I'll pre-empt the poll and suggest crystal oscillator can wins, but if so what particular brand do people prefer and are there any gotchas to look out for?

I ask as 1) I'm interested 2) will be useful for others 3) recent experience

My recent experience was thinking that there was an issue with my design and/or firmware on my original 6502 system which was limiting it to 10MHz. I was trying a 12MHz oscillator can and found that it didn't work. Measuring PHI2 showed a 12MHz square wave and harmonics, etc. all looked normal.
I eventually tried an IQXO 12.28MHz can and it worked...

Actually, I use an oscillator feeding into a flip-flop for a strong signal with a near-exact 50 percent duty cycle.


I use the ECS 2100AX series, but haven't really found any one brand to be superior to another.

One possible gotcha is the temperature coefficient rating. That could be important if you are dependent on the frequency being very stable with changes in ambient conditions. In practice, few of our creations are likely to be that fussy.

_________________
x86?  We ain't got no x86.  We don't NEED no stinking x86!

I voted for the oscillator can; but note that many of them take a huge amount of current. It might be an issue if you want to run the thing on battery power.

The first commercial 65c02 design I did (in the late 1980's) took about 2mA for the entire computer, including the LCD, on battery power. It did not need much speed, so it ran most of the time at 170kHz, and kicked it up to 1MHz for brief times it had to do a string of floating-point calculations. (I later learned I could have done it just as well with scaled-integer, reducing the performance requirements. There's a discussion of that in the top half of the page at http://wilsonminesco.com/16bitMathTables/ but you don't need the tables to take advantage of scaled-integer math methods.) It used a Schmitt-trigger inverter for an oscillator, with speed set by a resistor and capacitor. Under software control, another resistor was switched in and out to change the speed. It did not need crystal accuracy.

In 1993, I did another one to control our top-of-the-line aircraft intercom, and again exact speed was not critical, and I just hung a resistor and capacitor on the 65c02 itself as shown in the 3rd circuit on the clock-generation page of my 6502 primer. We sold that product for 16 years or so.

_________________
http://WilsonMinesCo.com/ lots of 6502 resources
The "second front page" is http://wilsonminesco.com/links.html .
What's an additional VIA among friends, anyhow?

I also voted for the Can oscillator. I use the CTS MXO45HS (half-size) which fits into either an 8-pin DIP or the bespoke 4-pin sockets. I drive the CPU and UART directly from the Cans... no additional buffering/shaping. Current draw is pretty light, 10ma avg for a typical CMOS load. I like the small size as well, and try and keep my PCBs pocket-sized, probably a side effect from driving small cars

_________________
Regards, KM
https://github.com/floobydust

It depends on how you define "huge." Contemporary cans aren't anywhere as power-hungry as they were in the past.

_________________
x86?  We ain't got no x86.  We don't NEED no stinking x86!

15mA is ridiculous for just an oscillator, but nowhere near as bad as the 75mA of 25 years ago.

_________________
http://WilsonMinesCo.com/ lots of 6502 resources
The "second front page" is http://wilsonminesco.com/links.html .
What's an additional VIA among friends, anyhow?

I did this earlier to get my original 65C02 system to behave at 12.5MHz (25MHz osc can divided by 2). Worked nicely.
My new design (still being debugged) has this built in to it.

While hunting around on ebay I found several which quoted 25mA which I thought a little extreme. 75mA is just insane lol.

Fifteen mills is inconsequential if the circuit is being run from a power supply. Naturally, the higher the frequency the greater the current consumption.

_________________
x86?  We ain't got no x86.  We don't NEED no stinking x86!

Most of these oscillators don't guarantee an exact fifty percent duty cycle, which might have been a factor in why you were having trouble. For example, the data sheet for the ECS 2100A units I use quotes an output symmetry of 50 percent, +/- 3, with a maximum of 55 percent either way. If you are operating your system near its timing limits that mas o menos 55 percent could be enough to cause a timing violation during one half the machine cycle.

Another possible gotcha has to do with the oscillator's output rise and fall time. Again, the ECS 2100A data sheet says thqat number is up to 5ns. The output of a 74AC74 flop slews in about half the time, 74ABT74 is even faster. Either is well within the requirements of the 65C02 and 65C816.

_________________
x86?  We ain't got no x86.  We don't NEED no stinking x86!

I was seeing approx 52-53% / 48-47% ratio using an oscillator can (on average). I have to admit that I didn't consider the ratio and was more thinking about the drive capability at the time/

BTW does it matter that I've used an HC rather than HCT part with regard to this?

Anybody ever use a ceramic resonator ? I've tried them a couple of time but found them to be too unstable.

_________________
http://www.finitron.ca

Currently I prefer microcontrollers. I let them run twice the rate of the planned clock rate of the 6502 and use a PWM output as PHI2 source. Preferably a baud-rate crystal is used, this also let you use another PWM output as clock-input for your USARTs. Besides providing PHI2 I use the microcontroller to load the ROM image via IML, Initial Machine Load as in old IBM days (or blindload as Michael calls it). As I'm using a 328P I have enough flash to hold even a larger ROM image. So I don't need any ROMs in my designs. I also added a PS/2 interface to the same microcontroller. The microcontroller runs a small control program with which I can control the clock speed or even stop the clock and perform single cycles or download an alternative ROM or Program Image via XModem. And all this you get for about 10-20mA.

Old IBM days... IML was Initial Microcode Load. Back then, the IBM machines had Control Storage and Main Storage. The IML loaded the microcode, which made up the processor's instruction set into Control Storage. The operating system and applications were then loaded into Main Storage. This allowed patches/updates to the actual processor microcode which was pretty cool

_________________
Regards, KM
https://github.com/floobydust

Only with PICs. They seemed to work ok with them.

_________________
Andrew Jacobs
6502 & PIC Stuff - http://www.obelisk.me.uk/
Cross-Platform 6502/65C02/65816 Macro Assembler - http://www.obelisk.me.uk/dev65/
Open Source Projects - https://github.com/andrew-jacobs