PAW1 - NU-Xtal

[matseng]

As mentioned in an earlier post I'll try to do one small PCB project every week and then send for PCB's of it every Monday. This it the first project.

NU-Xtal
The NU -part stands for NanoUtils which is a series of really tiny breakout boards that I'll do over time.

This board is a breakout containing a SMD crystal, two caps and a resistor meant to be used in breadboards when building something with an Atmel ATmega or a Microchip PIC that have the two crystal pins next to a GND pin. For instance the atmega328 or pic18f2550.

Eagleup rendering of NU-Xtal

As simple as can be...

NU-Xtal-Schematic.png (9.87 KiB) Viewed 2690 times

Design files and stuff can be found at https://github.com/SmallRoomLabs/NanoUtils.

The board can also be populated with a two pin header at the bottom and then use a short wire from the top of the board to GND if the GND is not adjacent to the xtal pins in your design.

The crystal Digikey Link is a 5x3.2 mm SMD that can be found every now and then for a rather low price - not as low as a standard through-hole HC49 crystal but you have to pay a little extra for the small size. A HC49 is huge in comparison.

Since the board is only 8 x 11 mm I did a step & repeat with 24 copies on a 5x5 cm pcb. Between each copy I have a 0.25mm (~ 10 mil) of tStop & Bstop so I will get a slight groove in the soldermask for easier cutting with a blade. I'll order them as 0.8 mm pcb instead of the standard 1.6mm thickness to make them easier to cut as well.

A 5x5 cm PCB with 24 copies of the NU-Xtal

[Sjaak]

For small boards <5x5cm I usually select .8mm. IMHO 1.6mm is too thick for small board, also when using light components (smd) I don't think it is very much needed too.

[t0mpr1c3]

PCB a week is a really cool idea. 240 boards is a lot though! For others who may not want so many, OSHpark might be a good option for small boards because you pay by area.

[matseng]

The resistor is commonly used to improve stability and startup of the crystals. Some of the later revisions of the Arduino boards added them. But they are not really necessary and can be unpopulated without any problem - still having a 1M resistor there will not have any ill effects either.

I was thinking of adding a jumper to disable the bottom ground pin, but then I realized that a complete board will cost less than a dollar so I'll just build a few 16 and 20 Mhz models populated with both two and three pins at the bottom.

[matseng]

For small boards <5x5cm I usually select .8mm. IMHO 1.6mm is too thick for small board, also when using light components (smd) I don't think it is very much needed too.

Same here, 0.8 is my standard and I go up to 1.6 only if I have heavy parts on the board or I by some reason need the extra rigidity. Even a 0.8mm 10x10 board feel very stable but it can be flexed a bit if you try to bend it.

I also hope to save the environment a bit by using half the raw materials as well as reducing the costs for shipping since the weight should be almost half.

I'd imagine that a 0.8 would not be a good choice for a 20x30 panel when automated pick&place - that might be a bit too flexy...

[hak8or]

Did you test it to see if the total capacitive/inductive load would present any problems with such a header setup at 16 Mhz using a crystal?

Have you considered a oscillator instead, or even a voltage controlled one with a pot on it?

I love the idea of a new simple board every week though! Very good way to learn the entire process of designing a PCB including the EDA package you chose, and then how to get it fabbed and whatnot. I did the same with my super simple lux meter to remind myself how to use Altium and learn OSH park.

[matseng]

Nope... no tests yet. This is the prototype :-) But I would imagine that the extra capacitance from this contraption would be less than the capacitance from the breadboard.

It's an golden opportunity to do some torture testing and add a lot of extra capacitance to it. If it can start with say 56pF caps added a pair of 18pF or 12pF should be fine.

An variable oscillator controlled by a pot is a nice idea, but then a flying vcc lead is required as well. And a jumper to select whether xtal1 or xtal2 is osc_in.

If the frequency changes is done so that the minimum on- or off-time of the clock is not violated due to glitches during adjustment it probably would work.

Yup, it will only be smallish projects where the research, design and routing can be done in like 10-15 hours maximum, but that should be enough for a lot of smallish things. Especially if cheating by using the autorouter (Bad dog!)

[matseng]

The Gerbers are sent to Seeedstudio. I hope that they will accept that I have 24 copies of my design on each 5x5cm pcb. They say that only 5 copies are allowed, but they did 9 without any complaints a while back.

I also ordered a Bus Pirate acrylic SOB case and a handful of battery holders for 23A (12 volt) batteries that might come in handy some day. I had planned to use a standard 9 volt battery to power my PAW 3 project, but these might be an alternative.
http://www.seeedstudio.com/depot/-p-1272.html?cPath=207

Edit: Nah, I think I'll stick to the PP3 9-volt battery - they have like 10 times the capacity....

[Markus Gritsch]

It's an golden opportunity to do some torture testing and add a lot of extra capacitance to it. If it can start with say 56pF caps added a pair of 18pF or 12pF should be fine.

@matseng: Not directly related to your project, but I have the feeling you could know the answer: I use a ceramic resonator for my project and wonder which model I should choose, depending on the value of the built-in load capacitance.

I understand that the external load capacitors of XTALs should match the values specified in the XTAL datasheet, but what makes it necessary for ceramic resonators to be available with different built-in load capacitances.

I tried to find an answer using Google, but without any success.

[Bertho]

I understand that the external load capacitors of XTALs should match the values specified in the XTAL datasheet, but what makes it necessary for ceramic resonators to be available with different built-in load capacitances.
I tried to find an answer using Google, but without any success.

You may find these Microchip app-notes interesting:
- AN826: http://ww1.microchip.com/downloads/en/A ... 00826a.pdf
- AN849: http://ww1.microchip.com/downloads/en/A ... 00849a.pdf
- AN943: http://ww1.microchip.com/downloads/en/A ... 00943A.pdf

The load capacitors should match the load-capacitance of the crystal. The quick way to calculate them is to use (where C1 and C2 are the added capacitors):
Cload = (C1*C2 / (C1+C2)) + Cstray

For most designs it will suffice if you match the crystal load capacitance using this formula. You'll normally end up with values for C1 and C2 somewhere between 10p and 33p.

What I do not understand is why the series resistor was omitted while the parallel resistor was designed in. Matseng, care to comment?

[Markus Gritsch]

@Bertho: My question was not about crystals, but about ceramic resonators. They have the load capacitors already built-in. My question was why are there different models of the same ceramic resonator available only differing in the value of the built-in load capacitors. Depending on what should one choose one with 15 pF in contrast to one with 30 pF built-in load capacitance.

[Bertho]

@Bertho: My question was not about crystals, but about ceramic resonators. They have the load capacitors already built-in. My question was why are there different models of the same ceramic resonator available only differing in the value of the built-in load capacitors. Depending on what should one choose one with 15 pF in contrast to one with 30 pF built-in load capacitance.

Yes, your question was about resonators, but it applies to all types or resonators. The load capacitance and drive capability of the driver needs to be taken into account for a stable setup. If you have a high stray-capacitance, then you offset the frequency too much (maybe to the point of unstable operation) with high-load ceramic resonators. The combination of components dictates the setup.

The main difference between quartz resonators and ceramic resonators is accuracy and stability, where a quartz crystal is much more accurate and stable. Otherwise, you can generally treat them as the same device.

Se f.ex. app-note AN588 http://ww1.microchip.com/downloads/en/A ... 00588b.pdf where on page 3 they give a very good description of the inner composition of a ceramic resonator in contrast to a quartz crystal.

[matseng]

What I do not understand is why the series resistor was omitted while the parallel resistor was designed in. Matseng, care to comment?

From what I remember the series resistor is only used for not overdriving sensitive crystals, and most current common crystals aren't classified as "sensitive" today. And most modern microcontrollers are made for standard crystals without the series resistor. I can be totally wrong here - I'm not an expert on the subject.

The parallel resistor on the breakout doesn't need to be populated, the pads are there if needed. I think rev 3 of one of the Arduinos added a parallel 1M resistor to improve startup problems.

[Bertho]

From what I remember the series resistor is only used for not overdriving sensitive crystals, and most current common crystals aren't classified as "sensitive" today. And most modern microcontrollers are made for standard crystals without the series resistor. I can be totally wrong here - I'm not an expert on the subject.

You are not wrong, but there are other reasons as well.

The series resistor is also used to limit the Q of the crystal setup to limit the voltage upswing, and thereby protecting the clock input of the driver. When the tuning is right, then the upswing will cause the voltage to be significantly higher than the Vcc of the driver, which is dissipated by the protection diodes.

The series resistor adds a better phase-margin to the oscillator, so it will start more reliably. High frequency crystals are often 3rd overtone type, which requires a parallel resistor of significantly smaller value (order of 1k..10k). However, the parallel resistor is often not enough to dampen the system to a reasonable low level without creating a load on the driver that is too high. There you need to limit the drive-power, for which you add the series resistor. Without the dampening, the crystal refuses to swing at the overtone and will get stuck at the ground-frequency.

The app-notes I linked previously are very good at going into the details of the setup.

[Markus Gritsch]

I like your idea of a plug-and-play XTAL solution for breadboarding. I had to use some XTAL to get USB communication working with a PIC32. (It won't enumerate using the built-in RC oscillator.) Until I saw your post, I had a fearsome XTAL and SMD-capatcitor air-wire construction on my breadboard.

I had a DP protoboard [1] lying around, which contains an SMD XTAL footprint together with two capacitors.

Soldering was done using solder paste and my new (really cheap 16 €) hot air gun [2].

The result looks ok and works fine.

Cheers,
Markus

[1] http://www.seeedstudio.com/depot/-p-786.html
[2] http://www.amazon.de/Mannesmann-49500-H ... B000B9RK9O