Help: Crystal Oscillator is not working.

[ProfessorCagan]

The scope is working correctly! I tested it's 1KHz output and it worked just fine, I also got 50Hz out of the oscillator, and the voltage was peaking around 2.5v.
Is the oscillator defective?

[Dr Jefyll]

Fifty Hertz?? That sounds suspiciously like ambient noise from the AC mains (assuming you're in a country that uses 50 Hz, such as UK and much of Europe). As well as the signal, have you attached the ground lead of your scope probe to ground on the project?

-- Jeff

[ProfessorCagan]

I live in the states, so I'd expect 60Hz if that were the case?
I've used probes quite a bit for troubleshooting lessons in Devices II, usually on OPAMPs. Admittedly we've never spent much time drawing schematics or creating actual circuits in my classes, we mostly went over theory and formulas.
But yes, the probe is connected correct,

[leepivonka]

On the scope: I don't see much for specifications on it, but it does mention 1Msps - does this mean it can do up to 1M samples per second? If that's true, it won't be able to reliably show the fundamental frequency of a 1MHz signal, let alone any higher multiples to show how sharp the edges are. The 50Hz signal you're seeing could be the beat between the oscillator signal & the scope's sampling rate.

On the oscillator: 2.5v peak could just barely meet TTL voltage levels ( >2.4v hi, <.8v lo).
Check the 6502 datasheet for your CPU version - this may not meet the clock input requirements.
You could improve the signal by adding a pull up resistor (try 470 ohm to +5v) or running it through a 74hct gate (try 1 section of a 74hct04).

Most 6502 variants can run at clock frequencys down to 100KHz, any some down to DC. Check the datasheet for your CPU version for details.

[Dr Jefyll]

ProfessorCagan wrote:

But yes, the probe is connected correct,

OK, just checking.

What about the fourth pin on that oscillator? Besides the Gnd and power pin and the Output, some units also have an Enable input that needs to be tied high. As an experiment, I'd just go ahead and try attaching that pin to +5.

And your little test circuit could use a bypass cap, BTW.

-- Jeff

Edit:

leepivonka wrote:

The 50Hz signal you're seeing could be the beat between the oscillator signal & the scope's sampling rate.

Good point.

[BigEd]

If the oscillator is a bit fast for the scope, perhaps you can divide by 2, 4 or 8 using a TTL part, and check the output of that divider.

[ProfessorCagan]

Leepivonka, I couldn't find the CPU's clock voltage requirement in its datasheet, but I'm gonna go ahead and try Dr. Jefyll's suggestion, and leave the results in this post.
Ok, so I continued to get around 50Hz measured when I tried connecting 5v to that last pin.

[cjs]

ProfessorCagan wrote:

The scope is working correctly! I tested it's 1KHz output and it worked just fine, I also got 50Hz out of the oscillator, and the voltage was peaking around 2.5v.

Looking at the pic you posted, you forgot another critical piece of information: it's not only peaking at around 2.5 V (actually 2.9 V in your image; I assume that the 'scope display is not compensating for the 10x probe), but Vmin is 2.1 V. So from a TTL point of view there's no signal there; it never swings below 0.5 V to become a '0'.

But as leepivonka points out, it didn't occur to me to look at the 'scope specs. I don't see anything for sampling rate there, but it does say that the time base range starts at "10μs / Div", which is not a good sign; if it sampled fast enough to capture a 1 MHz waveform surely they wouldn't make you display ten cycles of it in a division, making it unreadable. (The picture of a running screen shows a 140 kHz waveform.) So yeah, I'm thinking that probably this 'scope can't be used to check this clock.

No fear, though; you can run the W65C02 at any speed, so if you can get hold of a slower crystal, or divide this one down per Ed's suggestion, that will let you check the clock. If neither of those options is handy, I'd start by hooking the clock back up to the CPU, pulling up/down the data bus to generate a NOP, and then trying your 'scope on the high address lines, which I hope might be down in a range that your 'scope will show. On a CPU running at 1 MHz, and at two cyles per NOP, A0 should be changing at 500 kHz, A1 at 250 kHz, and so on all the way up to A15 at 15.25 Hz. (Someone check my math there.) And this is exactly what I saw on both NMOS and early CMOS 6502s, last time I tried this out. However, I don't know if the W65C02 behaves in the same way, maintaining the address bus value throughout the entire cycle until it needs to change it.

(If this does work, you also now have built yourself a handy tester to see how high your 'scope goes. :-))

If this doesn't work, I'd do a quick check of the /RESET line, triggering your 'scope for a one-shot capture when you press the button, if it can do this. If you're having a good day it will look fine, but if the reset line is a big bouncing mess when you press the button, you might want to consider building a proper reset circuit.

If none of that works out, while you're waiting for a slower oscillator maybe try building a manual-step clock so that you can click a button to cycle the CPU, and see if that, with a NOP generator, produces some expected results on the address lines.

BTW, 32.768 kHz oscillators are really common and cheap, and well within the range of your 'scope unless the picture is a complete lie.

[BigEd]

Although packaged chips are quite robust, crystals are fragile mechanical systems: it might be worth noting that a crystal oscillator could be damaged in transit or by falling onto a hard surface.

So making sure you do have a clock signal is a crucial first step.

[Dr Jefyll]

The 32.768 kHz oscillator is a great idea. The CPU won't mind running that slowly, and there's cause to suspect the lower speed will make a world of difference regarding the 'scope.

BigEd wrote:

a crystal oscillator could be damaged

Well, until a new oscillator is obtained, it might be worth testing the old one with a voltmeter (or multimeter). An attempt to read DC voltage on the output may seem counterintuitive, but it's better than nothing. What you'll get is the approximate average voltage. A reading of 0 or 5 volts surely indicates a defect, whereas a reading of 2.5V or slightly less suggests (doesn't prove) all is well.

Moreover, a multimeter may feature a Frequency range, which of course ought to indicate 1 MHz (assuming the meter can read that high).

-- Jeff

[cjs]

Dr Jefyll wrote:

Moreover, a multimeter may feature a Frequency range, which of course ought to indicate 1 MHz (assuming the meter can read that high).

Oh right. Actually, a frequency counter (whether built in to a multimeter or stand-alone) can be a useful tool if you don't have an oscilloscope, or the frequencies you're checking are out of range of the 'scope. I was going to mention this earlier but it dropped out of mind once I found out that there was a 'scope available.

That said, a 'scope provides significantly better information, so it's not worth spending additional money on a frequency counter unless it's significantly cheaper than buying a 'scope that can go to a couple of MHz or more.

Another tool that's useful when you don't have a 'scope is a logic probe, and some can actually identify high frequency signals. (E.g., a logic probe might be able to tell you whether that clock line is really switching between 0 and 1 TTL levels, or is just hovering around some intermediate voltage.)

[ProfessorCagan]

Ok, guys, I do have a meter that can read frequency, and to be honest, I don't know why I haven't tried it yet, slipped my mind I suppose.
So, I can either use the scope for checking pulled address lines, or the meter for checking the oscillator outright (if it can,) or I can purchase a slower crystal.
Then get back to debugging the CPU?

(Side questions, using a smaller crystal would let me see the LED's change correct, if it was slow enough of course?
When I last tried it, the LEDs that came on stayed on, and the ones that were off stayed off. I assume that is because was not receiving a clock signal?)

Thanks again for the help you've all given me so far, and I hope Christmas is treating you all well!

[ProfessorCagan]

Ok, so, the oscillator works correctly, tested it with the multimeter, I should've done that from the beginning, would've saved us all a headache, I imagine, my apologies.
But now that it is out of the way, I'd still like to figure out (see quote below:)

ProfessorCagan wrote:

(Side questions, using a smaller crystal would let me see the LED's change correct, if it was slow enough of course?
When I last tried it, the LEDs that came on stayed on, and the ones that were off stayed off. I assume that is because was not receiving a clock signal?)

[1024MAK]

Yes, a lower frequency crystal or crystal oscillator can will clock the 6502 microprocessor slower, hence it’s whole operation will be slower including the frequency that the address pins change.

With a NOP generator, the microprocessor will read the instruction (NOP - telling it to do a ‘No Operation’), it will then carry out NO register or ALU operation, but will increment the instruction register as normal. Then at the next cycle, it will put the new address on to the address bus (the A0 to A15 pins) and read the next instruction from the data bus (the D0 to D7 pins), which will still be a NOP. So it will repeat everything again and again.

Because the number on the address bus is being incremented, A0 will change the fastest, with each higher numbered address pin changing state at half the frequency compared to its lower numbered neighbour. This applies all the way up to the A15 pin.

So connect your ‘scope, or multimeter to each address pin in turn from A0 to A15 and see what you get.

Mark

[cjs]

ProfessorCagan wrote:

Ok, guys, I do have a meter that can read frequency, and to be honest, I don't know why I haven't tried it yet, slipped my mind I suppose.

It slipped everybody's mind for quite a while, it seems. But it's a good lesson: there are a lot of different ways to debug things and get insight into what's going on in a circuit. If one approach isn't being helpful enough, get creative and find another. Each way of measuring something gives you something different (and invariably incomplete) information.

ProfessorCagan wrote:

Ok, so, the oscillator works correctly, tested it with the multimeter...

That's good to see. But do remember, just because it's a 1 MHz signal does not mean that it's of the appropriate shape to properly drive the CPU, nor does the signal you see when the output is not connected to anything necessarily match what the signal is when the oscillator is plugged into the circuit. But this is definitely a good additional piece of data!

Quote:

So, I can either use the scope for checking pulled address lines, or the meter for checking the oscillator outright (if it can,) or I can purchase a slower crystal.

I'd do all three, since you're at a moderately early learning stage with this stuff. In particular, a slower crystal will give you a lot more flexability when walking through your circuits with your 'scope, since much more of the circuit will be visible to it.

Quote:

...using a smaller crystal would let me see the LED's change correct, if it was slow enough of course?

Yes, exactly. But "slow enough" is probably somewhere around 4-5 Hz; much beyond that and the LED will start to look more like a constant but dimmer light. (In fact, turning LEDs on and off quickly is how many LED displays work.)

If the CPU with NOP pull-ups/downs on the data bus is working correctly, at 1 MHz the A15 line should be toggling at around 7.5 Hz. That may be too fast to see as actual blinking, depending on the particular conditions of the LED circuit, and of course each successively lower address line will be twice the rate of the previous. At 32.768 kHz A15 should be blinking at 0.25 Hz, or be on for two seconds followed by off for two seconds, which will be easily visible.

Quote:

When I last tried it, the LEDs that came on stayed on, and the ones that were off stayed off. I assume that is because was not receiving a clock signal?

No, that's not a good assumption to make, though it may be a reasonable hypothesis to start with. As I mentioned above, an LED being "on" may simply indicate that it's being turned on and off too fast for you to see with the human eye, an LED being unexpectedly on or off may indicate a problem with the LED or its part of the circuit, it could be that you're getting a clock but the CPU wasn't reset properly or has another signal preventing the exact operation you're hoping for (e.g., READY being low), the clock or other waveforms might be dodgy, or any of a number of other things.

So really, when things start going weird, it's often helpful to go back to the very start by checking the power lines, then the clock, then working out from there to every other signal in the system.