Looking for fast transistors to build own cmos logic.

dl324

Joined Mar 30, 2015
11,282
Welcome to AAC!

What do you plan to do with this high speed discrete CMOS logic?

You're asking for nothing but trouble. You're going to have problems with device matching and parasitics from interconnects.
 

BobTPH

Joined Jun 5, 2013
2,503
AT 45GHz, the maximum distance a signal can go in one clock cycle is 0.66 cm. I hope what you are building is small.

In ICs that function at 1/10 this speed, clock skew and speed of light delays of signals is a significant problem. With discrete logic it would be a non-starter.

Bob
 

crutschow

Joined Mar 14, 2008
25,267
Well, I'm hoping to dabble in processor designs.
I want to use fast components as saving on ordering extra.
Any advice?
CMOS digital IC's will be a lot cheaper, faster, and use many fewer parts than you can build using discrete transistors.
Why do you want to use transistors instead of IC's?
You will end up with hundreds if not thousands of transistors to build a processor. :eek:
 

Thread Starter

Paul Griffiths

Joined Nov 21, 2017
28
CMOS digital IC's will be a lot cheaper, faster, and use many fewer parts than you can build using discrete transistors.
Why do you want to use transistors instead of IC's?
You will end up with hundreds if not thousands of transistors to build a processor. :eek:
As yet to know everything but believe may save count as some transistors may be used for a few logic math than ic's that can't get to each transistor.
 

Papabravo

Joined Feb 24, 2006
14,244
As yet to know everything but believe may save count as some transistors may be used for a few logic math than ic's that can't get to each transistor.
This is essentially a non-starter. What you want to do is design a processor to be implemented in an FPGA. You'll do the design in Verilog or VHDL and synthesize it inside a single chip. There is NO OTHER PRACTICAL way to do this.
 

Thread Starter

Paul Griffiths

Joined Nov 21, 2017
28
This is essentially a non-starter. What you want to do is design a processor to be implemented in an FPGA. You'll do the design in Verilog or VHDL and synthesize it inside a single chip. There is NO OTHER PRACTICAL way to do this.
FPGA's not fast due to the programming logic's.
 

dl324

Joined Mar 30, 2015
11,282
FPGA's not fast due to the programming logic's.
The way the Big Boys do this is they have simulators that are set up for the target process. You draw the circuit, run parasitic extraction, and simulate circuit performance. Once the circuit is fabricated, you try to find out why it doesn't perform as expected. That's what we did for microprocessors that had hundreds of millions of transistors *before* the process was anywhere near production worthy.

The last design I'm aware of that did an actual breadboard was the 80286. I knew a technician who worked on it.
 

Thread Starter

Paul Griffiths

Joined Nov 21, 2017
28
You made a statement.

"FPGA's not fast due to the programming logic's."

That statement is total nonsense.

The programming logic does not affect the speed of the FPGA and any representation that it does is...well, total nonsense.
Each connected gate requires a programmed part which slows down it's entirety. All part of the game...
Or do you believe there's an intel engineer inside with a soldering iron?
 

Papabravo

Joined Feb 24, 2006
14,244
This is what an old 8-bit microprocessor looks like as transistors.
From https://monster6502.com/
View attachment 139808
This is cool project, but at less than 1/10 speed is mainly a tour-de-force to show it can be done. We did such things with Xilinx chips to simulate custom communication ASICS back in the 1990's. Clock recovery for Manchester encoded data ran at 1/10 th speed using 110 FAST IC's (74F series) and PALS. The rest of the chip was in 56 XC4000 series FPGAs all daisy chained together for programming. The power up delay for the programming was hardly noticeable. The 1st silicon of the part worked as expected and went directly into production.
 

absf

Joined Dec 29, 2010
1,947
Paul Griffiths said:
I had a BBC model B:
SOUND 10, -15, 128, 1000 :)
It is the BASIC statement used in BBC BASIC:

The format of the SOUND statement is:

SOUND C,A,P,D
where

C is the channel number.
A is the amplitude (volume) of the note.
P is the pitch (frequency) of the note.
D is the duration (length) of the note.
 

Papabravo

Joined Feb 24, 2006
14,244
I had a BBC model B:
SOUND 10, -15, 128, 1000


It has been a long time (maybe half a century) since I had any need for BASIC, so excuse me for not recognizing it or understanding it's significance to the discussion, which was building a computer out of discrete transistors. As the designers of the MONSTER 6502 pointed out you cannot buy the discrete transistors such a design requires, and you can't do anything about the massive amount of gate capacitance you need to overcome.
 
Top