6502.org Forum  Projects  Code  Documents  Tools  Forum
It is currently Wed Nov 13, 2024 1:33 am

All times are UTC




Post new topic Reply to topic  [ 49 posts ]  Go to page Previous  1, 2, 3, 4  Next
Author Message
PostPosted: Tue Oct 24, 2017 11:05 pm 
Online
User avatar

Joined: Fri Aug 30, 2002 1:09 am
Posts: 8540
Location: Southern California
See Jeff Laughton's animated visualizations of timing margins. These excellent, drawn-to-scale (unlike most in data sheets) .gif's help understand what timings are constant and what varies with clock speed. Discussion about them is in the forum topic here.

_________________
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?


Top
 Profile  
Reply with quote  
PostPosted: Wed Oct 25, 2017 7:52 am 
Offline
User avatar

Joined: Thu Dec 11, 2008 1:28 pm
Posts: 10977
Location: England
mdpye wrote:
Quote:
However, in the design of my POC V2's logic, I worked off the fall of Ø2.


I'll have to take a look at that. I'm still pretty hazy on how one adds deliberate (and especially fixed, rather than expressed as a percentage of a cycle) delays to a signal to account for e.g. setup times after an edge rather than working directly from a clock edge.

Using phi2 is very convenient and extremely common for that reason. Adding up the worst-case paths through logic, or the best-case paths as the case may be, might also result in a reliable and robust design. It's not unknown to use RC delays - especially to produce the sequenced strobes which DRAM needs - or to use divisions marked out by a higher-speed clock. It's not necessary for phi2 to be a 50% duty cycle, although it usually is.

From an engineering perspective, it's a matter of understanding what's a convention and what's a constraint.

FWIW even something as popular and long-lived as the BBC Micro isn't actually a very good design from this perspective. And Bil Herd has written several amusing stories about the state of play within Commodore - not good at all! But by trial and error and some luck and judgement, they made some very successful machines.

For a hobby project, built as a one-off, an empirical approach can take you a long way. Build a 100 to sell, and if you get 10 or 20 returns or field failures, that's going to be a lot of a grief.


Top
 Profile  
Reply with quote  
PostPosted: Thu Oct 26, 2017 9:43 pm 
Offline

Joined: Mon Oct 23, 2017 9:16 pm
Posts: 22
Quote:
See Jeff Laughton's animated visualizations of timing margins. These excellent, drawn-to-scale (unlike most in data sheets) .gif's help understand what timings are constant and what varies with clock speed. Discussion about them is in the forum topic here.

Thanks. I'm familiar with the animated diagrams, they were a fantastic aid in understanding the implications of the timing requirements.

I'm fairly clear on the *requirements*, what I'm more hazy on is the other side. If I'm told that a signal I want to latch will be stable x nanoseconds *after* a particular clock edge, what are good approaches to relating my latch enable to said clock? What is canonical good practice for introducing a precise deliberate delay in to a signal?

My guess from my experience with design in FPGAs would be to take it from the edge of a faster clock, counting edges if required. Or an deliberately introduced rc delay, although that feels a little... soft or imprecise to the software engineer in me (that's my background).


Top
 Profile  
Reply with quote  
PostPosted: Thu Oct 26, 2017 11:20 pm 
Online
User avatar

Joined: Fri Aug 30, 2002 1:09 am
Posts: 8540
Location: Southern California
mdpye wrote:
My guess from my experience with design in FPGAs would be to take it from the edge of a faster clock, counting edges if required. Or an deliberately introduced rc delay, although that feels a little... soft or imprecise to the software engineer in me (that's my background).

Counting edges from a faster clock is of course valid, but that clock might have to be a lot faster, and you might need quite a few logic parts for the fine-grained counting, like one part in 20 or 30. If they're fast enough, fine. The RxC has an advantage is that it can be continuously variable (with a trimmer) and may not need parts to be as fast. Temperature coefficient, at least using X7R or NP0 capacitors, will be negligible for the application. Going a different way, without counting edges on a faster clock, just adding gates to get their delay is not nearly as precise, since manufacturers don't specify a minimum propagation delay, and the delay varies considerably with temperature and load. They don't specify a minimum input capacitance, either, nor the hysteresis thresholds for Schmitt-trigger inputs, but I don't think the latter will vary substantially over temperature, so once you set the trimmer, things should be reliable. I have a little experience doing digital delays this way, but not a lot. Commercially available digital delay lines are not clocked.

_________________
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?


Top
 Profile  
Reply with quote  
PostPosted: Fri Oct 27, 2017 11:28 pm 
Offline

Joined: Mon Oct 23, 2017 9:16 pm
Posts: 22
What I'm getting from this, more or less, is that best case is use a cpld/fpga (where a bunch of counters are "cheap", and in the fpga case at least, clock multiplication is a common feature) or hope you can work from existing signal edges. I had a Google round commercial delay ICs, they look pretty... hardcore (a delay line needs 16+ pins?! Granted, many are for selecting the desired delay, but still...)


Top
 Profile  
Reply with quote  
PostPosted: Sat Oct 28, 2017 7:52 am 
Offline
User avatar

Joined: Thu Dec 11, 2008 1:28 pm
Posts: 10977
Location: England
It might be that you don't need a delay: can you share your timing diagram? Personally I find it near-impossible to follow textual stories of what's going on, and words like 'gate' and 'latch' are potentially misleading if they mean different things to different people.

What I'm getting at is this: say, for example, the address bus from a CPU becomes valid about 30ns after the fall of Phi2. There's no need to capture this new value at that time, or any other time. If the bus stays static and valid until say 5ns after the following Phi2, then it will be valid for the entire time you need it.

Of course, your situation might be different.


Top
 Profile  
Reply with quote  
PostPosted: Sat Nov 04, 2017 3:06 pm 
Offline

Joined: Mon Oct 23, 2017 9:16 pm
Posts: 22
So, I've been working hard when I can find the time, and I've come to the attached schematic (now presented in glorious black and white!)

It is considerably more involved that before - after looking more carefully at the serial handshaking, I decided to being the UART-USB interfaces on board.

Both the UART and RAM selects are now qualified with PHI2, with /OE driven by PHI1 (now properly generated from the clock circuit). The datasheet puts /OE as don't care during a write cycle.

The memory map and decoding has been re-done. It's possibly somewhat unconventional, but it seems pretty convenient to me, I could even place jumpers on the board to be able to reconfigure the memory map after fabrication. A15 on the RAM is wired to the bus to cover this eventuality.
The main motivation was the realisation that as this board has no provision for mass storage, I should perhaps be more generous with the addressable ROM so I can fit more code in to it.

All unused inputs are properly tied, and I've added a complement of decoupling caps, status LEDs and power in via a micro-usb port so that this design is also proposed as actually complete.

On the other hand, this is now way more than I might be able to test on a breadboard, and will be considerably larger on the PCB, so I'm more concerned than ever about potential bugs!

Speaking of board space, I am really starting to see the advantage of the PLCC package format. I have some PLCC EEPROMs, and they're less than half the size of the identical model DIP. On the other hand, ZIF sockets are hard to come by (for the ROM) and I already have the DIP parts for all the other major ICs. I'll be using SOIC for the 74 series ICs.

Edit: Schematic removed, because it was badly oriented. See download/file.php?id=5671 (comment below) for PDF.


Last edited by mdpye on Sat Nov 04, 2017 4:59 pm, edited 1 time in total.

Top
 Profile  
Reply with quote  
PostPosted: Sat Nov 04, 2017 4:40 pm 
Offline
User avatar

Joined: Tue Oct 25, 2016 8:56 pm
Posts: 362
As I am unable to read sideways circuit schematics, I shall need to print this before I can give any meaningful feedback.

_________________
Want to design a PCB for your project? I strongly recommend KiCad. Its free, its multiplatform, and its easy to learn!
Also, I maintain KiCad libraries of Retro Computing and Arduino components you might find useful.


Top
 Profile  
Reply with quote  
PostPosted: Sat Nov 04, 2017 4:58 pm 
Offline

Joined: Mon Oct 23, 2017 9:16 pm
Posts: 22
I was having some difficulty with my PDF generator, but sorted now.

Here's a correctly oriented copy, where I've also found better symbols for the RAM and ROM ICs (with logical, rather than physical schematic presentation)


Attachments:
main.pdf [194.86 KiB]
Downloaded 128 times
Top
 Profile  
Reply with quote  
PostPosted: Sat Nov 04, 2017 5:33 pm 
Offline
User avatar

Joined: Tue Oct 25, 2016 8:56 pm
Posts: 362
Nope, that was still portrait by default. I have however just discovered that adobe has a rotate view function! woo!

_________________
Want to design a PCB for your project? I strongly recommend KiCad. Its free, its multiplatform, and its easy to learn!
Also, I maintain KiCad libraries of Retro Computing and Arduino components you might find useful.


Top
 Profile  
Reply with quote  
PostPosted: Sat Nov 04, 2017 5:50 pm 
Offline
User avatar

Joined: Tue Oct 25, 2016 8:56 pm
Posts: 362
That is probably one of the more interesting uses for an '85 comparator i've seen, but I see no reason why it wouldn't work.

  • CS_ROM is not gated by PHI2 like the other two are, I would suggest it probably should be. Also as it stands, it will be possible to write to the ROM chip - whilst this is fine, generally speaking the 28C' chips have specific requirements as to write cycles, its not like writing to RAM, so be careful you don't write to the ROM space when you don't mean to.
  • I would recommend a bulk capacitor - say 10uF - on your system power input.
  • If all inputs have weak internal pull-ups as stated, why do DCD and RI need to be pulled connected to Vcc?
  • I feel like your PHI1/2 generator: I would be concerned about the ability of that circuit to drive PHI2 properly, and PHI1 will be a little out of phase with PHI2 due to the delay in the second inverter. Rather than rolling your own crystal circuit, it'd be much easier to chuck in a oscillator can. If you use a 4MHz can and divide it down to 2MHz with a 74HC74, you get a 50% duty cycle clock and you get both PHI1 and PHI2 for free, with no phase difference between them. You can even use the second flip-flop in a 74HC74 package to implement debounce for a reset or NMI button.
  • A suggestion: Have two USBs instead of three; one of them can do power and data. If you power it off a wall-wart then the data lines will do nothing anyway, and if you've plugged it into a PC chances are you want to power AND talk to it, so why use up two USB ports where one will do?

Other than that, it all looks good to me.

_________________
Want to design a PCB for your project? I strongly recommend KiCad. Its free, its multiplatform, and its easy to learn!
Also, I maintain KiCad libraries of Retro Computing and Arduino components you might find useful.


Top
 Profile  
Reply with quote  
PostPosted: Sun Nov 05, 2017 1:20 am 
Offline

Joined: Mon Oct 23, 2017 9:16 pm
Posts: 22
Quote:
CS_ROM is not gated by PHI2 like the other two are, I would suggest it probably should be. Also as it stands, it will be possible to write to the ROM chip - whilst this is fine, generally speaking the 28C' chips have specific requirements as to write cycles, its not like writing to RAM, so be careful you don't write to the ROM space when you don't mean to.

Oops, wiring R/W to the ROM /WE was entirely accidental, introduced when I switched the symbols in too much haste! Without the ability to write to the EEPROM in place, PHI2 gating of the /CS line shouldn't be necessary, right? The ROM is by far the slowest component to respond, and I wanted to give it as much time as possible, in case I am able to push the clock up a little on completion. It was mentioned that the UART will likely limit me to 3MHz, but even that would be a 50% increase in raw power, which would be neat...

Quote:
I would recommend a bulk capacitor - say 10uF - on your system power input.

Thanks. I wondered a little what to strap across the rails there. I'll add a nice big one.

Quote:
If all inputs have weak internal pull-ups as stated, why do DCD and RI need to be pulled connected to Vcc?

Haha, for no good reason, it appears. I think I must have connected them before finding the pull-up note in the data-sheet and neglected to re-visit.

Quote:
I feel like your PHI1/2 generator: I would be concerned about the ability of that circuit to drive PHI2 properly, and PHI1 will be a little out of phase with PHI2 due to the delay in the second inverter. Rather than rolling your own crystal circuit, it'd be much easier to chuck in a oscillator can. If you use a 4MHz can and divide it down to 2MHz with a 74HC74, you get a 50% duty cycle clock and you get both PHI1 and PHI2 for free, with no phase difference between them. You can even use the second flip-flop in a 74HC74 package to implement debounce for a reset or NMI button.

While a can sounds tempting, I do have a lot of bare crystals in a wide variety of frequencies already, so I would prefer to put them to use. But using a faster one and a flipflop would be a sensible move from the sounds of it. And debouncing using the spare unit is a great idea, thanks! That also reminds me that an NMI button for debugging lock-up crashes would be super useful.

Quote:
A suggestion: Have two USBs instead of three; one of them can do power and data. If you power it off a wall-wart then the data lines will do nothing anyway, and if you've plugged it into a PC chances are you want to power AND talk to it, so why use up two USB ports where one will do?

I thought about this. I guess I was a little worried about what the total power draw might be. I can use a decent wall wart, but laptop USB ports are - as I understand it - often somewhat under powered. That said, the unit doesn't have much in the way of useful IO besides a serial connection to a PC, so I will always be using it plugged in to one, and a single cable would be much more convenient!

Quote:
Other than that, it all looks good to me.

Woohoo! Thank you very much for looking over it.


Top
 Profile  
Reply with quote  
PostPosted: Sun Nov 05, 2017 3:39 am 
Online
User avatar

Joined: Fri Aug 30, 2002 1:09 am
Posts: 8540
Location: Southern California
I couldn't find where your GPIOs go (which the "weak" pull-ups are connected to), but if you truly want weak pull-ups, that would be more like a megohm, not 10K. I've had situations where I'd want something with 5 or 10 K of impedance to pull down to a solid logic '0' against a weak I/O, and they wouldn't be able to against a 10K.

_________________
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?


Top
 Profile  
Reply with quote  
PostPosted: Sun Nov 05, 2017 9:56 am 
Offline
User avatar

Joined: Tue Oct 25, 2016 8:56 pm
Posts: 362
mdpye wrote:
Oops, wiring R/W to the ROM /WE was entirely accidental, introduced when I switched the symbols in too much haste! Without the ability to write to the EEPROM in place, PHI2 gating of the /CS line shouldn't be necessary, right? The ROM is by far the slowest component to respond, and I wanted to give it as much time as possible, in case I am able to push the clock up a little on completion. It was mentioned that the UART will likely limit me to 3MHz, but even that would be a 50% increase in raw power, which would be neat...

If it is physically impossible to write to the ROM chip, and it will never become possible, then no, you shouldn't need to gate it by PHI2. However, if you ever intend for your SBC to be able to write to the ROM then you'll need it to be gated. On my own boards, I always have /WE pulled high with a resistor and a jumper that can connect it to the relevant control line; if I should ever feel like allowing the ROM to be written.

mdpye wrote:
While a can sounds tempting, I do have a lot of bare crystals in a wide variety of frequencies already, so I would prefer to put them to use. But using a faster one and a flipflop would be a sensible move from the sounds of it. And debouncing using the spare unit is a great idea, thanks! That also reminds me that an NMI button for debugging lock-up crashes would be super useful.

Yes. If you want to use your crystals, keep your current circuit but connect the output to the flop, and get PHI1 and PHI2 from the flop's output. As to the debouncing, just in case you're not familiar with this method, this is what you're going for... Image.

mdpye wrote:
I thought about this. I guess I was a little worried about what the total power draw might be. I can use a decent wall wart, but laptop USB ports are - as I understand it - often somewhat under powered. That said, the unit doesn't have much in the way of useful IO besides a serial connection to a PC, so I will always be using it plugged in to one, and a single cable would be much more convenient!

That is a fair point, I had not considered current draw. You can only guarentee that a USB port can supply 500mA. Many can supply more, but to be compliant they only need that half an amp. [Also, technically a USB port is only supposed to supply 100mA initially and extra current draw is supposed to be negotiated between the host and the device, but this is widely ignored] If your system runs at 500mA or less then great, otherwise yes it might be best to keep them seperate. Infact it might even be better in that case to use a barrel-jack so you don't accidentally plug it into something that cannot power it in future.

_________________
Want to design a PCB for your project? I strongly recommend KiCad. Its free, its multiplatform, and its easy to learn!
Also, I maintain KiCad libraries of Retro Computing and Arduino components you might find useful.


Top
 Profile  
Reply with quote  
PostPosted: Tue Nov 07, 2017 4:30 pm 
Offline
User avatar

Joined: Fri Dec 11, 2009 3:50 pm
Posts: 3367
Location: Ontario, Canada
Alarm Siren wrote:
If it is physically impossible to write to the ROM chip, and it will never become possible, then no, you shouldn't need to gate it by PHI2.

Hmm, there's a point in question that's more basic than writes to the ROM. The point pertains to transients of bus contention.

We all know that on each cycle there can be only one device driving the data bus, and that a gross violation of this rule will be a show-stopper. What's less well known is that the boundaries between cycles also present the threat of contention. Specifically, it's desirable that the previously active device complete the process of transitioning its outputs to high impedance before the incoming device exits high impedance and commences turning on. This seems almost impossible to guarantee. The enable and disable times of devices on the bus will vary -- and in some cases won't even be fully specified by the manufacturer.

The best -- and easiest -- solution is to observe the convention that nothing drives the data bus when PHI2 is low. (The 6502 chip itself observes this convention.) Thus the PHI2-low time provides a timing cushion during which each of the devices can promptly or lazily complete its part of the handover. Notice it's impossible to transfer data during this time anyway because the address bus isn't yet known to be valid. So there's little or no downside; the worst that can happen is you might need an extra gate.

Admittedly it is possible for a design to work despite transients of bus contention, but tolerating the problem is a decision with certain consequences regarding signal integrity and power consumption. It's also possible to "get lucky." For example you may experimentally find that even the minimum time taken for the (E)EPROM to begin driving the bus is longer than the maximum time taken for the preceding device to relinquish the bus. But the minimum time is often not specified by the manufacturer, in which case this -- like overclocking -- becomes a seat-of-the-pants practice rather than rigorous engineering.

_________________
In 1988 my 65C02 got six new registers and 44 new full-speed instructions!
https://laughtonelectronics.com/Arcana/ ... mmary.html


Top
 Profile  
Reply with quote  
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 49 posts ]  Go to page Previous  1, 2, 3, 4  Next

All times are UTC


Who is online

Users browsing this forum: Google [Bot] and 10 guests


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to: