Is there anything wrong with these oscillators?

[Signal]

This is the survey! I hope the provocative one
Not an actual question - rather occasion for fun (i.e. academic interest).

The same PIC32MX795 TQFP64 on both pictures.
Quartz resonators are 16 MHz in similar packages: YIC XT224 and ILSI ILCX18.
Resistor and capacitors are 0805. Two layers.

Layout 1:


Layout 2:

[Old_but_Alive]

I think both layouts are OK

I prefer the second photo.

The length of the track from the resistor is less

as to why one has 1.2K, and the other 3K9 ohms I dont know.

its probably a series resistor from the output of the oscillator to limit the drive to the xtal.

overdriving xtal causes problems.

I usually use 200ohm though.

[mars01]

1. I assume that you have a ground plane/polygon on the bottom layer. I would not connect the crystal capacitors with a trace on the top layer but put the GND vias very close to each capacitor.
2. The series resistor might be not necessary or might need some tweaking the value using an o-scope so you have a clean sinus wave.
3. The biggest issue is that I think that you connected the crystals in a wrong way meaning you connected the NC pins instead of actual active pins.

[Signal]

as to why one has 1.2K, and the other 3K9 ohms I dont know.

120 Ohm actually.

[vern]

since the trace with the series resistor (PIN40) is the oscillator output, I would try to make this trace as short as possible for EMC reasons.
In the lower picture you could rotate the entire block of components by 90 degrees left, than you get a short oscillator output trace and a somewhat longer osc input trace, which is better for EMC.

[Signal]

I would like not to unveil the intrigue (if it ever was here) prematurely. Maybe here will be some more opinions. So I will add my responses on each point later. I already wrote down some of them to be prepared for future posting.

If I think that I see or know something then that doesn't mean I do not expect that many know more or better than me.
These are my designs. Earlier I thought they are fine. Then found an issue. Then found another. It costs me some time of study. I'm going to share results of this work. And something could still be missed. That is the reason for this thread.

Now I offer an experiment.
I have 1 invitation credit. If you are looking for sonsivri membership and have something to say about certain (not common phrases!) issue(s) of discussed oscillators then you can publish your opinion in NOTE area of invitation request (after your self-presentation part for example - it is a free form). Mark your participation in this experiment by adding suffix _OSC to your nick, it can be cleaned later.

I will check invitation board for Usernick_OSC requests untill special notification about the end of this experiment appears here in this thread. If I meet an explanation of an issue that has not less importance than the most important issue that I have in mind (or exactly the same issue) I will invite you. Limitation: do not repeat already posted variants - only first appearance has a value for experiment. (That limitation is not for members - of course you are free and appreciated to share your agreement.)

You can think that it will be very subjective decision. Yes, of course. After all I could be wrong. But asking for membership you perhaps want to participate. And my offer is the fun opportunity to do it.

[frnando]

Not sure what kind of problems this layouts could cause, but if I'm doing such layout, I would use a shield of GND around the oscillator components coming DIRECTLY from the neighbor GND pin (as I think thats why it is so close).
I suspect that other thing that could cause problems on high impedance lines, is the oscillator OUTPUT line running so close to I/O pins...
That's what my limited knowledge tells me.

[Old_but_Alive]

what is this all about.

people have been giving you their technical views.

this is not a game

[Signal]

what is this all about.

This is about common engineering practice in electronics - making damn simple quartz oscillator. That short task in words shows itself by dozen of variants in practice, part of them quit contradicting that corresponds usually to conflicting "official" recommendations often given without enough technical arguments.
There are many factors remaining unclear for most and perhaps for not the worst ones. For example read errata for new PIC32MZ processors. Think about how many decades Microchip makes controllers with integrated oscillator. Then what the hell is with oscillators in new flagchips that can not be corrected for several chip revisions? Corrected in one series - reappears in next. So Quartz oscillator is a very difficult task for XXI century

After all this is about our personal practice and experience. That is all that forums exist for.

people have been giving you their technical views.

And I need it. That is what I asked for. I want to know as many opinions and arguments as I can collect about task I interested in. Don't you need the same? That is why I started this thread.

this is not a game

I see a negative sound. But with all respect I do not know what was done inappropriately nor do not think that this is a game nor do not think that a game is a bad thing.
We are engineers because we like technical stuff - tasks and solutions. And we value a beauty of them. If I like chess - I like to experience just another great chess puzzle. And it will not be about game it will be about beauty of beloved things.

By first post in this thread I tried to define it precisely - academic interest. I posed a technical problem to test it, to find new points without influence of my own solution that also will be shared later. And I stated that my own experience with this task tells me that already found problem could be not last.
You have experience to share - great, many thanks. You have arguments - even better.

Or maybe you think that the experiment I started is a game. Yes the game as our life is. But not a game at the same time - all serious and fair. I call new participants because I very interested in professionals that knows particular things better then I do. I do not know if the problem bigger than I already found exists, so the same level will trigger my invitation. My initiative may look strange but I see no problems or disrespect.

I tend to be too verbose. So Old_but_Alive, be more definite, please - what is the problem if it still is?

[Old_but_Alive]

@signal

I found your reply #5 very very strange.

I wont reply to this thread again.

[Vineyards]

As a non-engineer who plunged into this childhood hobby out of interest and a series of coincidences, I tried to ask similar questions in the past and I was confronted with a similar attitude. For me, it is OK to come up with an interesting question and to present it as a puzzle as long as there is enough common interest in it. If you don't find it interesting just skip it and focus on another post. That is exactly what some users are doing at the moment.

[rtm]

Both are bad.

1. You need to locate quartz as close as possible to MC oscillator pins and connect it with shortest path to them. So, your second board is better, but you need to move right cap up, and move quartz and resistor to right.

2. Capacitors must be located in symmetry. So, they must be located symmetrically above quartz in your second board.

3. You need to make a separated ground plane under oscillator, it must be separated from main ground plane and be connected to the nearest ground pin of microcontroller (in your case it is pin 41). I think that you are wrongly using main ground plain here.

4. You need to make guard ring around oscillator (in upper layer) connected in few points to your separated ground plane and directly to pin 41. You have such unfinished guard ring in first board but it was not connected directly to pin 41.

[Vineyards]

There is always a trick to get better accuracy on an oscillator but the question is how much better and how necessary would that be. For majority of cases, the standard accuracy is good enough. We don't know what kind of an application the circuit will be used for. However, we know that there is a long and narrow path towards better accuracy that starts with pcb design, pcb production, choice of crystal oscillator and goes all the way up to the use of OCXO's. It all depends on what you will use this circuit for. In some situations, even the use of an XTAL is quite redundant. Things like guard rings or shields are more necessary for high z circuits but may be used to eliminate parasitic capacitance caused by pcb leakage.

[bigtoy]

rtm makes some good points. These 2 boards both look like they're "height constrained". In both cases the circuits drag out to the left, resulting in long trace lengths. No doubt they still work, but they're not optimal. More common (when possible) is to place the resonator or crystal right at the MCU pins, then have the caps directly above, with vias straight down into a ground-plane below. Doing that keeps a tight layout with very short trace lengths.

[roscoe]

I think for a 16MHz oscillator, with the frequency tolerances that make sense for a part like a PIC32, pretty much any reasonable layout will be fine.  This is not something like a TCXO design where you really care intensely about noise etc. 

[Signal]

size

[vern]

#2 has no ground connection for the 2 caps, the vias look like they are not metallized.
#1 I'm not sure but here the vias look metallized, there might also be a ground connection on the top left under some part or screw?

[Signal]

#2 has no ground connection for the 2 caps, the vias look like they are not metallized.
#1 I'm not sure but here the vias look metallized, there might also be a ground connection on the top left under some part or screw?

All vias are ok. Both are "working" oscillators in fact. My previous message was the answer on my own question "what is wrong?". The size is wrong.

[bobcat1]

One question - are you plan to manufacture your design by the +100,000 unit?
Whether the answer is yes - I rest my case , but whether the answer is no :

Why do you use Quartz resonators when Quartz oscillator are available from the shelf for very low cost
using Quartz oscillator is far more simple (1 decoupling cap is needed) no critical components placement, accurate trimming, low cost.

All the best

Bobi       

[Signal]

Quartz oscillator is far more simple (1 decoupling cap is needed) no critical components placement, accurate trimming, low cost.

I thought about it long time ago. Many doubts. Simplicity is questionable. You need additional power trace. In case device sleeps this line should be controlled too - so two uC pins anyway. Operational current is comparable with consumption of the whole device without oscillator. Sharp edges on several MHz - additional EMI considerations (maybe I am wrong). So my ancient conclusion - not for every project.

<...>are available from the shelf for very low cost

I should check if they are actually very low priced now. Thank you for pointing.

The size is wrong.

By the way, size reflects power, max power determines max Vp-p, that could be not enough for work or reliable work of given generator in uC, that usually depends on supply voltage.

[Catcatcat]

All schemes will work. A series resistor is needed only in certain specific cases. I recommend using ceramic resonators with built-in capacitors, they save space. In PIC, the generators are so reliable that they will run any resonator with any strapping. I almost never use resonators in my projects, the reliability and stability of the internal generator completely suits me.