Substituting 27c16 with 57c45, how to?

[neazoi]

Hello,
I would like to replace a 27c16 eprom in a digital circuit with a 57c45 one.
I attach you a screenshot of the circuit that uses the 27c16.
For your convenience I also attach the datasheets.

The data is obvious, D0-D7 in both eproms.
The address seems obvious too, use only A0-A5 of the 57c45 and ground the rest of adress lines.

What is confusing are the other pins /CS, PGM, VPP on the 27c16. How should I connect the 57c45?

 

[betwixt]

I think you might have problems with this device, obviously it has greater storage capacity but it also has other features which your original circuit may not cater for. Primarily, it has a latched output although the data sheet suggests it may be possible to bypass it. Without seeing the schematic you are using, I would suggest you tie PGM and VPP to +5V and connect /CS to /OE.

Note that the programming requirements of the two devices are completely different, 2716 needs +25V VPP and 57C45 needs +13.5V and the programming algorithms are different, you can't just plug one into the programmer unit instead of the other. Do you have a suitable programmer for the 57C45 before going further?

It would be much simpler to use a larger but compatible device such as 2732, 2764, 27128 and so on but they have more pins so they won't fit straight in to the same socket.

Brian.

 

[neazoi]

Thanks Brian, this is what I am trying to make http://neazoi.com/mycomputer/COMPLETE/#Clock_and_micro-instruction_logic

I have a suitable programmer for the new eprom, but it would be interesting to try the manual programmer as well at the end of the page, by altering the programming voltage.

Could you guess, what are these features you refer to by seeing the schematic?Is it just enough to tie PGM and VPP to +5V and connect /CS to /OE or I need to do something else?

PS. the purpose I would like to use the newer eprom is speed, I have one with 25ns.
Of course I have thought of this issue for the other chips in the circuit as well, the page is not complete, so do not judge from the schematics...

Another thing to notice is that both 2716 eproms are connected in parallel, if I could find a 16-bit wide eprom I could use only one. Any thoughts of a suitable eprom type?

 

[betwixt]

I don't think speed is an issue here, in fact the glitch problem highlighted in the text will be worse with a faster device. The actual micro-instruction sequencing is timed by the RC clock in the top left corner of the schematic and it's speed is very low.

In principle, yes you can use a 16-bit EPROM but old style ones are difficult to source and are normally in 40-pin DIP packages so the layout would be significantly different. The only one I can think of is the AT27C1024 but it's an OTP device so you don't get a second chance if it doesn't program correctly. You might be able to use PAL devices but again these need a different programming technique.

Brian.

 

[betwixt]

I've just cleared out a lot of cobwebs and made several spiders homeless. Underneath them I found a box of EPROMS of various sizes ranging from 2708 to 27C2000 and amongst them were two 2716s. If you are trying to work around not being able to source them, you are welcome to the ones I have here. I doubt I will ever need them, most of my work is in microcontrollers these days and most of them have more memory than a 2716 on board already!

Brian.

 

[neazoi]

Thank you very much for your kind offer Brian. The problem is not sourcing the eproms but speed as I mentioned. The top clock according to the datasheet can go up to 20MHz and It would be interesting to test how the computer will behave in higher frequencies, since I have managed to get 54F series chips for the circuit. However I think you are right about the glitch problem, but this is due to the eprom used and may not apply to the faster eprom, this is something I need to try.
I think I will make the original version on the PCB and later on try modifications to accept the newer eprom, based on the pin connections you indicated.

My final idea is to completely replace the two eproms with a 1n4148 16-bit diode matrix and a address decoder out of a few demux chips. That is true ttl but I think the speed can never go higher than a 200ns eprom, even if using F series chips for the decoder.
Based on your experience, what do you think about it?

 

[betwixt]

'F' series devices may not make it any faster and potentially can cause 'race' situations. The timing as a whole has to be considered and reducing propagation delays in any single part may skew the timing with other parts.

You might be able to get 200nS from a diode array if you keep the current fairly high but that sacrifices power and logic thresholds. I'm not sure which clock you want to push for 20MHz but if it's the 74624, please remember that this is the sequencer for each cycle of each instruction so the actual execution rate will be much slower.

Brian.

 

[neazoi]

This is a bit out of topic but how should I connect a 27C512 to the circuit?

I have found some ST M27C512-45XF3 at 45ns, not bad!

 

[betwixt]

Connect /E and /G (VPP) to ground. All the unused address lines should be tied low (or to ground) and it should then work like an 8-bit 64 address memory. In the instruction sequencer the outputs are always enabled so it may give a slightly faster than 45nS response. The figure quoted in the data sheets are measured from the /E pin going low to the data becoming available on the 'Q' pins and in your case it is already enabled so the only delay is in the internal address decoder.

You have almost re-invented the Intel 4004 processor!

Brian.

 

[neazoi]

Thanks Brian i will follow your suggestions and design the PCB that accordingly.

I wish the core was my design, but at least I have already made some modifications and I will do more, so hopefully at the end I will take some credit for this as well
I wish to make the project very educational, it educated me in understating many things of how computers actually work!
Thanks again for you very informative help!

 

[betwixt]

Good luck!

I started working with logic circuits in the mid 1970s and most 'computers' of that time used the kind of technology you are working with. It gives you a good understanding of what goes on inside a microprocessor and why they work. I see too many students these days who just assume you plug a mouse and keyboard into a box and HD graphics appear by magic. Those of us who understand their inner workings are rare and I fear, a dying breed!

Brian.

 

[neazoi]

I am younger, but I understand why you are absolutely right.
Even universities (even in UK, where I had my degree) have moved to a higher level of things because they must cope with the new technology fast. There are only the basic digital electronics lessons who teach students about primitive stuff like flip-flops and gates etc. Students interested in digital electronics, usually work with prepackaged microcontrollers writing just software (in C...), where in the best cases they design something on fpga using vhdl. There are even fewer of these students that design their own processors as parts of their final year projects and when the final open-day presentation comes, they take the risk of being asked "why didn't you do that on a modern fpga?".

Things are moving fast and we are trying to catch. I think that is the reason why. But those who know the primitive stuff, will always know the difference between building/understanding something or just using it. That does not apply only in electronics.

That's a bit philosophical, but I hope someday my/our children (flooded by the vast amount of information of the touch/smart phones) will understand this. People in your age, can teach how these things actually work, as you have worked with these, whereas people in my age are just in the edge or a bit too late. That is why these forums are so important; listen, talk and learn from older ones...