and design decisions. Simple, really. As is the solution.
Maybe it's just because I'm new and relatively inexperienced, but I don't feel comfortable using a chip outside the rated specifications as given in the datasheet. I'd prefer to use the 10MHz crystals due to their smaller size. How can I tell my employer that "in my experience", the 10MHz crystal works? I've never used one, and I have no idea if the modules that Parallax sells with the 10MHz crystal actually work as designed. It could be that nearly all fail, except for those on the west coast, and I just never hear about it.
In short, I don't have experience with the 10MHz crystal, and I can't point to a forum thread as 'my' experience and expect it to be good design.
Maybe this has already been pointed out, but all FORD wants is more Spin Stamps. Apparently he, his company, and his customers, have been happy with a 10 MHz resonator. I can't imagine FORD wanting anything different at this point. Certainly he doesn't want to have to support two versions of code for two different boards. I'm glad Phil has found a suitable replacement.
WRT bridges, I suspect there's something like a 2x safety factor designed in. I think that's all we are dealing with here: A 160 MHz VCO is gobbling up some of the the designed-in reserve.
What is a problem with Propeller working @ 10 MHz crystal? It works without problem with 14 MHz crystal, giving 112 MHz CPU speed. PLL have to work @ 224 MHz then and if it is capable to work at this speed, then what is 160 MHz?
Specifications have to be changed.
Today's Propellers all works fine @ 100 MHz and this should be considered its standard clock.
PLL - time to change this 64..128 MHz range. I experimented with VGA drivers. The picture quality was bad if I used PLL at ~64 MHz. To get 40 MHz pixel clock result was better when using pll @ 160MHz than @ 80 MHz. Its maximum frequency is something about 230 MHz. Then I think 200 MHz is safe, and 160 MHz is very safe - 2/3 of maximumm available.
Today's Propellers all works fine @ 100 MHz and this should be considered its standard clock.
If Parallax does decide to make 100MHz the standard clock, they should let the marking people in on the release. "Can't wait for Propeller 2? We now offer...(drum roll) Propeller 1.5! Your favorite 8 core microcontroller, now 25% faster! Get things done, faster!"
I think the issue with finding a crystal for the current Spin Stamp layout will be resolved fairly easily and will enable Parallax to get it back on the shelves. As for the frequency of said crystal, I am in agreement with jmg in regards to the design of the Spin Stamp being outside the parameters defined in the Propeller manual and datasheet. It is very clearly stated "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." One of the key purposes of a datasheet is to define the functional parameters for proper utilization of the component. While I agree with Phil in that there is a considerate amount of usage supporting stable operation of the Propeller with a 10mhz crystal, I feel that jmg is correct in abiding by the datasheet for his designs to be reliable.
Perhaps knowing why Parallax established the 64-128MHZ window onto the PLL would explain what the true risks are of operating outside those specifications.
And it is working at 14 MHz. Without any voltage change.
75% safety margin. Too high, I think. Today made CPUs cannot work even at 1.33 of its standard clock if you don't set a higer voltage.
The Propeller production started in 2006, 6 years ago and I think this is the reason - production quality increased in these 6 years, but specifications was not changed. It is time to Parallax to retest their chips and set a new standards according to 2012 reality.
100 MHz clock standard
112 MHz - recommended overclocking for multimedia and other non critical purposes
Safe PLL - 80..180 MHz. Overclocked PLL - up to 220 MHz
These are results of my experiments with Propeller overclocking. Max reached was 118 MHz (7.3xxxx *16) main clock, 14 MHz crystal and 230 MHz PLL frequency
From 4 to 8 MHz...
And it is working at 14 MHz. Without any voltage change.
What Vcc was that, and how does it very with Vcc ?
The variables are usually Process, Temperature and Voltage.
eg Try it at 3.0V and 85'C, and you get an idea of the remaining typical process adder.
Where I've pushed vendors in the past to expand their datasheets, they usually go into a huddle, and then come back with a reduced Voltage Spread, and sometimes, a reduced temperature spread, on the new Data Sheet.
They know their silicon, & test flow yields, best.
The Prop has room to specify +/- 10% in Vcc, instead of the present -18%, and to relax the 125'C (!) too, if needed.
105'C and 85'C are common alternatives.
Jauch offers a 10MHz part even in an even smaller 3.2 x 2.6 mm package, the JXS32 series. Also they have the 5 x 3.2 with 4 terminations, JXG53P4. It comes standard at 12pF, but they can supply it for 8-30 pF. (The Prop needs 20pF). I'm using their 7x5 part on my boards at 5MHz, and it is hard to source 5MHz in that size -- Transko couldn't do it.
Here's an idea for an easy way for Parallax to clarify stability of overclocking the PLL as defined in the datasheet.... Add another application note that covers any proven method of using crystals that are outside the specs in the datasheet. By proven, I mean any method that Parallax will agree to being a method that still operates the propeller in a stable state.
Here's an idea for an easy way for Parallax to clarify stability of overclocking the PLL as defined in the datasheet.... Add another application note that covers any proven method of using crystals that are outside the specs in the datasheet. By proven, I mean any method that Parallax will agree to being a method that still operates the propeller in a stable state.
The easiest way is to fix the data sheet.
Others have done this type of change, it is not hard, just a few more lines, and qualifiers.
Most of the work is in deciding exactly where to peg things, and that is what customers need to know.
Comments
Now suppose the Propeller was another company's product, and Parallax had no control over the datasheet. Would you then absolutely proscribe them from using a 10 MHz crystal, even though a significant experience factor showed that it was okay?
-Phil
Nope, experience says : if you gamble to ignore the data sheet, you are on thin ice.
Heck, sometimes, even following the data sheet can get you into trouble, but then that is clearly the vendor's fault.
Correct.
Sounds like the old debate on 4D's uOLED-96-Prop display, They used both 8MHz and 10MHz xtals. 8Mhz was first, 8Mhz x PLL16 (128MHz) was considered norm by 4D Systems, but when they switched to the 10 MHz XTAL, PLL x 8 (80 MHz) was advised IIRC.
http://forums.parallax.com/showthread.php?97730-Tiny-new-OLED-display-from-4D-Systems-WITH-A-TWIST!&p=679308#post679308
-Phil
Only if you ignore production variations, temperature variations, and drift over time...
I am less interested in 'if it works, ship' than in ensuring a design will continue to work, in the next manufacturing run, and the one after that.
Perhaps we have different customers, with mine, folklore simply does not cut it.
What you or I say, does not really matter : It is Parallax who are shipping a product, outside what their own Data Sheet says is OK.
If, as you claim, that really has zero field/process issues and is 100% reliable, then they should simply fix the data sheet to say so.
Problem solved. Easy.
I think they were an early adopter and had the idea that the Prop should work fine at 10 MHz * 16 = 160 Mhz.
And, it did really. But, if you tried to start a second cog, then it froze up...
I think 4D was a bit disgruntled by this actually...
But, it works just fine at 80 MHz anyway, even with the PLL at 160 MHz.
Where a restriction is needed 'to meet a better spec', that is common to see that in Data sheets.
I have many Micro data sheets here, that specify Multiple MHz values :
Some give x1 and x2 Clock specs, and others elevate the MHz for a higher/tighter Vcc spread.
Others specify a de-rating for industrial (wider) temperature ranges.
I've also seen ripple specs for Vcc, on devices with VCOs.
It may be that Parallax needs to do some of that, in order to spec the device to guarantee the 10MHz / 160MHz VCO operation.
It might need 3.0V Vcc Min, and/or a lowering of the 125'C spec, or they may be happy to guarantee it at 2.7V and 55'C - 125'C
2. Ken stated that the reason for the product being in limbo was the crystal being discontinued by the manufacturer.
3. I stated that there was a suitable drop-in replacement for said crystal, and provided a link.
As far as I'm concerned, this issue has been resolved, and Parallax can continue manufacturing the Spin Stamp. All other discussion is superfluous and off-topic.
-Phil
I don't think anyone is advocating running the Prop at 160MHz. It has just been shown that production boards, (Hydra, Backpack, uOLED-96-PROP), all have run with a 10 MHz XTAL. But they all run at PLL x 8. Within spec.
Maybe Parallax could relax the XTAL frequency a bit to in include 10 MHz.
EDIT: Phil , didn't see your post until i posted mine. Your correct. Case Closed.
Yes, that is the correct and expected action.
Now who is pontificating ?
4) the OP is using the Spin Stamp in industrial applications, in good volumes
5) The Spin Stamp that Parallax actually ship, is operating outside the Guaranteed Data sheet values.
Only to the 'she'll be right' camp, can the combination of 4) and 5) be considered merely 'superfluous'
The issue is Parallax Specifications and design decisions. Simple, really. As is the solution.
-Phil
Only to you, perhaps. Meanwhile, a problem remains unresolved.
In short, I don't have experience with the 10MHz crystal, and I can't point to a forum thread as 'my' experience and expect it to be good design.
-Phil
WRT bridges, I suspect there's something like a 2x safety factor designed in. I think that's all we are dealing with here: A 160 MHz VCO is gobbling up some of the the designed-in reserve.
Based upon what I know at this point, you're probably safe to continue working on your project. The Spin Stamp ain't dead yet.
-Phil
Specifications have to be changed.
Today's Propellers all works fine @ 100 MHz and this should be considered its standard clock.
PLL - time to change this 64..128 MHz range. I experimented with VGA drivers. The picture quality was bad if I used PLL at ~64 MHz. To get 40 MHz pixel clock result was better when using pll @ 160MHz than @ 80 MHz. Its maximum frequency is something about 230 MHz. Then I think 200 MHz is safe, and 160 MHz is very safe - 2/3 of maximumm available.
Thats good. I hate running a train of thought for a week and then having to stop it and run something else.....
If Parallax does decide to make 100MHz the standard clock, they should let the marking people in on the release. "Can't wait for Propeller 2? We now offer...(drum roll) Propeller 1.5! Your favorite 8 core microcontroller, now 25% faster! Get things done, faster!"
Perhaps knowing why Parallax established the 64-128MHZ window onto the PLL would explain what the true risks are of operating outside those specifications.
And it is working at 14 MHz. Without any voltage change.
75% safety margin. Too high, I think. Today made CPUs cannot work even at 1.33 of its standard clock if you don't set a higer voltage.
The Propeller production started in 2006, 6 years ago and I think this is the reason - production quality increased in these 6 years, but specifications was not changed. It is time to Parallax to retest their chips and set a new standards according to 2012 reality.
100 MHz clock standard
112 MHz - recommended overclocking for multimedia and other non critical purposes
Safe PLL - 80..180 MHz. Overclocked PLL - up to 220 MHz
These are results of my experiments with Propeller overclocking. Max reached was 118 MHz (7.3xxxx *16) main clock, 14 MHz crystal and 230 MHz PLL frequency
What Vcc was that, and how does it very with Vcc ?
The variables are usually Process, Temperature and Voltage.
eg Try it at 3.0V and 85'C, and you get an idea of the remaining typical process adder.
Where I've pushed vendors in the past to expand their datasheets, they usually go into a huddle, and then come back with a reduced Voltage Spread, and sometimes, a reduced temperature spread, on the new Data Sheet.
They know their silicon, & test flow yields, best.
The Prop has room to specify +/- 10% in Vcc, instead of the present -18%, and to relax the 125'C (!) too, if needed.
105'C and 85'C are common alternatives.
The easiest way is to fix the data sheet.
Others have done this type of change, it is not hard, just a few more lines, and qualifiers.
Most of the work is in deciding exactly where to peg things, and that is what customers need to know.