Oscillator

It's a good idea to check that your crystal is working as well as you need.

There's ways to just probe the oscillator itself if pins are scarce, particularly if you're willing to build a board specifically for probing, but driving a pin on your microprocessor should work. If you don't bring that pin out to any traces, but just drive it from the micro, you should be able to get a probe on it (assuming that it's a DIP or PLCC -- obviously that won't work on a BGA).

--

Tim Wescott 
Wescott Design Services 
http://www.wescottdesign.com

The thing is that the micro will be on the bottom side of the printed circu it board. The tester will not be able to acess the micro or crystal pins us ing scope. So, I am planning to route that clock trace to a 2.54mm spacing single row three pin connector.

The connector will allow the user to hook upi the scope to one of the pins of the connector. There will addition two other signal avialable on the re st of the two pins.

Is this a good idea? Will it cause any EMC problems during EMC testing or a ny other interference problems?

jess

The printed circuit board will be mounted inside the box and its top side can be accessable to tester. The micro will be on the bottom side of the printed circuit board. jess

Any digital signal has this potential. Are you going to divide the clock down? That'll help.

--

Tim Wescott 
Wescott Design Services 
http://www.wescottdesign.com

On a sunny day (Wed, 31 Jul 2013 14:34:56 -0700 (PDT)) it happened snipped-for-privacy@gmail.com wrote in :

I do that often with PIC, just to make sure I have the right clock working. When all OK re-program the pin for I/O.. I see no problem leaving it on, as other pins will carry fast edges too. But that costs an I/O pin.

I would not bring out the crystal frequency directly. It should be possible to program one of the micros timers to output a divided-down version of the clock on an I/O pin, on some signal trace that is already accessible. Perhaps as part of a special test mode. The frequency/period of this will scale directly and precisely with the crystal frequency.

E.g. some of the real time clock chips have a "1 Hz" output that can be used in a similar way to calibrate the 32768Hz crystal.

--

John Devereux

If you have a shortwave receiver that can listen to single-sideband transmissions (most can), and if it works at your 17MHz (most do), then you can pick up the signal from the crystal oscillator on the radio without interfering with the oscillator operation.

You may need to hold the board up to the radio, or make an antenna probe that you can hold right next to (but not touching!) the oscillator, but it should come through like gangbusters.

With care you can verify the frequency of operation as well as the presence of the signal, all without disturbing the operation of the oscillator.

--

Tim Wescott 
Wescott Design Services 
http://www.wescottdesign.com

That's the way to go. If the uC has to blink an LED let that come off a timer that provides an exact divider ratio. Then switch the frequency counter to period measurement and ... bingo.

--
Regards, Joerg 

http://www.analogconsultants.com/

It is kind of obvious if you pull out crystal signal from logic circuit, that would affect crystal frequency. Coupling is changed.

It is not so obvious if you pull out divided frequency, that influences crystal as well, although the effect is smaller.

This is very noticeable with MCU built-in oscillators.

Vladimir Vassilevsky DSP and Mixed Signal Designs

On a sunny day (Fri, 02 Aug 2013 17:51:57 -0700) it happened Joerg wrote in :

Actually it is not. Ther is no way to tell if you programmed any of the prescalers and counters right. You START with a clock verification.

Afraid I don't quite follow what you are saying. How can a divider inside the uC affect the crystal frequency?

--
Regards, Joerg 

http://www.analogconsultants.com/

Huh? If I have a 16MHz crystal on the uC, run it through a timer that is supposed to deliver 0.5Hz. When I then measure a stable 0.500000000Hz, how is that not verification that the crystal oscillator frequency is correct?

It is quite inconceivable that a crystal operates in a stable fashion on some unwanted frequency and it just so happens that I misprogrammed the timer register so that it exactly compensates this.

--
Regards, Joerg 

http://www.analogconsultants.com/

On a sunny day (Sat, 03 Aug 2013 08:04:48 -0700) it happened Joerg wrote in :

So you are used to simple micros. How do you know your timer code is coreect or you osicllator? You are an idiot.

You should really for example try reading a datasheet of a real micro, start with a simple PIC like 18F14K22

There are more catches there in the config lines that you have ever programmed code in your entire life. Oscillator configurations, PLL configurations etc. Idiots are those who measure outside temperature in the bathroom. You measure where you want to know if possible, and that is really simple in this case as I pointed out. Will teach you how to program the thing too.

If you connect divided output to anyting, that affects crystal frequency. The divider is coupled to oscillator through power rails; so changing load conditions influences the oscillator. Feedback from connected probe through the air affects the oscillator, too. I was surprised to see how significant is this effect even at high division ratios.

VLV

I have never seen any modulation by that. Ok, if you drive drive a coffee cup warmer with it and have no power plane, maybe :-)

A lot of old-style (not so integrated) PLLs from the 80's relied on the fact that such modulation would not happen, and in mine it didn't.

--
Regards, Joerg 

http://www.analogconsultants.com/

All kinds of unexpected coupling occurs thru the substrate of a chip. It's not nearly so bad now, in a CMOS world, but, in TTL days, I had to use all kinds of gimmicks to keep saturating stages from injecting currents into substrate. ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    | 
| San Tan Valley, AZ 85142   Skype: Contacts Only  |             | 
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  | 
| E-mail Icon at http://www.analog-innovations.com |    1962     | 
              
I love to cook with wine.     Sometimes I even put it in the food.

Yes, substrate can be nasty. Though on a well designed IC they should have taken care of "cocooning" the oscillator enough. That's what we do with pulse drive paths on ultrasound chips, got to do it. Sometimes we even sink off-phase dummy currents. Like they do with the balance shafts on luxury cars, to make them run smooth.

But I doubt Jessica will care much if the clock is 17.000000MHz or

17.000050MHz. If she does then I would never even think about using the on-board oscillator. That's where dedicated TCXO's come in, they don't cost much anymore these days.

--
Regards, Joerg 

http://www.analogconsultants.com/

How do you know?

Think it depends a bit on the situation. During initial development of the system, bringing up new hardware and writing drivers, yes I would just probe the clock pins directly as a quick check that things were working. Not as a precision frequency measurement, just to see if it was oscillating properly.

But if the purpose is to do a production calibration of the oscillator, (or compensation of a frequency error), then you need to stay well away from it. And using a divided-down version is ideal for that.

And the last thing you want to be doing with a high frequency clock crystal signal is bringing it out any significant distance. They can be very susceptible to interference, and can dominate emissions too.

--

John Devereux