MARC-2 project

[lordbubsy]

Thanks for the article.

A crystal with two 22pF caps is what I also use with my AVR projects for years without any issues.

[GARTHWILSON]

A crystal by itself won't oscillate; but if your UART has an onboard oscillator (like many microcontrollers do) and is made to just hang a crystal off it to control the frequency, then there's nothing wrong with that. A capacitor from each side of the crystal to groung is commonly needed.

[Aslak3]

GARTHWILSON wrote:

A crystal by itself won't oscillate; but if your UART has an onboard oscillator (like many microcontrollers do) and is made to just hang a crystal off it to control the frequency, then there's nothing wrong with that. A capacitor from each side of the crystal to groung is commonly needed.

Yes indeed. Obviously an internal oscillator circuit is required, but this is something that is pretty generic; certainly the MC6809 (c. 70s) which I use have this - I assume the contemporary NMOS 6502s did too, as do the current AVR microcontrollers, and pretty much everything in between.

Obviously a can oscillator will outpace a crystal plus internal oscillator plus caps in at least dealing with temperature compensation. I'm sure there are other advantages.

[GARTHWILSON]

Quote:

I assume the contemporary NMOS 6502s did too

The CMOS one does, and I have used it; but the datasheets for the NMOS one specifically call for an external clock source. I haven't tried using the NMOS circuit as an onboard oscillator. I don't know if maybe the problem is insufficient gain, even if the configuration might be right. Designing oscillators that are accurate and stable under all anticipated conditions is not trivial, but the details should be already taken care of in any design that is specified to work with just an external crystal and two load capacitors. In the early years of can oscillators, it seems like they were all power hogs (like 75mA for just the oscillator!). I haven't shopped for any recently but I get the idea that they've improved a lot.

[BigDumbDinosaur]

Figure 7 on page 24 of the 26C92's data sheet shows the hookup for using a crystal as the frequency source.

Figure 7: 26C92 Oscillator Circuit

C1 and C2 are dependent on the crystal used. I can't vouch for the performance of this circuit, as I haven't tried it.

[BigDumbDinosaur]

GARTHWILSON wrote:

In the early years of can oscillators, it seems like they were all power hogs (like 75mA for just the oscillator!). I haven't shopped for any recently but I get the idea that they've improved a lot.

The ones I've been using with POC V1 consume 17 to 33ma total at 25° C, depending on frequency, and generate 16ma output. As output symmetry is not guaranteed to be perfect (spec is 50 ±3 percent) , I derive my Ø2 clock from a flop (74ABT74 or 74AC74), which also sharpens the rise and fall times and provides more drive.

ECS (the oscillator manufacturer) states that a maximum of 10ms are required for oscillation to start following power application. As the Maxim DS1813 econo-reset holds /RESET down for at least 150ms, the Ø2 clock will be stable by the time the MPU comes out of reset.

[BigDumbDinosaur]

Aslak3 wrote:

Obviously a can oscillator will outpace a crystal plus internal oscillator plus caps in at least dealing with temperature compensation. I'm sure there are other advantages.

If using a half-size can oscillator, there's usually a savings of board real estate.

[clockpulse]

BigDumbDinosaur wrote:

Figure 7 on page 24 of the 26C92's data sheet shows the hookup for using a crystal as the frequency source.

26c92_oscillator.gif

C1 and C2 are dependent on the crystal used. I can't vouch for the performance of this circuit, as I haven't tried it.

I used a crystal with the 26C92 when I first started using the chip.
After a few weeks the UART began screwing up rx/tx data. It took me a while to figure it out.
As a last resort I switched to a half can osc. and my troubles were over. Could have been the crystal was defective, but it was odd that it worked fine for a while.

[lordbubsy]

I’ve come up with a solution to ease myself with leaving two options open to use on the circuit board. Given all the information I find the flip-flop the most appealing.

As a backup option, I’ll leave room for a crystal and two capacitors.

I also looked at the datasheet of the can oscillators I have (1,8432 and 7,3728 ), and indeed they are improved since the 75mA. They need 20mA to 35mA depending on the frequency.
The odd thing is that my “local” supplier has can oscillators with a lot of different frequencies, except for the 3,6864MHz.

[cbscpe]

I suppose the Flip-Flop uses a VCC=5V, so I think you better use a 74ACT74 so the 3.3V Logic Levels will be clearly recognized as high level. The AC only version are a bit tight, they need 3.15V at VCC=4.5 and for higher VCC it will be clearly more than 3.3V. The ACT version will recognize 2.0V as High Level regardless of Supply (4.5 to 5.5V). You also don't need R16.

Peter

[GARTHWILSON]

Quote:

You also don't need R16.

+1. CMOS already pulls up pretty much all the way to its positive power supply rail. 74ACT is also quite a bit stronger than 74HCT. The resistor is only recommended if you have to use something like 74xx TTL or 74LS.

[lordbubsy]

cbscpe wrote:

I suppose the Flip-Flop uses a VCC=5V, so I think you better use a 74ACT74

Thanks, your right. I have the 74HCT74 here at home. If it won’t suffice, I have to search for a place where to buy a few ACT’s as a private person. I normally order from http://www.rechelt.de. They don’t have a lot of special or rare parts, but for the most parts they’re quite fast and not too expensive.

cbscpe wrote:

You also don't need R16.

GARTHWILSON wrote:

The resistor is only recommended if you have to use something like 74xx TTL or 74LS.

OK, I wasn’t sure about R16, it was mentioned in the datasheet, so I added it for the case it would be needed.

[BigDumbDinosaur]

lordbubsy wrote:

cbscpe wrote:

I suppose the Flip-Flop uses a VCC=5V, so I think you better use a 74ACT74

Thanks, your right. I have the 74HCT74 here at home. If it won’t suffice, I have to search for a place where to buy a few ACT’s as a private person. I normally order from http://www.rechelt.de. They don’t have a lot of special or rare parts, but for the most parts they’re quite fast and not too expensive.

It appears that Reichelt doesn't stock anything in 74ABT or 74ACT. As it is expected to only drive the xCLK input of the 26C92, I don't think you need to get too excited about how hard the 74HCT74 drives its output.

[lordbubsy]

MARC-2 is taking shape.

• The system is based on the W65C816.
• It is ROM-less with 512 KiB RAM.
• All pins of the CPU and RAM are connected individually to the CPLD, a Xilinx XC95288XL 144 pin. The CPLD will also provide 65SPI and VGA.
• The ATMega1284P provides RESET, master clock and the firmware to the RAM during RESET.
• The DUART, a SC26C92, provides RS232 serial and TTL serial.
• There are two VIA’s for future expansion.
• One of the eight 65SPI interfaces is used for a Real Time Clock with environment temperature sensor, a RV-3049-C2.

Initially I’ll try to get the computer going by only using the TTL serial port which also provides the 5V power. Later on I’ll add 65SPI and VGA. The CPLD should have enough resources to do even more.

At this moment the PCB is etched and all holes are drilled. Next, I have to solder the vias by hand and check all traces.

[BigDumbDinosaur]

lordbubsy wrote:

MARC-2 is taking shape.

I like your illustrations. What software do you use to render them?