Discussion in 'Electronic Design' started by [email protected], Jan 10, 2013.

1. ### Guest

Could some electronics guru please shed
some light on this ? First of all, may
I point out that my work so far has
involved either pure analog or pure digital
circuits, with hardly any analog-digital
conversion or vice-versa. So my questions
might sound silly to some experts. Here we
go.
Suppose I am sampling an audio signal, whose
raw voltage levels range from +2.5 Volts to -
2.5 Volts. I understand that a sigma-delta
encoder could tackle this, but can a flash
this ? If not, is there any viable modification
so that the last two ADCs can handle it ?
on this.

2. ### RockyGuest

Depends on the specific ADC. If it has an input range from +2.5v to -2.5v then it will work fine. If not you will have to scale your signal and possibly add an appropriate offset.

3. ### Guest

Whether a particular ADC chip can handle negative voltages is not a function of the method it uses internally to do the conversion: it is a function of the voltage range of the chip itself. There are a number of SAR and flash ADCs that are designed to work with inputs that swing below digital ground, and there are a number of sigma-delta ADCs (mostly the ones that come attached to microprocessors and less expensive audio ones) with inputs that are constrained to not go below 0V.

Look at data sheets. If you get onto the websites of the major semiconductor companies (TI and Analog Devices spring to mind) you should find selection guides that list, among other things, the chips' input voltage ranges.

4. ### Guest

Sure. Simple.

In the case of the SAR, you simply need a bipolar DAC with the
appropriate encoding in the feedback of the comparator. The sign bit
is the MSB (tested first).

A flash ADC would simply have the resistor chain between +2.5V and
-2.5V, with the comparators tapped off it. The output will be a
"thermometer code" but it's not a huge deal to convert that to
whatever code you want (same problem as unipolar, really).

5. ### Guest

Good Lord, you're an asshole, Fields.
Yes, really. Any second-year student would have studied this sort of
thing.
A CPLD, if I had an idiot boss who asked for such a dumb thing. IT's
a simple matter of logic. Maybe more than one would want to hand wire
with 555s (all you know how to use), but simple nonetheless. The
trivial solution is 250ish 2in muxes; a trivial CPLD or FPGA exercise.
Kinda a dumb problem though. Not surprising coming from you.

6. ### mikeGuest

First thing is to specify the problem.
"audio signal" is way to vague.
If you really mean "audio signal"
couple it with a capacitor sufficient to
pass the lowest frequencies and be done with it.

Even if you did build some kind of DC level shifter,
you still have the problem of deciding where zero is.

The solution is easier if you know what you're
solving.

7. ### mikeGuest

Well, we now know what YOU mean by audio.
That may or may not be what the OP means by audio.
I've seen many an engineer lead himself astray with the words
"generally regarded". Why not just ASK for actual data?Well, I'd start with a cap and two resistors.
But that's just me.
Unless you're the OP, YOUR edification is not what's required here.

Last thing we need is to jump to conclusion and apply tunnel vision.

8. ### Uwe HercksenGuest

Hello,

audio may be the range from 300 Hz to 3,4 kHz, or from 20 Hz to 16 kHz,
or 10 Hz to 20 kHz. It depends on telephone voice only or music and the
desired quality.

Bye

9. ### Syd RumpoGuest

Probably Mike means a high-pass, with a resistive divider to set the
mid-point. R1 +Vref to Ain, R2 Ain to -Vref (which may well be Agnd)
and C1 fron signal to Ain. R1 = R2 with C1 and R1//R2 setting the LF
rolloff. Don't forget the ADC input impedance if it's unbuffered.

Cheers

10. ### mikeGuest

Did you just say that the question is unimportant because you have the
I know a number of product development teams that would be delighted to
hire that talent.
You post a lot here. My conclusion is that you're a smart guy and this
is just an exercise in being contrary. You're going way out of your way
to be dense.

There aren't many ways to hook three components together to level-shift an
AC signal. You could have tried them all in less time than it took you
to bitch at me.

The more you know about the requirements, the better chance you have to
create an optimal solution. Being a hammer can be good...unless you
mistake a finger nail for a carpentry nail.

Ability to run a sim has value.
Knowing what to sim...priceless.

11. ### Guest

I answered the question, moron. All of this stuff is on the web if
you know how to look for it. I gave a *lot* more information than you
have. A level shifter (LOL). A typical Fields hack.

You were the one who put parameters on the project that made it a dumb
solution. He asked if it could be done. Certainly it can. Question
But that wasn't the question asked, dumbshit.Is the one trick pony getting sensitive to the truth?
A *really* dumb solution that no one doing this stuff professionally
would have dared offer.
You really are proud of your Fields' hairballs, aren't you?
Because it's *much* easier to do a web search, given the right
keywords. It took me less than 30 seconds. Here, I know you're a
hacker, so maybe you can learn something (doubtful).

http://www.es.isy.liu.se/staff/alumni/eriks/tmp/NORCHIP_2005B_rev2.pdfo

12. ### P E SchoenGuest

wrote in message
so that the last two ADCs can handle it ?
If your source impedance is very low (such as the output of an op-amp), it
is fairly easy to shift the level with two equal value precision resistors,
one of which goes to the Vref. For instance, I'm using a 5V PIC with a 2.5
VDC reference, so when the signal is zero, the ADC sees 2.5/2 = 1.25V. At
the extremes of +/- 2.5V, the ADC sees 2.5V to 0V. Just make sure the
resistors present an impedance low enough for the ADC sampling to acquire a
solid reading. A small capacitor helps. I use two 3.01k resistors and a 10
nF capacitor which is good for most audio signals (I'm reading 60 Hz).

It is easy to convert the resulting 10 bit unsigned digital result to a
signed result by subtracting 512 counts. I use this for an instrument that
calibrates to at least 0.25% which is about the limit of a 10 bit signed
integer value.

Paul

13. ### Guest

Only because you cannot read. ...or understand.
Another Fields' lie. BTW, how's your twin, DimBulb?
Level shifter? The whole concept is a hack, here. Where is "zero",
moron?

Another Fields' lie. Keep going...
That wasn't the question asked. But you don't care. You're only here
to prove how "smart" you are. Really, all you do is show what a hack
you are by publishing your Fields brand of hairballs. What a loser.

Actually, I did. Because you can't read, it's no surprise you don't
get it.
Yes. The truth is not an ad hominem. It's the truth. Sorry if it
Of course you would. It's obvious that you're not an engineer.
Obviously you don't understand the article. It *is* trivial. A
little hierarchy and it's a coupe of minutes work.

14. ### Jasen BettsGuest

if 0V is of special signifigance you may need an input of 0V

15. ### josephkkGuest

Keith, if you were much of a human you would be ashamed to have posted
nearly pure ad hominems. Have fun with the strawman.

?-)

16. ### Guest

[..../]
IRONY

<article not snipped intentionally>