0

Would the old series 4000 CMOS logic series have evolved over time?

The manufacturers surely have bought new machines and are not using the same machines from some odd 40 years back.

Would they have adapted the 4000 logic series to a smaller scale (nanometre) transistors / FETs / diodes / Zeners and what not to reduce power consume or are they still using the same masks?

JRE
  • 71,321
  • 10
  • 107
  • 188
NaturalDemon
  • 327
  • 1
  • 9
  • 1
    The Manufacturers surely have bought new machines and are not using the same machines from some odd 40 years back. Are you sure? What if the old equipment still works and only requires maintenance? Don't underestimate the fact that written-off equipment could be much cheaper than buying new equipment. 4000 series logic consumes almost no power so why would there be a need for even lower power? Don't underestimate the cost of making a new design. And for "jelly bean" chips like 400 series logic price is all that matters. All your suggestions potentially increase cost. – Bimpelrekkie Oct 22 '21 at 14:06
  • 1
    And "newer" CMOS logic already exists: 74HC and 74AC and LVT: https://en.wikipedia.org/wiki/LVCMOS series logic. Just an example: https://www.ti.com/lit/ds/symlink/sn74aup2g00.pdf?ts=1634911885202&ref_url=https%253A%252F%252Fwww.ti.com%252Fproduct%252FSN74AUP2G00 – Bimpelrekkie Oct 22 '21 at 14:09
  • @Bimpelrekkie, just wondering if they also chased the nano meter scale race. i know some factories burned down, like a Harris plant. Yeah, new machines seam to cost like 150 million @ ASML. well they don't have the invent the chip again, just reduce it's size. once i Asked NXP to produce a certain type of chip and at first the refused, but later i discovered they did went to make them. – NaturalDemon Oct 22 '21 at 14:10
  • @Bimpelrekkie, but these are for 3v333 or 5v operations, than i need t scale down to use them and later scale back up again. – NaturalDemon Oct 22 '21 at 14:11
  • 1
    Just wondering if they also chased the nano meter scale race NO because these logic chips aren't about cramming as much transistors as possible into a chip. The challenges with these logic chips are: price, price (yeah again), logistics (can a manufacturer deliver as promised?). There is No point in making a chip smaller when its size is already limited by the size of its bondingpads. So if you make the transistors smaller, you cannot make the chip any smaller. You only increase wafer cost. – Bimpelrekkie Oct 22 '21 at 14:15
  • but these are for 3v333 or 5v operations, Yep because that's the result of using a newer process. Anything "new" runs on 5 V, 3.3 V or even 1.8 V. If you need logic at 12 V: use 4000 series. It is still made and it serves a purpose. – Bimpelrekkie Oct 22 '21 at 14:17

3 Answers3

6

Usually, IC's built 40 years apart with the exact same part number have the exact same circuit inside (sometimes, something very close). This is because reverse-compatibility is critical. An LM358 built last month has the same high noise and crossover distortion as one built in the 1970's. And a CD4011 will be just as slow as it used to was.

Many old designs will fail with a newer part that is 10x faster because that faster output edge can create several different types of noise, including inducing ringing in downstream circuits that can be interpreted as false pulses.

Also, a shorter propagation delay could cause all kinds of timing problems, including a complete lockup of the circuit.

Yes, the series has evolved over time into dozens upgraded variants. No, the original part numbers have not evolved.

AnalogKid
  • 19,998
  • 1
  • 13
  • 32
6

One practical difficulty is that modern logic processes can't tolerate the 12V supply voltage (and usually 18V abs max rating) for classic 4000 series devices.

Anything breaks down (including silicon dioxide and other insulators) in a sufficiently high electric field (measured in volts/metre), and as you scale features down, you find you can get a lot of volts/metre across a micron, let alone 14 nM.

It seems like only yesterday to an old codger like me, that 3.3V and now 2.5 and 1.8V, took over from 5V as the supply rails for general purpose logic, because the I/O transistors could no longer support voltages as high as 5V ... let alone 12V.

And of course the core in a CPU or high end FPGA now runs on a fraction of a volt, with only especially large and beefy transistors around the I/O pads, capable of tolerating 2.5V or so. Newer FPGA families stopped supporting 5V I/O about ten years ago.

I believe that CD4000 was designed around a 10um or 15um gate width, easily achieved with what was basically optical microscope technology used in reverse...

  • 1
    I like this answer better, so far. +1. I'd add, without necessarily saying any particular 4000 series device qualifies for such use, that the large feature sizes of devices such as these also provide a possibility for more reliable rad-hard use (nuclear disaster robots for Fukushima or for use in vehicles designed for use in space, for example.) Since these aren't often "on a budget" they may not use the 4000 series devices but instead have custom-designed ICs (which themselves aren't overly expensive anymore as they use old-gen Fabs and processes and brokering is so readily available now.) – jonk Oct 22 '21 at 19:34
  • @jonk I'm guessing old generation fabs are running flat out just now, catching up with trailing edge semiconductors for the gasoline auto industry. I've seen some stories about how far behind Tesla the tech they use is ... nobody's saying, but it's possible that tolerance to the voltage spikes you get around a spark ignition engine makes redesign using more delicate parts a bit of a challenge. –  Oct 22 '21 at 20:42
  • 1
    That's also a good point, I suppose. And automotive can afford (today, though not "back in the day") to have custom IC designs fabbed. Heck. *I* can afford to have custom ICs fabbed. As a hobbyist, for gosh sake. It's getting down-right cheap. Better, there is FOSSi and on a 130 nm SkyWater node! If only I'd could be 20, again!! What a heaven I would live in. (Well, except for all the rest that happened in transitioning from 2 billion people on Earth to 8 billion+. There is that.) – jonk Oct 22 '21 at 21:23
  • @jonk, http://www.thatcorp.com, http://www.soundsemiconductor.com, That corp offers this service to help you. http://www.thatcorp.com/THAT_Semiconductor_Fabrication.shtml – NaturalDemon Oct 23 '21 at 09:41
  • @jonk there is a guy on YouTube constructing ic's, he has a manual bonding machines and what not

    Upgraded Homemade Silicon Chips ( published 2 months ago ) https://www.youtube.com/watch?v=IS5ycm7VfXg ... using Abode Photoshop

     – NaturalDemon Oct 23 '21 at 09:49
  • hmmm, i just found out, there is some web site indexing this site. – NaturalDemon Oct 23 '21 at 10:27
  • @NaturalDemon Those links look like they want money! FOSSi doesn't seem to be asking for cash if the work is open source. Which is good for hobbyists! In any case, I'll try and watch the video when I get a moment. Thanks! – jonk Oct 23 '21 at 10:34
  • @jonk, i did look at the site, but didn't read with full attention and i didn't see any downloads on how to design chips or masks. – NaturalDemon Oct 23 '21 at 12:20
1

They have evolved. They are still made on a high voltage CMOS process, but the process is much better controlled for parameters. Modern 4000-series parts still made - basically those from TI - all handily beat their speed specifications from old datasheets and the old jedec standard. Usually their performance on the low end of the supply range is especially improved vs. the datasheet. They are much better parts although still fundamentally “same”. Semiconductor manufacturing processes nowadays have quite tight parameter control vs. what was possible 40-50 years ago.

I use the 4000 series extensively in retro computing projects and while they are not as fast as TTL, they can come within a small integer factor sometimes.

Kuba hasn't forgotten Monica
  • 40,960
  • 1
  • 39
  • 120