Alexander A. wrote:
> I have some skills in software engineering and i think than it's enough
> to start for learning FPGA. Now i'm filling the gap in physical layer.
> In your opinion from your experience what topics needs an FPGA Engineer
> for work (exclude digital circuit design, of course)? Do you using an
> analog circuits or power electronics theory? Have you had to deal with a
> deep theory of transistor amplifiers, analog transmitters, antennas? Use
> you an electronic technician skills (for example soldering, circuit
> troubleshooting)? For understanding ADC & DAC how deep knowledge of
> analog signals i need?

It depends on the level of proficiency you want to achive ... You can 
buy cheaper FPGA dev boards which are ready to go - like Arduninos - and 
you wouldn't have to cope with antennas and transistor amplifiers and 
power supply and so on.

On the other hand, you could be motivated enough to completly design an 
own fpga dev board - then you would have to have a lot of electronic 
skills.

Either way you need a concrete project you want to implement - otherwise 
it will become the 1000th knight-righter-light.

Most importantly, you would have to figure out for what purpose FPGAs 
are used and where the advantages are compared to normal 
microcontrollers.

>If it's relevant, informing you about my interests and attachments of
>FPGA design: electronic music (especially sound synthesis)

You can buy this board:
http://www.mouser.de/new/Lattice-Semiconductor/lattice-machxo2-breakoutboard/

and connect a 100Ohm,100nF to your favorite MP3 speakers and starting 
hands on with this:
Beitrag "VHDL Grundlagen Tonerzeugung"

Mampf wrote:
> 1000th knight-righter-light.

Most probably he shall become the 100.000 FPGA-engineer trying to learn 
from scratch and providing design services to finally rxperience that 
the market is totally fill up already.

My girl friend is working with a personell service provider having more 
than 10.000 engineers in germany under contract / employed and has a 
clear view on the electronic market:

At almost every university in germany, europa, china and inda, jung 
engineers learn FPGA design such as they started to learn about C 
language 30 years ago.

In 1985, C was new and once you learned C well enough you were the hero 
in the 90tees and could earn a lot of money because market was growing 
faster than the students went of of university!

Ten years later in 1995 FPGAs were new and once you leraned Verilog or 
VHDL very well, you were the hero in the 2000th. The market was rapidly 
growing and still in the mid of 2000ths there not enough engineers with 
FPGA knowhow.

From 2005 on, when high level synthesis was invented to drive the 
designs more with VHDL rather than with ABEL oder GHDL oder gate 
placing, and synthesis was easier and more stable and FPGAs became 
bigger, there were already enough people to fill the market, because it 
came more to software and not bot hardware and software. Meaing: The 
companyies splitted VHDL-Design, C-Design and electronic aspects of 
FPGA. There were more engineers involved but more specalist. C-Design 
für FPGAs was new, so there were some issues to solve, but genal 
C-Design was already dead! Electronic design with FPGAs was growing and 
people were required in masses.

But now, 2015, there are about 10 times more FPGA people in the world 
then required, says my mate.

1) The lower market "VHDL only" is 15 times overflooded with people so 
only a small number of students finally work with VHDL. These persons 
are the "below 5 years FPGA desig expepricence" guys and prices are 
going down.

2) In the "middle market", with high speed FPGA design covering DDR, GTP 
and special solutions, there are already three times more people with 
knowledge and only the experienced guys with more than 5 years of 
knowledge get contracts and the guys with more than 10 years can earn 
good money.

3) Only in the "high market" with total coverage, namely electronics, 
signal processing, FPGA special problems, high speed design, SOC, SSO, 
TDC and such things, there is still demand for experts. But you need 15 
years to get all this knowledge and if you start know, there are many 
others in front of you and market is away.

VHDL-Programming and FPGA eelctronics are a dead horse in the long run 
view. You should focus an new aspects:

The demands now are high speed board design, dealing with 500MHz on the 
PCB, EMI, Transitions, fiber optics and high speed internet + 40GSPS. 
Physical design has more demands than simple VHDL or Verilog.

Next point is regulary affairs: 50% of my contracts are placed because I 
have knowledge about market, suppliers, regulary affiars, legal issues, 
patent issues and have nothing to do with electronic knowhow or FPGA.

New FPGA-engineers are not required because market ist steady and 
students are growing faster. Today, FPGA goes to china for developing. I 
recently wrote a specification covering an amont of no less than 4000 
men hours of work totally done in Shenzen. It is around €50 per hour in 
germany currency with is 40% cheaper than FPGA design here. Experts from 
germany are only involved when it comes to special problems and people 
are required to stay inhouse to solve issues are cooperate with the 
inhouse engineers.

There is no FPGA industry anymore. FPGA suppliers are sucking and only 
employ new FAE for marketing. But you have to be tall and good looking 
to do so:-)

FPGA suppliers like Lattice and Actel have market problems and even 
Altera had which was the reason to accept the offer of Intel. There 
design tools for FPGAs becames easier because silicon compiling 
progresses. 10 years ago you needed about 25 people to verify a new 
Silicon Design for a new Spartan FPGA, now there are less than 10 doing 
this in the same time.

I guess companies like Xilinx will shrink down to 60% of their nowadays 
employees.

Regaring usage of FPGAs: Lothar is right. Many companies employ cheap 
C-Designers to click around in the IDE to configurate a system provided 
by a third party supplier. There are a number of platform suppliers from 
italy, canada and also germany I know with total environments for image 
processing, radar and object recognition, which can be totally 
controlled with MATLAB to test and develop new designs.

The number of FPGA-engineers they need, is small too.

Well, this sounds like FPGA programming is a really dead horse.
Are there any other opinions here?

From my point of view, Linux is still becoming stronger on FPGAs and 
growing more quickly than developers do. The reason is, that many 
embedded Linux experts dealing with FPGAs formerly were normal 
electronic engineers, having started with soft core and embedded CPU and 
then lately coming to Linux, like me. Those people have a limited 
knowledge on Linux and high level software programming techniues only. 
There should be a godd point to start.

So one can "overjump" the electronics and focus on software. This is, 
because the FPGA lowlevel issues are already mainly solved, interfaces 
are present and FPGAs are very big and fast today to hold large 
C-Designs. So several Linux machnines can be used, all programmable in 
simple C, but though somehow effective.

Of course the silicon is generally inefficient then, but development 
time is very expensive and time to market has more weight than cheap 
production costs for medium sized projects with medium lot numbers.

For small lot numbers like prototypes and "below 100" pcs - rapid 
prototyping platforms shall be preferred. MATLAB is the point to start 
there. Signal processing is also mainly done for the most cases.

For large lot numbers like 10.000 and above, FPGAs will be optimized in 
VHDL and resource usage, so optimization experts (and low current usage 
tricks) are required. This shall be the field for engineers. There won't 
be a Linux inside then. So SW knowhow is obsolete for those engineers.

For very large lot numbers, ASICs became cheap enough. So 
ASIC-Prototyping in FPGAs is permanently done and requires some 
electronic knowhow.

-----------------

So either concentrate on hardware, like FPGA electronic + PCB, or 
concentrate on software, Linux, conncetions, C++, API, database and data 
handling.

"Big data" handling shall be the next point to serve with FPGAs. My 
customer works at a FPGA Server farm, with tons of FPGA platform PCBs. 
They all can work the same way or also differently. Just by programming. 
There is also an embedded Linux running on each FPGA interacting with 
other linux installations dynamically.

chris wrote:
> and connect a 100Ohm,100nF to your favorite MP3 speakers and starting
> hands on with this:
> Beitrag "VHDL Grundlagen Tonerzeugung"
Not convinced that this helpes since this is thread in german language.

Also not convinced that the idee is fine:

FPGAs are quick circuits and not cheap at all compared to a simple 
microcontroller like for example the STM series. For $25, you can get a 
microcontroller being able to aquire, process and recreate audio signals 
in realtime with existing algorithm libraries like VisualDSP++ rapidly.

For $25 you can get only a small FPGA doing nothing without much Verilog 
programming and is oversized with audio demands. An Arduino PCB hast 
already interface chips on board and interface to ethernet, digital 
audio and has serial programming and USB. Doing USB for USB-Audio is a 
giant project in FPGAs, in Arduino it is boring and easy and done by non 
engineers, just technicians without university graduation.

>in Arduino it is boring and easy and done by non
>engineers, just technicians without university graduation.

well .. no.

>Also not convinced that the idee is fine:

>FPGAs are quick circuits and not cheap at all compared to a simple
>microcontroller like for example the STM series. For $25, you can get a
>microcontroller being able to aquire, process and recreate audio signals
>in realtime with existing algorithm libraries like VisualDSP++ rapidly.

Probably you should tell this Juergen.

http://www.96khz.org/htm/fpgamusicsynthesizercyclone4.htm

He is often active in this forum.

chris wrote:
> Well, this sounds like FPGA programming is a really dead horse.
> Are there any other opinions here?

Totaler Bullshit, FPGA-Entwickler werden immer noch gesucht, aber eben 
die mit einigen Jahren Erfahrung und nicht die Greenhorns von den 
Hochschulen. Von denen mag zwar jeder FPGA an der Hochschule gehört 
haben, selbstständig ein FPGA-Projekt durchziehen können aber 
bestenfalls 5%. Besonders die FPGA-Programmierer aus der Informatik 
entpuppen sich schnell als Totalversager;
Techische Informatiker mal ausgenommen.

Und mit Prototypenboards und Linux auf FPGA's ist schnell ende wenn es 
in den Bereich Echtzeit und Sicherheit geht. Auf fertiglösung greift man 
im Bereich Phased array technologie auch ungern zurück weil man sich 
dann hinsichtlich Anpassungen, Optimierungen auf Jahre vom IP-Anbieter 
abhängig macht, wenn man nicht seklbst FPGA-KnowHow im Haus hat.
Kleiner Tipp, nicht jeder IP-Core der angeboten wird, findet auch 
wirklich einen Käufer.

Mit Simulink allein kriegt man auch kein FPGA-System ans Laufen, FPGA 
ist halt mehr als Regelstrecke und digitalfilter.


PS:
Ebenso ist es totaler bullshit mit deutschen Muttersprachlern ein Thema 
auf emglisch zu diskutieren, erst recht wenn es eine parallele 
Diskussion auf dem deutschsprachigen Unterforum gibt.

Did you recognize, that this is an english forum? Even if there is a 
similar discussion in der german part, you can post your ideas there, 
and should respond in english language! If you are pretending to be so 
well informed then you should be able to write in english simple 
language at least.

For your convenience, I will translate your text:

---------- TRANSLATION OF THE ABOVE GERMAN TEXT ----------

"total bullshit" FPGA-Developpers are still required, but those with 
experices of years and not the green horns. These might have heard of 
FPGA design at their university, but are not able to drive a design 
fully on theirselves. Only 5% can do this.

People from the field if IT are th worst.

With Prototypingboards or Linux at FPGAs, there is a quick limit 
regarding real time and security. For example a "ready to use solution" 
in the field of "phased array technologie" is not a good option because 
one get dependent from the suplier of IP in terms of changes, 
optimizations for years. Have the IP inhouse is better.

With only Simulink it is not possible  to get an FPGA design runing at 
all, because FPGA is much more than just regelation loop and digital 
filtering.

It is also total bullshit to talk with native germans in english since 
there is another similar topic in the german part.

------------------------------------------------------------

Well what the self proclamed expert wanted to say is exactly the same as 
did, namely explaining that only very experienced people are still 
reuqired.

The normal market is filled up completely!

The rest of our wannabe Expert is in fact nonsense, because of course it 
is possible to operate with Simulink and his phased array example does 
not really stand fo the majority of the designs. In fact many designs 
are support by IPs, often vendor like Xilinx some times also third 
party.

Anyway this does not change the facts, that only experiences people are 
required. Experience here means Smulink, Electronics, Regulation, and 
such.

It does not mean "general FPGA" because there are too many as with 
general C design. In both fields, VHDL and C there are 10 times more 
people than needed.

Lothar M. wrote:
> Du bist hier im englischen Forum embdev.net (einfach mal die
> Browserzeile ansehen!)

Lothar was quicker in replying. Anyway same sense.


One sheet I got from my girl friend, what they search for

- Simulink experts beeing able to design regulation processes and model, 
but physicians and Math-People in the first place - no engineers

- Project Managers with knowledge in regulary affairs and legal affairs, 
being able so sign contracts with offshore, but econmists in the first 
place - no engineers

- Team Leaders for offshore residents, driving designs there - engineers 
yes, but only with perfect english and mandarine knowlege

- High Frequency Designers to create PCBs with Antennas - engineers yes, 
but only with strong analog background and PCB-Design techniques

- FPGA-Designers with special knowledge of DDR3 issues, GigabitNetwork - 
engineers yes, but only with special knowledge to solve prolems

- Linux-Experts to solve real time issues - no engineers, but IT people 
with IT knowhow

- embedded Linux experts to solve particular NIOS problems, engineers 
yes, if they have the knowhow, typically IT-guys employed!

- cheap programmers, engineers yes, if not too expensive, physicians 
with no other options preferred

- general FPGA design, no queries. Companies have enough applications.

- general C Design, no queries. Companies have enough applications.

- general mechanics, no queires

- special mechanics to solve special problem with Auto-CAD libs

- sepcial engineers to solve problems with Pulsonix-Libs or Mentor 
knowhow