I have tried to learn the DSP by reading and play with MatLab. Now I want to do real time DSP by programming floating point DSP(it is easy to program compared w/ fixed point DSP). Two line of chips come to my mind: Analog Device's 21161/21262 and TI's 6713. Could anyone give me advice which one is better based on: 1. Market share - especially for job market. 2. Ease of use - learn curve is not steep. 3. Good support - documentation, samples, third party support.
recommendation on Floating Point DSP chip
Started by ●June 3, 2004
Reply by ●June 3, 20042004-06-03
yqin_99@yahoo.com (yong) wrote in message news:<be3c7875.0406031035.74ef9d77@posting.google.com>...> I have tried to learn the DSP by reading and play with MatLab. Now I > want to do real time DSP by programming floating point DSP(it is easy > to program compared w/ fixed point DSP). Two line of chips come to my > mind: Analog Device's 21161/21262 and TI's 6713. > > Could anyone give me advice which one is better based on:Hi, From what I have learned/heard, and based on personal experience:> 1. Market share - especially for job market.Equal opportunity, in general. Specifically talking about floating point, Analog Devices has definitely become THE leader, with its SHARC platform.> 2. Ease of use - learn curve is not steep.The VERY easiest is Analog Devices. Its algebraic assembly language is rather a high level assembly language, in the sense that it is very easy to use, the instructions are easy to code, and let you take full control of the DSP resources without much effor, which is definitely not the case with the C6000 family from Texas Instruments (indeed, not the case with any TI's DSP family at all!). Besides that, you can always take advantage of C/C++ programming with the SHARC family, the tools are rather good and stable nowadays.> 3. Good support - documentation, samples, third party support.You can always get free engineering samples. If not possible from the manufacturer's website, ask a representative. Third party support is very good for both vendors, as well as the documentation. I think they are quite similar. But, specifically for floating point, I'd rather go for ADI's SHARC family. I hope it helped. JaaC
Reply by ●June 3, 20042004-06-03
jaime.aranguren@ieee.org (Jaime Andres Aranguren Cardona) wrote in news:14a86f87.0406031523.57a55029@posting.google.com:> yqin_99@yahoo.com (yong) wrote in message > news:<be3c7875.0406031035.74ef9d77@posting.google.com>... >> I have tried to learn the DSP by reading and play with MatLab. Now I >> want to do real time DSP by programming floating point DSP(it is easy >> to program compared w/ fixed point DSP). Two line of chips come to my >> mind: Analog Device's 21161/21262 and TI's 6713. >> >> Could anyone give me advice which one is better based on: > > Hi, > > From what I have learned/heard, and based on personal experience: > >> 1. Market share - especially for job market. > > Equal opportunity, in general. Specifically talking about floating > point, Analog Devices has definitely become THE leader, with its SHARC > platform. > >> 2. Ease of use - learn curve is not steep. > > The VERY easiest is Analog Devices. Its algebraic assembly language is > rather a high level assembly language, in the sense that it is very > easy to use, the instructions are easy to code, and let you take full > control of the DSP resources without much effor, which is definitely > not the case with the C6000 family from Texas Instruments (indeed, not > the case with any TI's DSP family at all!). Besides that, you can > always take advantage of C/C++ programming with the SHARC family, the > tools are rather good and stable nowadays. > >> 3. Good support - documentation, samples, third party support. > > You can always get free engineering samples. If not possible from the > manufacturer's website, ask a representative. Third party support is > very good for both vendors, as well as the documentation. > > I think they are quite similar. But, specifically for floating point, > I'd rather go for ADI's SHARC family. > > I hope it helped. > > JaaCAs an ADI partisan, I agree with all of Jaime's points. From strictly a marketshare point of view, and after adjusting for the huge sales of DSPs for very specific markets such as cell phones, disk drives, etc., the DSP market is evenly divided by TI and ADI, You also might get this impression by lurking on comp.dsp for awhile. The questions, comments, etc are about evenly divided between TI & ADI. Most of us have placed our bets on one company or the other. I chose ADI many years ago and its been a great relationship. The 21262 and the 6713 have approximately the same performance. The ADI parts are much, much easier to program in assembly (which is my preference for DSP programming) and probably equal in C. As an employer, I consider skills with a specific DSP family less important than many other factors. Work ethic and intelligence are much more important. A TI DSP guy can learn to program an ADI DSP proficiently in a few weeks, and the converse is true. I have programmed DSPs from four different manufacturers in my DSP career, as well as DSPs from different families from the same manufacturer. Its just not that difficult to adapt. I estimate that the useful technical half life of a good enginner is 3 years. This means that you better plan on constantly improving and maintaining your skills if you want to be competent. I would start (AND COMPLETE) a real project using a DSP (my choice is the ADSP-21262, but I'm partisan). It doesn't have to be incredibly complex but you need to sweat out all the problems. It wouldn't hurt to learn a few tech skills either. If you are just starting out, bring your project to job interviews (assuming an engineering manager is the interviewer and not HR). Good luck, I'm sure your decision will be fine. -- Al Clark Danville Signal Processing, Inc. -------------------------------------------------------------------- comp.dsp conference July 28 - Aug 1, 2004 details at http://www.danvillesignal.com/index.php?id=compdsp email: compdsp@danvillesignal.com Who says you can't teach an old dog a new DSP trick?
Reply by ●June 4, 20042004-06-04
yong wrote:> I have tried to learn the DSP by reading and play with MatLab. Now I > want to do real time DSP by programming floating point DSP(it is easy > to program compared w/ fixed point DSP). Two line of chips come to my > mind: Analog Device's 21161/21262 and TI's 6713. > > Could anyone give me advice which one is better based on: > > 1. Market share - especially for job market. > 2. Ease of use - learn curve is not steep. > 3. Good support - documentation, samples, third party support.If I'm going to interview you for a job I'm going to be much more interested in your ability to tackle _any_ DSP. So I'm not going to be impressed if you seem tied to any one chip. I _am_ going to be impressed if I see that you have programmed in more than one assembly language, and if you can tell me with a straight face that you can pick up a new one by reading the data books. I'm going to ask you questions that make sure that you stayed awake during your high school algebra classes, that make sure you know how a processor works and that you are aware of the implications of using finite data widths. If you're going to program DSP's for me I'm going to make sure that you know a little bit of signal processing. Most importantly, though, you'll have to know how to correctly respond when your partner bids two no-trump when you're holding seven spades to the 10 and have a hand that's otherwise a Mulligan. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com
Reply by ●June 4, 20042004-06-04
You'll be definitely fine with SHARC, it even became some sort of nickname a DSP nowadays. Their assembler is good and clear, so it could be easily integrated with the C/C++ framework. Note, choose project which _really_interesting_ for you, that's important. Nordic regards, Yuri
Reply by ●June 4, 20042004-06-04
"Tim Wescott" <tim@wescottnospamdesign.com> wrote in message news:10c02cf46g0aq5c@corp.supernews.com...> > If I'm going to interview you for a job I'm going to be much more > interested in your ability to tackle _any_ DSP. So I'm not going to be > impressed if you seem tied to any one chip. I _am_ going to be > impressed if I see that you have programmed in more than one assembly > language, and if you can tell me with a straight face that you can pick > up a new one by reading the data books. > > I'm going to ask you questions that make sure that you stayed awake > during your high school algebra classes, that make sure you know how a > processor works and that you are aware of the implications of using > finite data widths. If you're going to program DSP's for me I'm going > to make sure that you know a little bit of signal processing. > > Most importantly, though, you'll have to know how to correctly respond > when your partner bids two no-trump when you're holding seven spades to > the 10 and have a hand that's otherwise a Mulligan.I was on track to get the job until that last paragraph! What card game is that?
Reply by ●June 4, 20042004-06-04
Al Clark <dsp@danvillesignal.com> wrote in message news:<Xns94FDD7BAB5A2aclarkdanvillesignal@66.133.130.30>...> jaime.aranguren@ieee.org (Jaime Andres Aranguren Cardona) wrote in > news:14a86f87.0406031523.57a55029@posting.google.com: > > > yqin_99@yahoo.com (yong) wrote in message > > news:<be3c7875.0406031035.74ef9d77@posting.google.com>... > >> I have tried to learn the DSP by reading and play with MatLab. Now I > >> want to do real time DSP by programming floating point DSP(it is easy > >> to program compared w/ fixed point DSP). Two line of chips come to my > >> mind: Analog Device's 21161/21262 and TI's 6713. > >> > >> Could anyone give me advice which one is better based on: > >> > As an employer, I consider skills with a specific DSP family less > important than many other factors. Work ethic and intelligence are much > more important. A TI DSP guy can learn to program an ADI DSP proficiently > in a few weeks, and the converse is true. I have programmed DSPs from > four different manufacturers in my DSP career, as well as DSPs from > different families from the same manufacturer. Its just not that > difficult to adapt.I was thinking of that too, when writting my first reply. And I agree, too, thanks Al. I've always used ADI, but I consider it wouldn't be too dificult to move to TI, if I had to, which hasn't happened yet. Anyway, learning a bit about other manufacturer's stuff doesn't hurt. Regards, JaaC
Reply by ●June 5, 20042004-06-05
"Jon Harris" <goldentully@hotmail.com> wrote in news:2ibuq6FlfrnoU1@uni-berlin.de:> "Tim Wescott" <tim@wescottnospamdesign.com> wrote in message > news:10c02cf46g0aq5c@corp.supernews.com... >> >> If I'm going to interview you for a job I'm going to be much more >> interested in your ability to tackle _any_ DSP. So I'm not going to >> be impressed if you seem tied to any one chip. I _am_ going to be >> impressed if I see that you have programmed in more than one assembly >> language, and if you can tell me with a straight face that you can >> pick up a new one by reading the data books. >> >> I'm going to ask you questions that make sure that you stayed awake >> during your high school algebra classes, that make sure you know how >> a processor works and that you are aware of the implications of using >> finite data widths. If you're going to program DSP's for me I'm >> going to make sure that you know a little bit of signal processing. >> >> Most importantly, though, you'll have to know how to correctly >> respond when your partner bids two no-trump when you're holding seven >> spades to the 10 and have a hand that's otherwise a Mulligan. > > I was on track to get the job until that last paragraph! What card > game is that? > >Bridge, Partner has 21-23 high card points, at least 2 spades. You are not going to make 6 spades and you might not have transportation to run spades in no trump. Two no trump is not forcing to game, so three spades is likely to be passed. Three clubs is probably not a good bid either because Stayman doesn't really apply since you have so many spades. I would bid 4 spades and maybe make one overtrick. -- Al Clark Danville Signal Processing, Inc. -------------------------------------------------------------------- comp.dsp conference July 28 - Aug 1, 2004 details at http://www.danvillesignal.com/index.php?id=compdsp email: compdsp@danvillesignal.com Who says you can't teach an old dog a new DSP trick?
Reply by ●June 5, 20042004-06-05
>>> Most importantly, though, you'll have to know how to correctly >>> respond when your partner bids two no-trump when you're holding seven >>> spades to the 10 and have a hand that's otherwise a Mulligan. >> >> I was on track to get the job until that last paragraph! What card >> game is that? >> >> > > Bridge, Partner has 21-23 high card points, at least 2 spades. You are > not going to make 6 spades and you might not have transportation to run > spades in no trump. Two no trump is not forcing to game, so threespades> is likely to be passed. Three clubs is probably not a good bid either > because Stayman doesn't really apply since you have so many spades. I > would bid 4 spades and maybe make one overtrick. > >Bridge is a game where the bidding is as important as the play. His real questions are: 1. How do you size up your situation? 2. Are you willing to exploit your opportunities? (3 spades versus 4 spades) 3. How do you take into account your coworker's contribution and how well do you communicate with your coworker? 4. Do you go for broke even if its not very realistic (6 spades)? 5. And perhaps, what do you do when you don't even understand the question? -- Al Clark Danville Signal Processing, Inc. -------------------------------------------------------------------- comp.dsp conference July 28 - Aug 1, 2004 details at http://www.danvillesignal.com/index.php?id=compdsp email: compdsp@danvillesignal.com Who says you can't teach an old dog a new DSP trick?
Reply by ●June 5, 20042004-06-05
Al Clark wrote:>>>>Most importantly, though, you'll have to know how to correctly >>>>respond when your partner bids two no-trump when you're holding seven >>>>spades to the 10 and have a hand that's otherwise a Mulligan. >>> >>>I was on track to get the job until that last paragraph! What card >>>game is that? >>> >>> >> >>Bridge, Partner has 21-23 high card points, at least 2 spades. You are >>not going to make 6 spades and you might not have transportation to run >>spades in no trump. Two no trump is not forcing to game, so three > > spades > >>is likely to be passed. Three clubs is probably not a good bid either >>because Stayman doesn't really apply since you have so many spades. I >>would bid 4 spades and maybe make one overtrick. >> >> > > > Bridge is a game where the bidding is as important as the play. > > His real questions are: > > 1. How do you size up your situation? > 2. Are you willing to exploit your opportunities? (3 spades versus 4 > spades) > 3. How do you take into account your coworker's contribution and how well > do you communicate with your coworker? > 4. Do you go for broke even if its not very realistic (6 spades)? > 5. And perhaps, what do you do when you don't even understand the > question? > >And here I thought I was just being facicious, or only willing to hire folks who'd play Bridge on lunch break -- although one's play it bridge (if one plays) is highly reflective of how one approaches one's job. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com
