Crystal not Oscillating

Thread Starter

Art

Joined Sep 10, 2007
806
Hi Guys :)
I’m using the VS1003 chip from VLSI which is an MP3 decoder chip for micros.
I connected the example circuit from the data sheet as shown, and it uses a 12.288MHz crystal.

When first powered, I noticed a crackle from the earphones, and figured out the crystal was oscillating intermittently.
It was only when I made the cut in red that it fires up reliably. Can anyone tell me why this is?
Perhaps I should have used the 22pF caps I normally use for any crystal up to 25MHz,
but I was just following the data sheet. I also don’t understand why it works without the cap on the input side.
Cheers, Art.

 

SLK001

Joined Nov 29, 2011
1,549
Too much capacitive loading on the xtal. If you are using a breadboard, the stray capacitances need to taken into account for your circuit.
 

Thread Starter

Art

Joined Sep 10, 2007
806
I’d like to see one on a breadboard :D

I see now, on any other occasion on protoboard, I’d do a loop to make the leads jump over the supply rails,
but this time it’s very close (even though they are only 10mm long).

 

RichardO

Joined May 4, 2013
2,270
How is the ground for the crystal caps getting back to the MP3 decoder chip? For that matter, how are the cap grounds tied together?

The crystal and caps should be right next to the oscillator pins on the decoder and the cap grounds should go directly to the nearest ground on the decoder. I would put the caps mechanically in parallel with the crystal with theie common ground in the center.
 

RichardO

Joined May 4, 2013
2,270
@Art More thoughts on your circuit.
Do you have a ground plane under the decoder? You should. Use copper tape on the back of the board and connect all ground pins of the decoder to it.

I see no high frequency power supply bypass caps in your picture. These are a must. You need a cap for every power/ground pair. This is a bunch of them.
Use the smallest 0.1 uF caps you can find and connect them directly from the power pin to the ground plane.
 

Thread Starter

Art

Joined Sep 10, 2007
806
Hi :) That's a lot of questions.
The ground plane of the board surrounds the circuit except for the left, and also a few horizontal traces are connected to ground.
Every cap on the board besides the crystal caps are for a supply (there are three), but the distance to the chip is difficult.
The cap leads can't be soldered to the LQFP. The resistors are only soldered directly because they connect to two pins. I have copper leaf to go under the board, where about every second thru hole on the ground traces becomes a via.
 

Thread Starter

Art

Joined Sep 10, 2007
806
The example schematic wants 3x1nF caps for each supply, but I couldn't get them on the board itself. There is a little more to the audio out which is also done off this board. Regulation for the 3.x Volt supply is also already done off this board.
 

SLK001

Joined Nov 29, 2011
1,549
I take it you've never encountered Bob Pease! :)
What you see in the photo is an actual FUNCTIONING op amp, made of all discrete devices. :)
I knew Bob Pease. The circuit on the left is the opamp. I don't have any idea what the circuit on the right is. My comment is based on using what I hope is enameled copper wire connected directly to the device. The pins on the SMT device are not that strong.
 

Thread Starter

Art

Joined Sep 10, 2007
806
The legs will be ok. The wire was soldered into the board, and then formed and cut to accomodate the chip. Only the two wires for the crystal are bare wire, that's stripped silver plated wrap wire.

The discreet circuit, I don't know what to think about that!
Someone who would hang the two TO92 cases upside down in the air like that, I'm guessing wouldn't orient the resistors uniformly either :D
 

RichardO

Joined May 4, 2013
2,270
Hi :) That's a lot of questions.
The ground plane of the board surrounds the circuit except for the left, and also a few horizontal traces are connected to ground.
The trace around the outside of the board is not a ground plane. It is an inductor. Because of this, your ground trace is actually a ground loop. You really do need a good region of ground under the decoder to make all of the ground connections to.

Every cap on the board besides the crystal caps are for a supply (there are three), but the distance to the chip is difficult.
This distance is shorter if you put the caps right next to the decoder and run the ground side of the cap through a hole to the ground plane on the other side of the board. You have many ground wires running around all over the place.

Notice that the ground path for the crystal caps is almost 2 1/2 inches long! The common ground of the crystal caps should go directly to the ground pin next to the crystal pins on the decoder. None of the crystal connections should not be more than a fraction of an inch -- with a total for all three crystal wires of less than an inch.

The cap leads can't be soldered to the LQFP. The resistors are only soldered directly because they connect to two pins.
I would argue that the cap leads_must_ solder to the decoder power pins. The power is then applied to the cap lead, not the decoder pin.

I have copper leaf to go under the board, where about every second thru hole on the ground traces becomes a via.
Are you saying that you already have copper leaf on the bottom side of the board? If so, that makes it a lot easier to move the caps next to the decoder power pins. If not then add it.
 

RichardO

Joined May 4, 2013
2,270
The example schematic wants 3x1nF caps for each supply, but I couldn't get them on the board itself. There is a little more to the audio out which is also done off this board. Regulation for the 3.x Volt supply is also already done off this board.
I see two caps and an inductor for each of the three supplies. It is very important to have all of the caps on the board with the decoder. The 0.1 uF caps (the ones we have been discussing) must be right next to the decoder. The 10 uF caps can be near the edge of the board. The inductors can be off the board.
 

GopherT

Joined Nov 23, 2012
8,009
1) it looks like the case of the crystal is shorting the resistor below it.

2) I made a color print of your circuit and posted it above my workbench. Beautiful art, @Art.
 

Thread Starter

Art

Joined Sep 10, 2007
806
1) it looks like the case of the crystal is shorting the resistor below it.

2) I made a color print of your circuit and posted it above my workbench. Beautiful art, @Art.
1) The crystal still has about 5mm legs, and the resistor is in the air between the crystal and the ground trace underneath it.
... and like I said.. it has worked reliably since I made the disconnection in the first post.
2) Thanks :)

Richard, I see where you’re coming from, but most of the standards you are applying to the dead bug are better met by the dead bug
than the commercial board it came from :D and every other commercial vs1003 board (shield),
where all supply traces are longer than mine, and almost all caps are further away than mine.
The closer caps are only the transient protection for the headphone socket.
Granted I did other dirty things like crossing signal 45 degrees over supply lines, but it is dead bug.

 
Last edited:

Thread Starter

Art

Joined Sep 10, 2007
806
Oh, I see what you mean moving the ground side of the crystal caps being easy.
The copper leaf isn't there yet, but there are lots of closer grounds then.
About half of the unused holes of anything common to ground become vias with dwell time I'm comfortable with,
which is easy to tell because some wet solder will draw through.
 

RichardO

Joined May 4, 2013
2,270
Oh, I see what you mean moving the ground side of the crystal caps being easy.
The copper leaf isn't there yet, but there are lots of closer grounds then.
About half of the unused holes of anything common to ground become vias with dwell time I'm comfortable with,
which is easy to tell because some wet solder will draw through.
I would not do any changes until you have the copper in place under the decoder. That way you can clean up all of the weaknesses of the long wires at one time.

I am not sure I understand correctly but I believe that you intend "wicking" the solder from the present ground traces to the copper on the other side of the PCB.
If this is your plan I suspect that you will be disappointed. I don't think that the PCB has plated through holes. At best you can do is to poke a hole in the copper foil and then put a wire through hole from the ground trace.

If you move the bypass and crystal caps to be near the decoder then the copper foil does not have to be very large extend under the capacitors.
 

RichardO

Joined May 4, 2013
2,270
Richard, I see where you’re coming from, but most of the standards you are applying to the dead bug are better met by the dead bug than the commercial board it came from :D and every other commercial vs1003 board (shield),
where all supply traces are longer than mine, and almost all caps are further away than mine.
The closer caps are only the transient protection for the headphone socket.
Granted I did other dirty things like crossing signal 45 degrees over supply lines, but it is dead bug.
Just because someone sells it does not mean it is right. That PCB in you picture is very sloppy for no reason whatsoever. For instance the resistor and caps for the crystal could be moved to be between the crystal and U1. C15 could easily be moved right next to U1 where it belongs.

This tells me that the person doing the PCB layout did not care or did not know how to do it right. Sad, either way. If the rest of the PCB layout out is as bad I suspect that it does not work well and would not meet emission standards.
 
Top