# 8 Bit, 8 Bit Multiplier

#### Steven Stanley Bayes

Joined Dec 23, 2014
4
I posted this in another thread and someone replied the thread is very old. I post this now also to state : threads are here to stay for everyone to see and use whenever they want.

Another reason is : there have been many questions in regards to this as well as related to. Some answers suggest ROM addressing ( look up tables ) which is NOT a good idea unless there are preprogrammed ones for 8 bit, 8 bit multiplication which I have not seen. Also, the standard approach may as well prove to be much faster. So, here is the standard approach as posted in another thread :

http://en.wikipedia.org/wiki/List_of_7400_series_integrated_circuits : 74558 8 bit by 8 bit multiplier

#### tshuck

Joined Oct 18, 2012
3,527
Yes, they are there for others to view, but the thread belongs to the thread starter and you asking your own question on that that derails the thread.

There are many was to do multiplication, it's a matter of tradeoffs to accomplish the specifications. Any wide, sweeping statement is bound to fall short.

#### WBahn

Joined Mar 31, 2012
24,564
I posted this in another thread and someone replied the thread is very old. I post this now also to state : threads are here to stay for everyone to see and use whenever they want.

Another reason is : there have been many questions in regards to this as well as related to. Some answers suggest ROM addressing ( look up tables ) which is NOT a good idea unless there are preprogrammed ones for 8 bit, 8 bit multiplication which I have not seen. Also, the standard approach may as well prove to be much faster. So, here is the standard approach as posted in another thread :

http://en.wikipedia.org/wiki/List_of_7400_series_integrated_circuits : 74558 8 bit by 8 bit multiplier
So you are saying that the "standard" way is to use an obsolete MSI IC?

Why do you say that this might be much faster than using a look-up table?

What is your metric for deciding which approach is best? Speed? Power? Space? Cost?

#### kubeek

Joined Sep 20, 2005
5,587
I couldnt even find the datasheet of the 74558. What I found instead was many different approaches to multiplication. These range on one side from a hard-wired lookup table that uses a vast a mount of chip area but will be very fast, to the other side of the spectrum being a have a shift-add multiplier, which takes less space but will be very slow beacause of long path delays from input to output. Then you have a lot of different variation in things between.

You really need to define what the use case is and then you can choose the appropriate circuit. It simply is not possible to make a general statement which is best.

#### Steven Stanley Bayes

Joined Dec 23, 2014
4

Threads are for everyone to post relevant things regardless who have started them.

The standard way is to use multipliers for multiplication because this is their purpose.

Using ROM's is more historic than using multipliers as they predate these.

Digital IC's would have a response in the sub nanosecond range.

8 bit, 8 bit multipliers MUST be available as well as 16 bit, 16 bit and any other bitness as they are immensely useful. One of the reason is their fast response.

They may not be manufactured by the general purpose manufacturers but must be manufactured elsewhere.

Someone listed areas where multipliers would be advantageous. The answer is : all areas mentioned are better with multipliers.

The applications are endless. One of the applications I found is for power monitoring. Instead of slow, programmable or configurable IC systems, fast ADC's can be used, then multipliers, than, in case needed, fast DAC's. Fast ADC's can be accomplished by fast comparators. 256 of these is a lot of circuitry BUT the speed would be in sub microsecond range. Fast multipliers would not add to the ADC's and DAC's response . Without a DAC, the power measurement speed is defined by the speed of the ADC's which can be much faster than the fastest fuse and, theoretically, temperature independent, practically, the slight temperature dependence can be compensated for to a good deal of extent.

Power monitoring with these speeds can even react to spikes.

Linear multipliers, such as Maxim 4210, rely on the inaccurate and temperature dependent non linear VI characteristic of a diode. Temperature compensation is possible to an extent yet, the best Maxim and anyone has ever been able to achieve is 1.5% which is OK in case true but can be even better. Maxim 4210, however, is much simpler to use than ADC's and multipliers.

In addition, no power is consumed when multipliers are not used which is true for ROM's too although multipliers should consume much less. This is useful for triggered use after certain parameters, say voltage and current, are higher than a given value.

All 74 series must be manufactured and all or most are by manufacturers elsewhere. 4 bit, 4 bit multipliers, for example are largely available. There is no reason to have 4, 4 and not 8, 8; 16, 16, etcetera.

Out of all, speed is the most important advantage. This advantage is applicable when fast calculations are necessary which processors cannot achieve. Digital stress evaluation is another example. In some cases, analogue computation is still used because of the slow performance of processors and clock driven systems. In other cases, they record analogue parameters and then measure them slowly. This is only possible when no immediate action nor immediate measurements are possible.

The biggest advantage of 74 series over the present conventional approaches is speed and speed is still necessary in various applications where present conventional devices cannot perform.

#### Steven Stanley Bayes

Joined Dec 23, 2014
4
Also, the more posts and the more demand, the bigger the chance for whatever 74 IC to be made again.

Old parts are also available at huge prices from some distributors. Some conventional manufacturers may occasionally reopen a line.

A simple digital system made with 74's can be unthinkably faster than the present conventional approaches. Such speed may be required for LSD's, laser diodes, super fast communications, super fast control, etcetera.

#### #12

Joined Nov 30, 2010
18,076
Threads are for everyone to post relevant things regardless who have started them.
If you act on this belief, you will find opposition from the people that own this website.

#### WBahn

Joined Mar 31, 2012
24,564
Threads are for everyone to post relevant things regardless who have started them.
Says who? You? You do not own this website, therefore you do not set the rules for this website. Sorry, but that's life.

The standard way is to use multipliers for multiplication because this is their purpose.
That is ONE way. There are others. Each with it's advantages and disadvantages. Which method is used depends on the relative weight given to those advantages and disadvantages.

Digital IC's would have a response in the sub nanosecond range.
Really? Please provide the part number of a non-LUT based 8-bit by 8-bit multiplier IC with a subnanosecond response.

Someone listed areas where multipliers would be advantageous. The answer is : all areas mentioned are better with multipliers.
Then why aren't they used in ALL applications, since you claim they are BETTER in ALL areas? I can think of one very specific example from my own past -- the first design I ever did professionally was an electronic radon level monitor. I used an 8-bit PIC that had no multiplication capabilities even though I needed to be able to perform 24-bit integer multiplication and division. I did this in software. You are claiming that the better solution would have been to add a dedicated multiplier IC to the design, thereby increasing the cost of a product that was intended to retail for under \$20 when doing it in software added no per-unit cost at all.

Maxim 4210, however, is much simpler to use than ADC's and multipliers.
Now you are contradicting yourself. You've claimed that the use of a multiplier is better in ALL areas, and yet here you give an example of where they obviously is not necessarily the case.

All 74 series must be manufactured and all or most are by manufacturers elsewhere. 4 bit, 4 bit multipliers, for example are largely available. There is no reason to have 4, 4 and not 8, 8; 16, 16, etcetera.
Upon what basis to you claim that all 74 series parts must manufactured? Is there some international law that mandates this? Please provide a reference to back up this claim.

Out of all, speed is the most important advantage.
Bull. What is most important depends on the application. In my radon unit I was clocking the processor at 32kHz and I had literally hours to perform a computation -- I didn't give a damn about speed. I cared about power (the entire unit, including the analog amplifiers for the charged particle detector, had to consume less than 900uA), I cared about cost, and I cared about manufacturability. I would have been happy to clock the thing at 32Hz had that been possible without increasing the cost.

This advantage is applicable when fast calculations are necessary which processors cannot achieve.
Now you are contradicting yourself again. You say that speed is the most important advantage but then by giving an example of when it is application, you are acknowledging that there are situations in which it isn't.

#### WBahn

Joined Mar 31, 2012
24,564
Also, the more posts and the more demand, the bigger the chance for whatever 74 IC to be made again.
ROTFLMAO!

Do you REALLY think that ANY semiconductor manufacturer is going to make a decision about whether to resurrect an obsolete part based on the number of posts on a web forum?

A simple digital system made with 74's can be unthinkably faster than the present conventional approaches. Such speed may be required for LSD's, laser diodes, super fast communications, super fast control, etcetera.
Nonsense. The very fact that a design is using SSI and MSI chips on a board places severe limits on the achievable system speed compared to any of a number of single-chip approaches.

#### kubeek

Joined Sep 20, 2005
5,587
The biggest advantage of 74 series over the present conventional approaches is speed and speed is still necessary in various applications where present conventional devices cannot perform.
I don´t know what you THINK is the conventional approach, but I would say it is either hardwired inside a chip such as microcontroller, or programmed inside FPGA. Both will be miles faster than any 74xx solution.