Invisible wrote:
>> "normal" ones work at all sorts of voltages, it even varies from piece 
>> to piece quite significantly.  But you never drive an LED directly by 
>> applying a fixed voltage, you always drive it by regulating the 
>> current to a fixed amount (like 20 mA or whatever).  Adding a series 
>> resistor to a raw LED is a quick and crude method of fixing the 
>> operating current.
> 
> This doesn't make sense to me.
> 
> Presumably the resistence of the LED is finite and fixed. How does 
> adding another resistor help? There are several schematics in my 
> electronics kit that involve LEDs and no resistors at all.
> 


Sorry I've been away and missed the fun.


I hook up a LED this way:

A LED will always drop or *consume* a fixed voltage - again, that's 
voltage not current.
A LED is NOT a resistor - in the most basic sense it generally will not 
prevent current from flowing through it once it starts flowing - it 
starts flowing at it's rated voltage drop.

The typical voltage drop for a basic red LED is 1.7 volts.
For most LEDs you can use this number for calculations.
Be aware, that different colors may have a higher voltage drop.
A LED will generally not light up unless your voltage across its leads 
goes above it's drop.

A LED likes to have between 10mA and 20mA - to be safe you can use 10mA 
in calculations.  The current will affect brightness, but too much 
current will cause it to burn out.  You are better off to select a 
*bright* LED with a high light output rating than try to get more light 
by putting more current through it.


So some easy calculations:

lets say 5v power source
1.7v LED
10mA current

the LED drops 1.7v so the rest of the circuit will drop 3.3v
(5v)-(1.7v)=(3.3v)

If don't put a resistor in then the current through the circuit will be:
I=V/R   from V=IR

I=(5v)/(0 ohm)
not good  too much current - things get hot
	(note:  LEDs should not get hot)

we want the current (I) to be about 10mA
so R=V/I   again from V=IR

R=(3.3v)/(10mA)
R=330 ohm


330 ohm happens to be a common resistor value
so put one in

if your source is 12v then you need a 1,030 ohm resistor - not a common 
value.  But 1,000 ohm is - so use it instead.

Typical resistor values are only +-10% anyway - so a 330 ohm resistor 
could be as low as 300 or as high as 360 and you don't know it.

So picking something slightly off from the *calculated* value is OK.
After all, you are just trying to make a LED light up - not try to 
maximize it's brightness.



Tom




















Different LEDs drop different voltages, but you can usually safely bet 
on 1.5V and still be safe.

Then there is the current rating of the LED - usually 10mA to 20mA - you 
can pick 15mA and be pretty safe.

Post a reply to this message

Orchid XP v8 wrote:
> Darren New wrote:
>> Invisible wrote:
>>> How many resistors can you see? Because I count NONE!
>>
>> I look at an Intel chip stuck to my motherboard, and I don't see any 
>> transitors, either.  The gate-level logic is working with bits, not 
>> voltages. You're not going to see the analog stuff that makes the bits 
>> work there.
> 
> ....which is why I'd rather work at the gate-level. ;-)
> 

The problem is that digital electronics are based on analog circuits.  A 
lot of digital circuits can be wired up with out knowing anything about 
analog, but when something doesn't work as you expect, you might quickly 
get lost.


And as you have found out with LEDs, even when you think you are only 
dealing with digital, you are dealing with analog as well.

You don't have to understand breakdown voltages and the like to 
successfully work with LEDs, you just have to understand some of their 
base characteristics.

Even something as simple as a button or switch can cause problems.
When you close the contact there are literally thousands of connections 
made and broken in the very short time before the contact is completely 
closed.  If you have a counter, it might count 4 or 5 on each button 
press, not just one.  That's what debounce circuits are for, but they 
can be analog - go figure.



Electronics can be deceptively simple, but insanely complex.


I think you have the capacity to really work well in electronics - but 
you just don't have the background in how it all comes together.  That 
comes mostly with experience.

Books will tell you how to calculate you need a 524 ohm resistor - 
that's insane.
Experience will tell you that 470 ohm (a typical value) is OK.

Don't get discouraged when things don't work as you think they should.
Pretend that you don't know anything about what is not working and go 
research it.  You will learn loads and loads.
If a LED doesn't light, go learn how it works.
The same goes for most anything else.

Don't try to build a rocket ship - build something simple.
Get a book that doesn't just say hook this wire here and there, but that 
actually goes through how it all works.
The CMOS Cookbook I pointed out earlier is a good book in that reguard.




Tom

Post a reply to this message

>> ....which is why I'd rather work at the gate-level. ;-)
>>
> 
> The problem is that digital electronics are based on analog circuits.  A 
> lot of digital circuits can be wired up with out knowing anything about 
> analog, but when something doesn't work as you expect, you might quickly 
> get lost.

Yeah, this is the thing. I had assumed that the digital devices you buy 
in the shop are carefully designed to behave in a simple, predictable 
mannar, even though their internal construction is a tangle on analogue 
devices. I thought you could just treat a 7400 as a black box. 
Apparently not...

> And as you have found out with LEDs, even when you think you are only 
> dealing with digital, you are dealing with analog as well.

The LEDs actually worked just fine. It's the gate driving them that 
didn't do what it was ment to.

> Even something as simple as a button or switch can cause problems.
> When you close the contact there are literally thousands of connections 
> made and broken in the very short time before the contact is completely 
> closed.  If you have a counter, it might count 4 or 5 on each button 
> press, not just one.

I wouldn't find this surprising. Rather, to be expected.

> That's what debounce circuits are for, but they 
> can be analog - go figure.

Presumably a debounce circuit is merely a low-pass filter?

> Don't try to build a rocket ship - build something simple.

Heh. I got stuck just trying to get a truth table out of a logic gate. :-S

> Get a book that doesn't just say hook this wire here and there, but that 
> actually goes through how it all works.
> The CMOS Cookbook I pointed out earlier is a good book in that reguard.

...does it matter that I'm using TTL?

Post a reply to this message

Orchid XP v8 wrote:
> Um... the obvious problem here is that I'm obviously not intelligent
> enough to do a PhD. :-P

	Yes you are (compared to the folks I knew in grad school). And yes,
Darren is correct: Perseverance and desire is a bigger factor than
intelligence.

-- 
I think animal testing is a terrible idea. They get all nervous and give
the wrong answers.


                    /\  /\               /\  /
                   /  \/  \ u e e n     /  \/  a w a z
                       >>>>>>mue### [at] nawazorg<<<<<<
                                   anl

Post a reply to this message

Darren New wrote:
> bureaucratic BS, and yes some amount of learning graduate-level computer
> stuff, since you need the MS first.  (Assuming you're getting a PhD in
> computers.)

	Depends on the university. In a number of well known grad schools in
the US, they actively discourage you from getting a MS, and strongly
encourage you just go straight to the PhD - in fact, they often don't
offer funding if you don't go directly into PhD, because they assume you
just want an MS and will leave.

	(Talking of computer science departments here...)

> You probably need to be a lot smarter to get a PhD in (say) physics or
> engineering than in computers.

	Engineering?

	Not really. I think it's on par with CS. Both have areas that will
require a lot of smartness, and both have areas that don't.

-- 
I think animal testing is a terrible idea. They get all nervous and give
the wrong answers.


                    /\  /\               /\  /
                   /  \/  \ u e e n     /  \/  a w a z
                       >>>>>>mue### [at] nawazorg<<<<<<
                                   anl

Post a reply to this message

Orchid XP v8 wrote:
> Fact: I absolutely *suck* at research. At uni, every assignment
> involving any element of research was consistently graded very poorly
> indeed. Several such modules I failed outright.

	Oh please! First, what people call research at the undergrad level
often isn't.

	Second, you've been doing research all these years. Being curious and
learning stuff - especially technical stuff - is virtually research. All
that's left is to do some new interesting stuff in it. And generally
your adviser will guide you through that.

> Fact: I also suck at report writing. I'm good at writing technical
> stuff, but reports are supposed to have a specific structure and I don't
> really grok that. Also I'm not very good at structuring large documents;
> the flow tends to end up rather muddled.

	Writing papers for journals is a pain the first or second time. Beyond
that, you have the template figured out and you just write. It's
actually less painful than the lab reports I had to do in undergrad.

	And not all papers are large.

	And when you've spent time working on something, then trust me, you'll
probably have more difficulty trimming it down than trying to fill
space. Just look at your own Haskell evangelism!

> Fact: I don't actually need a PhD for anything. Certainly I don't have
> the money to pay for one. Time would also seem to be an issue.

	*Nobody* needs a PhD.

	And don't pay for one. Few people do. In the US, it's usually paid for
either by being a teaching assistant or being a research assistant (you
get a good enough stipend). I'm sure you can figure out how to get it
paid for in the UK (government scholarship, etc). If not, come this side
of the Atlantic or cross the Channel and do it in Europe.

	Seriously. I think it's the best advice you've been given. You get paid
to do all the fun geeky stuff your adviser wants you to do. And it's not
that rare that you'll also want to do it too! You'll be around smart
people (which *may* make you feel stupid for a while, but that'll go
away). And not sure about the UK and Europe, but (many) campuses in the
US have a good social environment. Lots of great speakers visit campus
and give talks - spanning the whole spectrum from science to politics.

	And no, you don't *have* to be social if you don't want to ;-)

	Others can give advice regarding the PhD in the UK and parts of Europe.
If you want to consider the US as an option, I can likely help you on
some of the details (GRE, etc).

> I could almost be tempted to do math classes - but again time and money
> would seem an issue. (As well as the minor detail of finding a suitable
> class somehow.)

	I went to grad school in electrical engineering. I told myself that
while there, I'd try to get a MS in either physics or mathematics. My
mathematics background when I joined was likely no better than yours. I
did the work for a MS in physics more or less as part of my EE PhD (not
yet completed) - so I won't get the MS in that. I was half way through
the mathematics requirements before I got sick of homework.

	Not to start a flame war or anything, but the MS in mathematics in the
US is generally not that advanced. Probably half of the courses are done
at the undergrad level in many places in Europe.

	Bottom line: At least in the US, they're very flexible about things.
Even if you're doing a PhD in engineering or CS, there's nothing
preventing you from taking courses in mathematics. And because of the
(*cough*) lower standards (*cough*) in mathematics, you won't have
trouble filling any missing maths prerequisites and going further
towards an MS in it.

-- 
I think animal testing is a terrible idea. They get all nervous and give
the wrong answers.


                    /\  /\               /\  /
                   /  \/  \ u e e n     /  \/  a w a z
                       >>>>>>mue### [at] nawazorg<<<<<<
                                   anl

Post a reply to this message

Mueen Nawaz wrote:
> [ lots of stuff ]

I'll second everything she said, including helping you find and get 
involved in the right place if you want.

Places in the USA pay you to get the graduate degree, because "famous 
alumni" are how they sell the school to the paying undergraduate 
schools. If you were doing an undergrad in physics, wouldn't you prefer 
to go to the school where Hawkings got his PhD?  Our schools aren't 
nearly old enough to have reputations on their own without famous people 
associated with them. :-)

It's also an excellent way to meet (a) friends and (b) business 
contacts. You'll be in a mile-square place with thousands of people your 
own age, most pretty smart, most rather more tolerant of differences 
than places with less explicit mixing of cultures going on, etc. You'll 
also be around people whose job it is to travel around the world meeting 
with and talking to people with interesting problems, and whose other 
job it is to talk people into giving them money to solve interesting 
problems they didn't know they had.

It's not about the computers. It's about the people.

-- 
Darren New / San Diego, CA, USA (PST)

Post a reply to this message

Mueen Nawaz wrote:
> 	Depends on the university. In a number of well known grad schools in
> the US, they actively discourage you from getting a MS, and strongly
> encourage you just go straight to the PhD - in fact, they often don't
> offer funding if you don't go directly into PhD, because they assume you
> just want an MS and will leave.

Odd. Never heard of that. I can see your point, but I would think the 
number of people who go thru the MS saying they'll get a PhD and knowing 
they're going to bail before then must be fairly low. Maybe where that 
has happened too often they instituted this policy.

>> You probably need to be a lot smarter to get a PhD in (say) physics or
>> engineering than in computers.
> 
> 	Engineering?
> 
> 	Not really. I think it's on par with CS. Both have areas that will
> require a lot of smartness, and both have areas that don't.

Maybe. The few engineering courses I sat in on seemed a lot harder. 
Maybe I'm just better at computers.

-- 
Darren New / San Diego, CA, USA (PST)

Post a reply to this message

Invisible wrote:
> 
> Yeah, this is the thing. I had assumed that the digital devices you buy 
> in the shop are carefully designed to behave in a simple, predictable 
> mannar, even though their internal construction is a tangle on analogue 
> devices. I thought you could just treat a 7400 as a black box. 
> Apparently not...
> 

To a point you can - knowing that point is the trick - and likely some 
of what you are facing.

>> And as you have found out with LEDs, even when you think you are only 
>> dealing with digital, you are dealing with analog as well.
> 
> The LEDs actually worked just fine. It's the gate driving them that 
> didn't do what it was ment to.
> 

One think to keep in mind, you are best off not trying to drive the LED 
directly from the chip.  Some chips can do it, but most cannot.  Lots of 
times you need to put a *switch* that the gate switches.
see:
http://www.rason.org/Projects/transwit/transwit.htm
http://www.eecs.tufts.edu/~dsculley/tutorial/transistors/transistors1.html

remember, you need a resistor in the path from + to ground with the LED.


>> Even something as simple as a button or switch can cause problems.
>> When you close the contact there are literally thousands of 
>> connections made and broken in the very short time before the contact 
>> is completely closed.  If you have a counter, it might count 4 or 5 on 
>> each button press, not just one.
> 
> I wouldn't find this surprising. Rather, to be expected.
> 


Good, something won't catch you by surprise.  But it's all those little 
things that can add up and frustrate you.


>> That's what debounce circuits are for, but they can be analog - go 
>> figure.
> 
> Presumably a debounce circuit is merely a low-pass filter?
> 

can be - it can be implemented purely analog, or it can be filtered in 
programming.
You can also use some logic gate tricks to debounce.


>> Don't try to build a rocket ship - build something simple.
> 
> Heh. I got stuck just trying to get a truth table out of a logic gate. :-S
> 

So, keep trying to do it.  That's how you start.  Overcome the things 
that keep it from working.  How else do you learn.
You can certainly ask questions here - you've already gotten an ear full.


>> Get a book that doesn't just say hook this wire here and there, but 
>> that actually goes through how it all works.
>> The CMOS Cookbook I pointed out earlier is a good book in that reguard.
> 
> ....does it matter that I'm using TTL?

TTL & CMOS are very similar in their logic, but completely different in 
how they work internally.

You can't just hook the two together.  At you level you are better off 
sticking with one or the other.

CMOS is much lower power and is what is typically found in newer 
electronics - tho TTL has its place as well.

There is a TTL cookbook by the same author - it is basically the exact 
same book.

You can use either book, they are both good starts.


The TTL Cookbook is what opened my eyes that I could actually do 
electronics and it was relatively easy.  I had to return it as it was a 
loaned book, but then I bought the CMOS cookbook and took off from there.



I highly recommend either of the books if you are really interested in 
learning some about electronics.  Unless you have $100 in TTL chips, I 
recommend that you go the CMOS route as it is more likely what you would 
encounter when dealing with more specialized chips (think CPU).

If you get one, start reading it from the beginning and learn the 
fundamentals.

Post a reply to this message

Tom Austin wrote:

> I highly recommend either of the books if you are really interested in 
> learning some about electronics.  Unless you have $100 in TTL chips, I 
> recommend that you go the CMOS route as it is more likely what you would 
> encounter when dealing with more specialized chips (think CPU).

Just keep in mind CMOS is very easy to cook.

-- 
~Mike

Post a reply to this message