CMOS compatible SRAM

[GlennSmith]

Hi all,
Sorry if this is a duplicate, I *have* spent time searching - to no avail.

I'm looking for part references of CMOS compatible SRAM, preferably in 32kx8 or 64kx8 formats. Preferably DIP but I do SMT also.
*AND* available from Farnell or Mouser - the only viable sources here in Europe (thx Brexit).
Pretty-much everything available is "TTL compatible", with Vhigh quoted as 2.4v minimum (but never an upper value) Some datasheets are actually quite clear (like Infineon) that the chips are unsuitable for CMOS applications...

Any ideas of other solutions? Level shifting?

Thanks in advance

[Martin A]

My "Solution" was to use alliance memory chips and interpolate for the blank spot on the datasheet! They're CMOS devices themselves, with a wide voltage range, they're not going to have the limited drive of TTL.

If the drivers can output a guaranteed 2.4v** from a 2.7v supply, and typically do 3v from a 3.3v supply. That's basically saying the high output is at least 90% of the supply voltage. CMOS devices typically ask for a minimum 70% of supply as a high.

For "homebrew" applications try it, and if it works use it, is perfectly acceptable, I tried them, they worked for what I wanted at both 3.3v and 5v. You'll see folk on these forums pushing parts way past the published specs, because it works for them, in that particular application.

** NB if you run the WDC version of the 65C02 off 3.3v then that 2.4v minimum is within the CMOS spec .

[GlennSmith]

Thanks for the input. I continued ploughing through the posts that talk about SRAM, and reading people's schemas, - and I came to the conclusion that if it work's for others, it aught to work for me.

Yes, I'm using WDC 14MHz parts, I target a system running at 12MHz (edit: at 5v). I'll do another search at Farnell and see what comes-up in prices that I can afford to buy 1-offs and try and discard if they don't work for me.

I feel that the forum is rather a victim of it's success - there is a *LOT* of very useful data out there. Often while looking for something specific I stumble upon a gem of unrelated information and I think "cool, I must come back and read this...", only to find that I can't locate it afterwards. The best maze/adventure game ever : at least you don't get killed by trolls!

Request to our honorable moderators, would it be possible to have a sticky topic about modern memory parts (or Mfrs) that reliably work with the different CMOS versions of the 65Cxxx chips? (another edit: or at least a (black)list of those that aren't very reliable.)

Thanks to all.

[GARTHWILSON]

GlennSmith wrote:

Request to our honorable moderators, would it be possible to have a sticky topic about modern memory parts (or Mfrs) that reliably work with the different CMOS versions of the 65Cxxx chips? (another edit: or at least a (black)list of those that aren't very reliable.)

Some forums have loads of stickies at the top of each section's index page, making it take an unpleasant amount of scrolling to get down to the new material with every visit.  Mike, the forum owner here, and I, the moderator, want to be careful that we don't end up that way; so we're pretty careful about what gets "sticky'ed."  Rather than focus on just the narrow subject of SRAM, what additional parts could we include in the same category, ones that could commonly get used with 5V 65xx parts?  That might be worthy of consideration.

[GlennSmith]

GARTHWILSON wrote:

GlennSmith wrote:

Request to our honorable moderators, would it be possible to have a sticky topic about modern memory parts (or Mfrs) that reliably work with the different CMOS versions of the 65Cxxx chips? (another edit: or at least a (black)list of those that aren't very reliable.)

what additional parts could we include in the same category, ones that could commonly get used with 5V 65xx parts?  That might be worthy of consideration.

I instantly think of all of the families of 74xx logic - e.g. LS, HC, HCT, AC, ACT and even the BiMos variants, and also the possible use (or not) of other CMOS families like the original 4000 series. Your site(s) also have a lot of useful info [big thanks] - but again it's often by theme. Perhaps some of the articles could be assembled and condensed into one article that describes the why's and where's of mixing (or not) the different technologies.

Also, because the 6502 (and the 6502 sites) have been around for so long, lots of technologies that are still viable have just simply disappeared for economic reasons - leaving us with problems that shouldn't really exist. I also do a lot of work with valve (tube) technology, because I fell in when I was small (nod to a famous French cartoon series), and that also has become weird because certain tubes have magically 'reappeared', but it's now almost impossible to find the appropriate O/P matching transformers or other HV components.

If I can help write drafts, or proofread, or something - I'd be very pleased to do so.

[Dr Jefyll]

GlennSmith wrote:

Some datasheets are actually quite clear (like Infineon) that the chips are unsuitable for CMOS applications...

It's true some RAM datasheets explicitly state that output levels from the RAM are unsuitable for driving typical CMOS inputs (such as the inputs found on all modern WDC processors; also the inputs of logic families such as 74HC, 74AC, 74AHC).

Just be aware that most modern RAMs are equally unsuitable even though their datasheets lack an explicit warning. What's important is the numbers contained in the electrical specs... not whether or not the manufacturer chooses to additionally include an explicit warning written in prose.

Martin A wrote:

For "homebrew" applications try it, and if it works use it, is perfectly acceptable

I am sympathetic to this outlook, and I've occasionally gotten away with "breaking the rules" myself. However I do NOT consider it perfectly acceptable, as there's a serious loss of noise immunity. Even hobby projects occasionally fall victim to this loss of noise immunity. Thankfully that's uncommon, but OTOH it has to be said that noise problems can be an absolute b*tch to recognize and resolve.

For more on noise immunity and CMOS (in-)compatibility, see my thread TTL Compatible... NOT! ( WDC ).

Martin A wrote:

If the drivers can output a guaranteed 2.4v** from a 2.7v supply, and typically do 3v from a 3.3v supply. That's basically saying the high output is at least 90% of the supply voltage.

Things are not always as simple as they seem. I haven't specifically examined datasheets from Alliances, Martin. But as a general trend, it appears as though many modern manufacturers of 5 volt RAM's are actually selling something else. What they seemingly produce are 3 volt RAM's that each include an internal regulator to reduce the incoming supply voltage from 5 down to approximately 3. This preserves TTL compatibility and it saves them money due to the reduced die area of lower-voltage geometries.

And it undermines the point you're making about output levels being proportional to the supply voltage. Although the voltage-regulator theory is presently unsubstantiated, the point about proportional voltage swings gets raised (and found questionable) more than once in the linked thread.

-- Jeff

[okwatts]

I remember this thread ( viewtopic.php?f=12&t=7371&start=150 )where at least 2 of us had trouble with the Alliance rams in our machines that used WDC 65C02 as the CPU. The issue for me is still unresolved except if I use the Hitachi ram. Hope you have better luck if you go that route.

[plasmo]

In that thread is a datasheet of CY7C1009D posted by Dr J that says “90nm” technology. You can’t run 90nm at 5.0V, so clearly there is an internal regulator for RAM array and possibly complexity related to voltage translation from internal core of probably 2.5V to 5V IO drivers. Fast RAM is troublesome in many ways.
Bill

[GlennSmith]

Dr Jefyll wrote:

... see my thread TTL Compatible... NOT! ( WDC )

Thanks! That was one of the gems I hadn't been able to relocate 'cos I couldn' remember what the topic was called, nor who posted it.

[BigEd]

We have in the past put together thread index posts - maybe such a one might be useful as a sticky, but even if not, you can search for posts with "index" in the subject, and there are not so many.

Edit: oh, maybe only twice!

[GlennSmith]

To finish-off this diagram, from the TI document "Selecting the Right Level-Translation Solution" posted by Dr J, shows the trend :

Digital Switching Levels Compared

... and from the same document, this little snippet :

Dont use pullups for CMOS outputs...

I've ordered a few different memory ICs, more news when my WBC is wired-up.

[Martin A]

Time for some examples I guess.

viewtopic.php?f=3&t=5728 It the write up of the Atom build I did, you'll note from the first board there's an Alliance AS6C62256 32k ram on there with an older 10MHz W65C02. The date code of the ram isn't clear, but it's 1127.

I've put the older video card on the backplane, with nothing else fitted and checked the outputs on a scope. In this case the memory is the smaller AS6C6264 8k device, date code 1249. Slightly newer but not ecaxtly new.

AS6C6264 date code 1249 D0 output

That't driving to the rail.

For reference the MC6847 doesn't drive quite so hard, but still a respectable 4v+

A0 driven by MC6847

That's another device that only guaranteed 2.4v highs on the address lines.

The PSU being lightly loaded with just the video board was otputting a little over 5v.

I come from a Z80 background, and the CMSO Z80 is TTL compatible, when it came to buiding the Atom build, the only ram chips "in stock" were the AS6C type. To be honest I didn't even check the specs at that time, I just used what I had.

That said neither board pushes the ram past it's limits in terms of loading, access times etc. the signals do have to to estsblish themselves. The CY7C and even newer AS6C may be totally different construction and might give different results.

The chips I have, they've not let me down so far.

[barnacle]

I have to confess to a slight 'slap it together and see if it ticks' attitude sometimes, which is a bit silly really as part of the day job is hunting down replacement parts to replace those becoming unavailable. I spend *far* too much time reading/comparing datasheets...

Neil

[fachat]

GlennSmith wrote:

Hi all,
Sorry if this is a duplicate, I *have* spent time searching - to no avail.

I'm looking for part references of CMOS compatible SRAM, preferably in 32kx8 or 64kx8 formats. Preferably DIP but I do SMT also.
*AND* available from Farnell or Mouser - the only viable sources here in Europe (thx Brexit).
Pretty-much everything available is "TTL compatible", with Vhigh quoted as 2.4v minimum (but never an upper value) Some datasheets are actually quite clear (like Infineon) that the chips are unsuitable for CMOS applications...

Any ideas of other solutions? Level shifting?

Thanks in advance

That is interesting. I've never seen any such notice before.

What is important is to consider the direction of the signal. A Vhigh value says nothing if you don't know if it is VhighIn or VhighOut.

What you have to make sure is that VhighOut of the driving chip is > VhighIn of the receiver. On bidirectional lines like data bus, this condition needs to hold in both directions.

Personally I always assumed that 'TTL compatible CMOS' devices would have TTL compatible inputs i.e. VhighIn 2.4V but CMOS outputs, i.e. driving close to the rail. I've never had problems with CMOS SRAM in 5V TTL designs except when I gated CE with Phi2 on one brand causing supply voltage issues... I am pretty sure I was deeply looking into the datasheets when I analyzed this, but I could have overlooked it or more modern chips have changed.

Due you have a specific link to a datasheet with such a warning?

Edit: I looked up the CY chip mentioned above and, lo and behold, there is this warning!
I even looked up what I recently used (https://www.mouser.de/datasheet/2/12/AS ... 265427.pdf ) and while it doesn't have the warning, it also only outputs 2.4V...
Thinking about my designs I think I was always conservative assuming TTL level anyway. The ones where I went 5V CMOS were mostly the W65816 reading from the RAM (and their datasheets are inconclusive...) or a CPLD that was 3.3V anyway and only 5V tolerant.
Good catch!

[Dr Jefyll]

fachat wrote:

Personally I always assumed that 'TTL compatible CMOS' devices would have TTL compatible inputs i.e. VhighIn 2.4V but CMOS outputs, i.e. driving close to the rail.

That used to be my assumption also... but unfortunately it isn't valid anymore. We old-timers need to adjust our thinking.

I addressed another old-timer in this post (which is part of the TTL Compatible... NOT! thread):

I wrote:

This is not CMOS like we had in the old days. So, correct me please if I've mistaken your point, but if you're saying that any CMOS chip is automatically OK voltage-level-wise to drive other CMOS, I can't agree. There's definitely been a change since you & I entered the biz.

There's more, but I won't quote the entire post here.

-- Jeff