Author Topic: Engineering Students being taught DSP on 20 year old processors RMIT  (Read 29905 times)

0 Members and 1 Guest are viewing this topic.

Online Benta

  • Super Contributor
  • ***
  • Posts: 5839
  • Country: de
Re: Engineering Students being taught DSP on 20 year old processors RMIT
« Reply #50 on: August 29, 2017, 09:03:46 pm »
Your head-banging on working with assembly code is also beside the point. Almost all the signal-processing code parts are done in assembly on DSPs, otherwise no time-critical analysis/prediction is possible. Housekeeping tasks like init, I/O etc. are mostly done in C.
That is another misconception. Who ever said signal processing needs to be fast and/or time critical? Like only signal processing can be time critical?
Are you trolling? I never said that. But a reason for using a DSP is real-time signal processing. Otherwise, just GOTO any micro and a BASIC interpreter.
I just want to point out there is nothing magical or complex to signal processing. A lot of people seem to think digital signal processing ALWAYS needs special processors, FPGAs, assembly code, lots of complexity, etc. Nowadays you only need a DSP if you need to process lots of data very fast with a limited power budget. In opther words: nowadays DSPs only serve a niche market because PCs, microcontrollers and SoCs have lots of processing power.

Yes, I heard that argument 25 years ago as well. In those days it was about "soft modems" that didn't need any hardware other than AD/DA converters to run a 14400 line, by just using the "incredible processing power" of a 486.

The basis of that argument was and is "just throw some more processing power at it". Extremely primitive.

As I wrote, most DSPs are hidden nowadays, but crucial for their applications. And I stand by my opinion that the DSP56800 is a good choice for education.

 
The following users thanked this post: Someone

Offline ehughes

  • Frequent Contributor
  • **
  • Posts: 409
  • Country: us
Re: Engineering Students being taught DSP on 20 year old processors RMIT
« Reply #51 on: August 29, 2017, 09:07:55 pm »
DSP is applied discrete mathematics.   You don't need a chip stamped "DSP" on top to do it. All algorithms used in DSP eventually funnel down to computation of an inner-product space.   I.E. multiply and add.

I could teach a course on DSP using an 8051, or an abacus.    Just because there is a new fancy processor that automates an algorithm doesn't invalidate the core knowledge base required.

Processing atmospheric infrasonic signals with wavenumbers << 1 can easily be done on an ARM... Or and 8051.   The hard problem is understand the math your are trying to implement,  not that particular platform.

There are *lots* of applications where you only a small Cortex-M do do some very interesting signal processing at very low power envelopes.     There always be the "high end" (Radar, image processing, etc)  but the 6-sigma can be done with general purpose chips nowadays.

Not everything needs a SHARC or a Virtex-7.       Once you get to the super high end, you will need a solid background in implementing algorithms which can be done on any processor platform.   

From my point of view,  I would hire someone with ASM experience over something who is only trained in modern tools.

Many times faculty do not update the tools as it is a ton of work and are lazy....

That all being said, there is nothing stopping from buying a modern kit and getting yourself going on some other tools.    Or you could offer to update the lab exercises!

 

Online nctnico

  • Super Contributor
  • ***
  • Posts: 26752
  • Country: nl
    • NCT Developments
Re: Engineering Students being taught DSP on 20 year old processors RMIT
« Reply #52 on: August 29, 2017, 09:09:08 pm »
Yes, I heard that argument 25 years ago as well. In those days it was about "soft modems" that didn't need any hardware other than AD/DA converters to run a 14400 line, by just using the "incredible processing power" of a 486.

The basis of that argument was and is "just throw some more processing power at it". Extremely primitive.
But very cost effective when done right and in the end that is what counts. One of my former employers killed and borged the entire competition by not using traditional DSPs but using the PC's CPU for doing the signal processing.
There are small lies, big lies and then there is what is on the screen of your oscilloscope.
 

Online Benta

  • Super Contributor
  • ***
  • Posts: 5839
  • Country: de
Re: Engineering Students being taught DSP on 20 year old processors RMIT
« Reply #53 on: August 29, 2017, 09:46:13 pm »
Again, you quote selectively from my post without regarding my other arguments, which I find an offensive, "cherry-picking" way of discussing, otherwise only known from desperate politicians up to an election.
I'm out, I don't like your style.
 

Online nctnico

  • Super Contributor
  • ***
  • Posts: 26752
  • Country: nl
    • NCT Developments
Re: Engineering Students being taught DSP on 20 year old processors RMIT
« Reply #54 on: August 29, 2017, 09:49:01 pm »
Again, you quote selectively from my post without regarding my other arguments, which I find an offensive, "cherry-picking" way of discussing, otherwise only known from desperate politicians up to an election.
I'm out, I don't like your style.
There is not much use in quoting entire posts because people can scroll up and read them. So in order to keep things short and to the point it is good practise to cut things which aren't relevant to the remark. Yeah sure there are niche applications but we already agreed on that so why leave that part in?
There are small lies, big lies and then there is what is on the screen of your oscilloscope.
 

Offline KE5FX

  • Super Contributor
  • ***
  • Posts: 1878
  • Country: us
    • KE5FX.COM
Re: Engineering Students being taught DSP on 20 year old processors RMIT
« Reply #55 on: August 29, 2017, 09:53:52 pm »
The basis of that argument was and is "just throw some more processing power at it". Extremely primitive.

The thing is, processing power scales, while dedicated hardware doesn't.  If you leave school without understanding that... well, I won't say your education was wasted, but I will say that it's just beginning.

I was watching some old videos of talks by Seymour Cray the other day.  Unfortunately there aren't many surviving ones, and the ones that have been posted on YouTube are fuzzy VHS transfers with bad sound quality.  But the lessons he learned the hard way are timeless.  Cray's designs were industry-leading in the 1970s, and were still going strong into the 1980s.  But by the 1990s, Moore's Law caught up to him.  He was forced into one reckless gamble after another.  He bet the company on processors made with discrete GaAs logic, alien-looking three-dimensional modules made with exotic chips for which his company was the only customer on the planet.  They had to be immersed in Fluorinert for cooling, with power consumption measured in the hundreds of watts per cubic inch.  We all know how that worked out for him.

This is basically what has happened to hardware DSP.  If you're studying 56000s in school today, you might as well be learning how 1970s-style vector processors worked.  Or studying how to build flip-flops out of 12AU7s.  Yes, the principles endure, but the hardware is just getting in your way.
 

Offline ez24

  • Super Contributor
  • ***
  • Posts: 3082
  • Country: us
  • L.D.A.
Re: Engineering Students being taught DSP on 20 year old processors RMIT
« Reply #56 on: August 29, 2017, 10:26:45 pm »
Or studying how to build flip-flops out of 12AU7s. 

 :-DD  I remember them
YouTube and Website Electronic Resources ------>  https://www.eevblog.com/forum/other-blog-specific/a/msg1341166/#msg1341166
 

Offline Sal Ammoniac

  • Super Contributor
  • ***
  • Posts: 1662
  • Country: us
Re: Engineering Students being taught DSP on 20 year old processors RMIT
« Reply #57 on: August 29, 2017, 11:04:30 pm »
If you're studying 56000s in school today, you might as well be learning how 1970s-style vector processors worked.  Or studying how to build flip-flops out of 12AU7s.  Yes, the principles endure, but the hardware is just getting in your way.

People still teach embedded classes in universities using 8051s, and those things are even older than the DSP56000.
Complexity is the number-one enemy of high-quality code.
 

Online nctnico

  • Super Contributor
  • ***
  • Posts: 26752
  • Country: nl
    • NCT Developments
Re: Engineering Students being taught DSP on 20 year old processors RMIT
« Reply #58 on: August 29, 2017, 11:15:43 pm »
If you're studying 56000s in school today, you might as well be learning how 1970s-style vector processors worked.  Or studying how to build flip-flops out of 12AU7s.  Yes, the principles endure, but the hardware is just getting in your way.

People still teach embedded classes in universities using 8051s, and those things are even older than the DSP56000.
That still doesn't make it a good idea. The defacto standard is ARM nowadays! If I need to choose between hiring someone who already knows ARM and somebody who only knows about obsolete hardware then the choice will be easy.
There are small lies, big lies and then there is what is on the screen of your oscilloscope.
 
The following users thanked this post: Unordung

Offline tggzzz

  • Super Contributor
  • ***
  • Posts: 19279
  • Country: gb
  • Numbers, not adjectives
    • Having fun doing more, with less
Re: Engineering Students being taught DSP on 20 year old processors RMIT
« Reply #59 on: August 29, 2017, 11:41:07 pm »
If I need to choose between hiring someone who already knows ARM and somebody who only knows about obsolete hardware then the choice will be easy.

I -- and every company I have worked for[1] -- have always looked for people that know the fundamental principles and are competent when it comes to applying them in practical situations.

Why? Because given that it is easy for them to learn today's latest tool. And tomorrows. And to shift onto significantly different technology when it become available. And to assess which class of technology is most appropriate for the next problem they are presented with. And, most importantly, knows what a tool can never be expected to deliver and why.

In contrast, anybody that only knows one tool, even if it is the one I expect to use on the next project, is of little interest. They are the people that hit screws with hammers, because that's what they have done in the past. They are also the people that believe salesmens' patter, and build in fundamental defects that are only discovered very late in the day including after delivery.

[1] except one. That company was more than happy for people to catch and swallow NullPointerExceptions. Why not, after all that stopped the unit tests failing! When asked if they enjoyed polishing the product, one person replied that there was only so much that you could polish a turd!
There are lies, damned lies, statistics - and ADC/DAC specs.
Glider pilot's aphorism: "there is no substitute for span". Retort: "There is a substitute: skill+imagination. But you can buy span".
Having fun doing more, with less
 

Offline Sal Ammoniac

  • Super Contributor
  • ***
  • Posts: 1662
  • Country: us
Re: Engineering Students being taught DSP on 20 year old processors RMIT
« Reply #60 on: August 29, 2017, 11:51:56 pm »
People still teach embedded classes in universities using 8051s, and those things are even older than the DSP56000.
That still doesn't make it a good idea. The defacto standard is ARM nowadays! If I need to choose between hiring someone who already knows ARM and somebody who only knows about obsolete hardware then the choice will be easy.

Basic principles are more important that knowledge of a specific part. Someone well-rounded in the basics can easily pick up a new MCU (like an ARM) with very little trouble. I'd rather hire someone like this than someone who knows ARM and nothing else.
Complexity is the number-one enemy of high-quality code.
 

Offline NorthGuy

  • Super Contributor
  • ***
  • Posts: 3137
  • Country: ca
Re: Engineering Students being taught DSP on 20 year old processors RMIT
« Reply #61 on: August 29, 2017, 11:55:41 pm »
The defacto standard is ARM nowadays! If I need to choose between hiring someone who already knows ARM and somebody who only knows about obsolete hardware then the choice will be easy.

This explains why ARM is so popular. But popularity is a fickle thing. Today it is popular, but year from now it isn't. What if 3 (or 10) years from now something else becomes popular. Say, FPGAs will be cheap and everyone decide to switch to FPGA and ARM will be well forgotten. What are you going to do with your workforce who know nothing but ARM. Send them to college to learn FPGA? Fire them? Hire newer, more progressive guys?

 

Offline KE5FX

  • Super Contributor
  • ***
  • Posts: 1878
  • Country: us
    • KE5FX.COM
Re: Engineering Students being taught DSP on 20 year old processors RMIT
« Reply #62 on: August 30, 2017, 12:07:23 am »
The defacto standard is ARM nowadays! If I need to choose between hiring someone who already knows ARM and somebody who only knows about obsolete hardware then the choice will be easy.

This explains why ARM is so popular. But popularity is a fickle thing. Today it is popular, but year from now it isn't. What if 3 (or 10) years from now something else becomes popular. Say, FPGAs will be cheap and everyone decide to switch to FPGA and ARM will be well forgotten. What are you going to do with your workforce who know nothing but ARM. Send them to college to learn FPGA? Fire them? Hire newer, more progressive guys?

Both ARM and 8051 are standard microprocessor architectures, programmable in C/C++ and targeted by free toolchains based on gcc.  They are more alike than different.  If you learn C programming on 8051, you can transfer those skills to ARM in a weekend. 

This will not be the case if you just spent half the semester dicking around with an antique proprietary toolchain for an obsolete DSP family.
 

Offline vk6zgo

  • Super Contributor
  • ***
  • Posts: 7563
  • Country: au
Re: Engineering Students being taught DSP on 20 year old processors RMIT
« Reply #63 on: August 30, 2017, 12:09:55 am »
Or studying how to build flip-flops out of 12AU7s. 

 :-DD  I remember them

The more usual device used was the 12AT7.
12AU7s were more often used in audio work, though I have used both interchangeably.

More on topic, EEs used to be able to apply fundamentals to come up with very effective answers to problems.
Us grimy old  Techs were more likely to "bodge" something up by digging through the National Semiconductor Data book.

That type of EE sadly seems to be extinct.

"Technical Education"  as represented by the present "TAFE" in Oz has been dumbed down to have very little theoretical education.
It is more "do this & that to achieve this end", which is, allegedly, what Employers want.

As this has become the new reality, students don't realise they are being short changed, & any attempt to teach real Electronics would be meant by shrieks of "it isn't relevant!

It is sad to see the same attitude raising its head in Universities.



 

Offline CatalinaWOW

  • Super Contributor
  • ***
  • Posts: 5170
  • Country: us
Re: Engineering Students being taught DSP on 20 year old processors RMIT
« Reply #64 on: August 30, 2017, 12:30:10 am »
Those who are asking for relevant experience with a specific processor for employment purposes better hope that their proposed employer uses that chip set.  Or maybe they should not set so much store in specific experience.

When I graduated from university it happened to be one of the lulls in engineering employment.  I had very carefully prepared my curriculum to be useful in three widely different patches of electrical engineering - power engineering, RF and semiconductor design/fab.  None of those fields happened to be hiring.  The place I got on worked with infrared.  In response to questions about how I could contribute to their problems I responded "It's all Maxwell's equations" and showed how they applied to some of the things they were doing.  They apparently liked me because they hired me and kept me on through subsequent layoffs.

I think you will find similar things in terms of processor architectures.  If the firm is hiring for a job that is going to last for less than a year they may select specific hardware experience.  If you hope to work in more than a few months you will need to convince them that you can solve all of their problems, not just one problem du jour.
 
The following users thanked this post: Unordung

Offline NorthGuy

  • Super Contributor
  • ***
  • Posts: 3137
  • Country: ca
Re: Engineering Students being taught DSP on 20 year old processors RMIT
« Reply #65 on: August 30, 2017, 12:34:52 am »
Both ARM and 8051 are standard microprocessor architectures, programmable in C/C++ and targeted by free toolchains based on gcc.  They are more alike than different.  If you learn C programming on 8051, you can transfer those skills to ARM in a weekend. 

And so are many others. For someone who had programmed before, it probably would take few hours to grasp the basics of the C syntax, and probably few weeks to a month to study the standard and memorize the majority of the rules. Then you should be able to program equally well (or bad) for any processor. If you look from this viewpoint, you won't see any difference between processors, so you grab whatever lies closer, which most likely will be ARM.

I don't think the OP's course is about programming in C, perhaps not even about programming. May be it merely explains what is DSP and how it is different from RISC processors. Perhaps, if the students spend some time programming DSP in assembler and then do the same with ARM (or other RISC processor), they will see some differences and perhaps start thinking about them. Next time they will not grab what lies closer, but what fits the task better.

 

Offline westfw

  • Super Contributor
  • ***
  • Posts: 4196
  • Country: us
Re: Engineering Students being taught DSP on 20 year old processors RMIT
« Reply #66 on: August 30, 2017, 01:39:36 am »
Quote
Something which has kind of irked myself and my fellow students is our perceived absurdity of our Digital signal processing micro apps class.
The instructor is having us program a Motorolla DSP56800 processor using assembly language.

Is that the whole class, or are you just starting the class and this is what you are learning first?
There are *many* hardware architectures that are very old, that still were so fundamental to the evolution of the industry that knowing about them is A Good Thing.  I'm not sure that there ARE any significant architectures that are "new" - most things are ARM or x86 these days, but both of those are older than the DSP56800...

Quote
D) If you were hiring staff, would you rather them know how to use C++ or assembly to a high proficiency?
First, let me say that I've never been a DSP person; I haven't used DSP hardware, and I haven't programmed DSP algorithms.  But I HAVE been a hiring manager.  If you have "DSP" on your resume, I expect that you'll have some knowledge about how programming a DSP chip in assembly language is DIFFERENT than programming a DSP algorithm in C on ARM.  (in fact, I'd make that an interview question...)  Does C even support typical DSP features like MAC?   (I did buy a TI DSP LaunchPad once, and read through the relevant processor manual, which spend LOTS OF TIME on how to manage the instruction/memory pipeline, for example.)  Anything else would be similar to "I have microcontroller experience: we used Arduino Zeros with micro-Python" - a sort of "ok.  But... reaction."


Quote
I once complained about obsolete material and the prof said to take it up with the Dept Dean.  I asked who that was and he said he was.  I knew I was in deep doo doo.  It turned out to be a hard course with him always calling on me.  The next semester the course was changed.
That sounds absolutely wonderful, actually!  I always learn more from two knowledgeable experts (whether or not I'm one of them( arguing over some fine point, compared to just trying to soak in one point of view...


Quote
There should be a compulsory class called "dabbling with new shit" where you just throw everyone in a room with a bunch of the latest dev boards and a weeks worth of pizza.
Oh yes!  One of the things I've noticed on-line is that forums or mailing lists with a large hobbyist contingent tend to be MUCH more interesting that vendor forums composed entirely of "professionals."  I think this is because even "professional" dabbling is more similar to hobbyist activity than "work."   The times in your career where you'll actually be able to spend a couple of months "comparing" different chips to see which ones you like better are going to be rare, compared to the number of times "I've been dabbling with X and I think it will do the job" might be influential.  This is why cheap hobbyist-friendly evaluation modules are SO important, compared with $1000 "professional" boards.

 

Online nctnico

  • Super Contributor
  • ***
  • Posts: 26752
  • Country: nl
    • NCT Developments
Re: Engineering Students being taught DSP on 20 year old processors RMIT
« Reply #67 on: August 30, 2017, 05:05:50 am »
People still teach embedded classes in universities using 8051s, and those things are even older than the DSP56000.
That still doesn't make it a good idea. The defacto standard is ARM nowadays! If I need to choose between hiring someone who already knows ARM and somebody who only knows about obsolete hardware then the choice will be easy.
Basic principles are more important that knowledge of a specific part. Someone well-rounded in the basics can easily pick up a new MCU (like an ARM) with very little trouble. I'd rather hire someone like this than someone who knows ARM and nothing else.
Why does everyone think you can't learn the basic principles on ARM?  major :palm: !
ARM isn't going away anytime soon and if I hire someone who already knows this architecture then I don't need to pay for that person learning it (which will take more than just a few hours)!
Besides that if someone learned to work with ARM (instead of some obsolete architecture) at the university it also shows the university keeps track on current technologies and the students probably learn a thing or two about other modern methods.
« Last Edit: August 30, 2017, 05:09:22 am by nctnico »
There are small lies, big lies and then there is what is on the screen of your oscilloscope.
 
The following users thanked this post: Unordung

Offline forrestc

  • Supporter
  • ****
  • Posts: 646
  • Country: us
Re: Engineering Students being taught DSP on 20 year old processors RMIT
« Reply #68 on: August 30, 2017, 06:16:10 am »
Let me see if I can stick myself in the middle of the "DSP on ARM" sub-discussion here.

First of all, an ARM is NOT a Digital Signal Processor.   Some ARM cores have Digital Signal Processing-friendly instructions, but they are not what I would call a digital signal processor.

Years ago, if you were doing signal processing of any meaningful sort, you really wanted a Digital Signal Processor.   General purpose processors didn't really have the steam.   Heck, I remember when you couldn't even reliably decode a MP3 on a PC.    Any time you wanted to do signal processing work you really needed a processor which was optimized for that. 

Time has gone on, and the applications that required a DSP 30 years ago can be done in a general purpose processor today in real time.   And many of the processor cores have added DSP instructions to make it easier.   If you're processing audio, generally you're going to be able to do that on a fast general purpose processor.  That doesn't mean that a general purpose processor with DSP instructions is a Digital signal processor, it just means that the applications where a real digital signal processor is needed have evolved.  And many of these applications have moved from a processor to a FPGA.

In a university class on "DSP" I think there are two main things which should/could be covered (I said some of this in a previous reply):

1) Digital signal processing algorithms, (filters, transforms, etc).  This doesn't need a digital signal processor - instead can be taught on pretty much any processor.  And for many applications, a general purpose processor would be the ideal end-product processor.

2) Digital signal processor architectures and how they are useful and how they differ from a general purpose processors.    *IF* you want hands-on work in relation to this topic, then you need to find a processor which is simple enough to comprehend that you could actually be able to utilize it in the limited time given during a course.   

There are still lots of applications out there where a real DSP is the ideal choice.   There are some $1-2 real DSP cores which can do incredible things at very low power (motor control (aka BLDC), power supplies).   There are also some very high end DSP's  (and FPGA's) which can do signal processing that we only dreamed about just a few years ago.   And yes, there are the general purpose cores which can handle far more applications than they did in the past.

I just think that making an argument that "you don't need a DSP to do signal processing" is just as bad as making the argument that "you need a DSP to do signal processing".   The key is to understand the underlying algorithms of signal processing, and then understanding the tradeoffs between a real DSP core and a general purpose one.   Sometimes the correct answer for a given product is a cheap general purpose core, sometimes the answer is a real DSP core, and sometimes the answer is a FPGA.   Or a mixture of the above.   
 

Offline nfmax

  • Super Contributor
  • ***
  • Posts: 1556
  • Country: gb
Re: Engineering Students being taught DSP on 20 year old processors RMIT
« Reply #69 on: August 30, 2017, 07:15:41 am »
1) Digital signal processing algorithms, (filters, transforms, etc).  This doesn't need a digital signal processor - instead can be taught on pretty much any processor.  And for many applications, a general purpose processor would be the ideal end-product processor.

Algorithms and general principles are best taught using a high level mathematical environment, and by far the best choice is Matlab. It is vey widely used in industry, and a CV lacking Matlab experience would definitely raise an eyebrow when recruiting for a DSP position.

Remember also, when recruiting, that any candidate with exactly the right combination of experience and knowledge of the specific processor/tools & hardware is either working for you already, or in a better-paid post with your main competitor.
 

Offline tggzzz

  • Super Contributor
  • ***
  • Posts: 19279
  • Country: gb
  • Numbers, not adjectives
    • Having fun doing more, with less
Re: Engineering Students being taught DSP on 20 year old processors RMIT
« Reply #70 on: August 30, 2017, 07:24:46 am »
... if you just spent half the semester dicking around with an antique proprietary toolchain for an obsolete DSP family.

That's an important point. In this context, the toolchain which allows people to concentrate on the DSP (rather than on the toolchain) is the better choice - and the choice of processor should follow from that.

Now I don't know that specific toolchain, but in my experience newer toolchains have many bells and whistles that obscure what is happening underneath. The worst ones, and there are many, make it easy to do common tasks in a predefined way, but are bloody awkward when you need to do something outside the way they think you ought to do things.

In this context it may be preferable to have a command-line driven toolchain, since then all internal operations are visible (always good), and a teacher can examine/guide/mark the student's work without having to create a project in an IDE.
There are lies, damned lies, statistics - and ADC/DAC specs.
Glider pilot's aphorism: "there is no substitute for span". Retort: "There is a substitute: skill+imagination. But you can buy span".
Having fun doing more, with less
 

Offline tggzzz

  • Super Contributor
  • ***
  • Posts: 19279
  • Country: gb
  • Numbers, not adjectives
    • Having fun doing more, with less
Re: Engineering Students being taught DSP on 20 year old processors RMIT
« Reply #71 on: August 30, 2017, 07:30:53 am »
Does C even support typical DSP features like MAC

That's invisible at the language level and hhat's up to the compiler; some will, some won't.

Saturating vs non-saturating arithmetic is different, since the intent has to expressed in the source code.
There are lies, damned lies, statistics - and ADC/DAC specs.
Glider pilot's aphorism: "there is no substitute for span". Retort: "There is a substitute: skill+imagination. But you can buy span".
Having fun doing more, with less
 

Online Kjelt

  • Super Contributor
  • ***
  • Posts: 6459
  • Country: nl
Re: Engineering Students being taught DSP on 20 year old processors RMIT
« Reply #72 on: August 30, 2017, 07:47:10 am »
Just quickly Looking at current DSPs and ARMs and hybrids (lot of modern DSPs have an DSP core and an ARM core) performances I do see some differences esp in floating point processing.
C6678 from TI can do 160GFlops and 320GMACs 1, even a Raspberri pi3 with 4 cores can not handle that (1-2GFlops)2.

If I look at inners of the last three generations of hometheater audio processors with now 14 channel audio and 3D sound (Atmos/DTS-X etc) they all have TI or AD DSPchipsets.
So I am still not convinced that anything a modern DSP can do an ARM can compete?



1http://www.ti.com/processors/dsp/c6000-dsp/c66x/overview.html
2http://browser.geekbench.com/geekbench2/2591431

 

Offline vealmike

  • Regular Contributor
  • *
  • Posts: 192
  • Country: gb
Re: Engineering Students being taught DSP on 20 year old processors RMIT
« Reply #73 on: August 30, 2017, 07:52:57 am »
I studied programming (IBM 370 assembly, FORTRAN, COBOL) using punched cards (years 1985 to 1990).  Ended having a successful career working for several multinational computer hardware and software companies.  The tools you use to learn is not important.

 :bullshit:  If you were a hiring manager, you'd be singing a very different tune.  Those graduates will be at a serious disadvantage relative to their peers who will be familiar with more modern DSP platforms.

As (until quite recently) the guy sitting next to the hiring manager asking the technical questions, I couldn't disagree more.
These are graduates, not battle hardened engineers. I do not expect them to hit the ground running. I do not expect that their tool chain or processor specific experience will be relevant  to me anyway.

What I do expect is a good understanding of the principles. I want them to know how the hardware does what it does. I want them to know how to approach problem solving.

But most importantly, I want them to have an aptitude and a thirst for engineering. A graduate engineer has a box of tools, but lacks the experience to apply them in a work environment. When they join the team they have to be taught. The more eager they are to learn, the better.

Technologies come and go. I couldn't give a flying poop what processor was learnt at Uni, it will be outmoded in seven years anyway. What's important is the principles.

I studied at a college that specialised in marine electronics ( I live in a city that has a large port.) Most of the kit that we practised on as students was donated to the college when ships were refitted. That means that, in the late 1980's we were working on valve based radios, radars, marine location systems etc.
Our digital logic was taught on the 1975 designed 6502. We used 7400 series TTL and were taught to prototype with wirewrap. Logic analysers were powered by Z80s and oscilloscopes were analogue.

The implementation of the technology is utterly irrelevant, as long as you understand the fundamentals you're good.
 

Offline KE5FX

  • Super Contributor
  • ***
  • Posts: 1878
  • Country: us
    • KE5FX.COM
Re: Engineering Students being taught DSP on 20 year old processors RMIT
« Reply #74 on: August 30, 2017, 08:02:06 am »
Just quickly Looking at current DSPs and ARMs and hybrids (lot of modern DSPs have an DSP core and an ARM core) performances I do see some differences esp in floating point processing.
C6678 from TI can do 160GFlops and 320GMACs 1, even a Raspberri pi3 with 4 cores can not handle that (1-2GFlops)2.

If I look at inners of the last three generations of hometheater audio processors with now 14 channel audio and 3D sound (Atmos/DTS-X etc) they all have TI or AD DSPchipsets.
So I am still not convinced that anything a modern DSP can do an ARM can compete?



1http://www.ti.com/processors/dsp/c6000-dsp/c66x/overview.html
2http://browser.geekbench.com/geekbench2/2591431

How many can a Motorola 56K do?  That's more relevant to the OP's original question.

The implementation of the technology is utterly irrelevant, as long as you understand the fundamentals you're good.

How familiar are you with DSP technology, personally?  (Not challenging or arguing, just curious.)  The impression I've gotten from the thread is that the people saying "It doesn't matter" have no idea how old the 56000 architecture is, and how development tools, practices, and processes have changed since 1986 when that particular chip was introduced.  That impression may be incorrect, so it seems worthwhile to ask.
« Last Edit: August 30, 2017, 08:10:38 am by KE5FX »
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf