I've been doing a lot of reading on clocks & jitter lately, particularly the excellent Well Tempered Master Clock thread, but have some questions for which I have not been able to find answers.
The first is how to select a decent crystal, and whether this is possible from datasheets alone or only though testing. I am talking from the point of view of identifying a crystal with respectable performance from the myriad available through Mouser/Digikey, not to target the extreme state of the art performance which can only be achieved with custom crystals, but to attain decent performance in a design where extreme cost or custom parts are not justifiable. Throughout the TWTMC thread the importance of a good quality crystal is stressed, and that particular project uses custom polished crystals from Laptech. Andrea Mori also mentions TFC crystals as good quality for a lower cost.
But looking at datasheets for those parts, what actually tells you they are good crystals vs. run of the mill crystals?
Q seems to be a big factor (high Q results in lower close-in phase noise).
The Laptech and TFC datasheets are pretty much the only ones that I have seen with a specification for Q, so I suppose this indicates they may be better quality than other crystals but is there any other information to base the decision on?
From what I have read the only way to know the Q is by measurement (which I cannot do), but ESR is linked to Q so it seems that choosing a crystal with low ESR is a route to follow.
This MtronPTI document mentions that round quartz blanks will have a lower ESR than small strips, essentially the larger the blank the better, so it seems choosing crystal in a larger package will be beneficial.
In the TWTMC thread Andrea also mentions a preference for cold-welded packages, but I cannot find any explanation of this. What is the advantage of a cold welded package over a resistance welded package?
My final question is on buffering/fan-out/distribution. NB3L553 gets a lot of recommendations and appears to perform well based on the datasheet, but many of the high quality clock designs presented on this forum just use the 74HCU04.
What are the pros and cons of each?
The basic points I can see between the chips are:
- NB3L553 is higher cost
- 74HCU04 output-to-output skew is not directly defined. NB3L553 spec'd at 50ps across its temperature range
- NB3L553 has higher output current (80mA vs 25mA)
- NB3L553 provides a phase-noise plot in the datasheet.
It seems that except for cost, the NB3L553 hold all the advantages. If this is the case why is it not used in the cost-no-object clock designs? Is it simply the case that the 74HCU04 already offers excellent performance which matches the performance of the oscillators?
Thanks
The first is how to select a decent crystal, and whether this is possible from datasheets alone or only though testing. I am talking from the point of view of identifying a crystal with respectable performance from the myriad available through Mouser/Digikey, not to target the extreme state of the art performance which can only be achieved with custom crystals, but to attain decent performance in a design where extreme cost or custom parts are not justifiable. Throughout the TWTMC thread the importance of a good quality crystal is stressed, and that particular project uses custom polished crystals from Laptech. Andrea Mori also mentions TFC crystals as good quality for a lower cost.
But looking at datasheets for those parts, what actually tells you they are good crystals vs. run of the mill crystals?
Q seems to be a big factor (high Q results in lower close-in phase noise).
The Laptech and TFC datasheets are pretty much the only ones that I have seen with a specification for Q, so I suppose this indicates they may be better quality than other crystals but is there any other information to base the decision on?
From what I have read the only way to know the Q is by measurement (which I cannot do), but ESR is linked to Q so it seems that choosing a crystal with low ESR is a route to follow.
This MtronPTI document mentions that round quartz blanks will have a lower ESR than small strips, essentially the larger the blank the better, so it seems choosing crystal in a larger package will be beneficial.
In the TWTMC thread Andrea also mentions a preference for cold-welded packages, but I cannot find any explanation of this. What is the advantage of a cold welded package over a resistance welded package?
My final question is on buffering/fan-out/distribution. NB3L553 gets a lot of recommendations and appears to perform well based on the datasheet, but many of the high quality clock designs presented on this forum just use the 74HCU04.
What are the pros and cons of each?
The basic points I can see between the chips are:
- NB3L553 is higher cost
- 74HCU04 output-to-output skew is not directly defined. NB3L553 spec'd at 50ps across its temperature range
- NB3L553 has higher output current (80mA vs 25mA)
- NB3L553 provides a phase-noise plot in the datasheet.
It seems that except for cost, the NB3L553 hold all the advantages. If this is the case why is it not used in the cost-no-object clock designs? Is it simply the case that the 74HCU04 already offers excellent performance which matches the performance of the oscillators?
Thanks
Best crystals are SC cut, details of cuts at: Quartz crystals.
NB3L553 can be exceptionally good, but to perform at its best it needs careful layout, optimized power, and carefully considered decoupling. Most of the implementations I have seen around here could IMHO benefit substantially from further refinement.
NB3L553 can be exceptionally good, but to perform at its best it needs careful layout, optimized power, and carefully considered decoupling. Most of the implementations I have seen around here could IMHO benefit substantially from further refinement.
Thanks,
So two questions:
1. How to tell a poor quality SC cut crystal from a good quality one? (based on datasheet)
2. Is an NB3L553 in optimum conditions any better than 74HCU04 in optimum conditions?
So two questions:
1. How to tell a poor quality SC cut crystal from a good quality one? (based on datasheet)
2. Is an NB3L553 in optimum conditions any better than 74HCU04 in optimum conditions?
Q is approximately the reactance of the motional capacitance divided by the motional resistance (which is approximately the ESR), so if you can find data for both, you can calculate Q. For AT-cut crystals, the motional capacitance C1 is usually 200 to 500 times smaller than the electrode capacitance C0.
Unfortunately, many crystal datasheets have only nonsensical information about C0 or even none at all. For example, SMD crystals normally have a C0 between 0.5 pF and 2 pF, but the datasheets often only state that it is less than 7 pF.
Unfortunately, many crystal datasheets have only nonsensical information about C0 or even none at all. For example, SMD crystals normally have a C0 between 0.5 pF and 2 pF, but the datasheets often only state that it is less than 7 pF.
Yeah thats exactly the problem, most datasheets don't seem to give any of the key info that is needed to judge quality!
Perhaps the conclusion is that they're all the same (excluding the custom polished crystals of course)
Perhaps the conclusion is that they're all the same (excluding the custom polished crystals of course)
That the two components are not equivalent is obvious if you read the datasheets.
The 74HCU04 is an unbuffered inverter. It is a required component in many oscillator circuits. There is no skew specified because there is no skew.
The NB3L553 is a 1:4 clock fanout buffer. It is used to increase an oscillator's fanout at the expense of added noise, jitter, and propagation delay.