AD & DA

Indeed it would be possible, and of course that's how it was done originally. Nowadays chips are more practical for most applications, but simple pulse averaging D/As are still used on the outputs of microprocessors that don't have true D/As built in.

On older systems with parallel ports, you can make a simple 8-bit D/A with a handful of resistors. See for a discussion of simple D/A circuits.

With a D/A in hand, you can turn it into an A/D via "successive approximation", where the system (computer, or internals of A/D chip) toggles the D/A bits as it compares the D/A output to the input signal. When they match close enough, that bit pattern is the converted A/D value. See for more info.

Best regards,

Bob Masta DAQARTA v5.00 Data AcQuisition And Real-Time Analysis

Scope, Spectrum, Spectrogram, Sound Level Meter Frequency Counter, FREE Signal Generator Pitch Track, Pitch-to-MIDI DaqMusic - FREE MUSIC, Forever! (Some assembly required) Science (and fun!) with your sound card!

Thanks Bob!

Your product is interesting. How about Linux version? My hobby at the moment is to write Linux software for the Velleman PCSGU250 scope. It is not much yet. My math is poor so I have to study little to be able to do the Fourier and sin(x)/x stuff. I don't even yet to know what that sin(x)/x stuff is for...

Happy hacking to you !

Kari

--
PIC - ARM - DISPLAYS - RELAYS - MODULES - CONVERTERS - I2C - SPI -
KEYPADS - ACCESSORIES
http://www.byvac.com   (I am just a satisfied customer)

This post just got me thinking about the A/D concept. Just curious about the basic concept (in 1 or 2 paragraphs). I think I understand how the process works but just wondering if you sample a microphone and store its value, that single sample represents the amplitude of the wave at that time and the frequency is reconstructed by stringing multiple samples together.

I hope this is clear. I guess another way of saying this is how are both the amplitude & freq derived from a single number (please keep it simple).

There's very little that came about because of ICs. Most of them are based on things that existed before ICs. The problem is that a lot of those things were not practical in the days of tubes or even transistors, they existed but either only in the laboratory or in very expensive equipment, because it took up too many parts and too much space.

So synthesized tuning of radios was possible in the old days, but it was only in very high end commercial equipment, but once ICs allowed for high integration, they became common and indeed have certain advantages in terms of cost and space (now they actually use up less space than analog tuning did).

Even with things relatively recent, they often existed as discrete components or at least using very common low integration ICs. So as others have pointed out, D/A converters were in the form of a latch on a databus and a set of resistors, and once you had that you could add a comparator and use the computer to control the D/A converter and make an A/D converter. Or even do it without a CPU, requiring more hardware. Digital Signal Processing was being done before ICs to handle the task came along.

A lot of the lowering of prices of consumer electronics comes because the design goes through a few iterations. So a first VCR or computer used a lot of common ICs that weren't very high density, and the cost and size of the equipment would reflect that. As sales rose, the manufacturer could afford to go to higher density components, which also cut manufacturing costs, so the ICs become higher density and less generic. That often happens a few times until there are virtually no parts in a piece of equipment, and the ICs can't be used except for that very specific use.

Michael

CC Inscribed thus:

Google "Nyquist Criterion".

--
Best Regards:
                     Baron.

Yes. In theory, to reconstruct the signal, you need to sample at a rate at least twice the highest frequency in the signal. In practise,

3x or 4x works well. The resulting ADC samples may look ratty to the eye, but if you later run them through a DAC and a lowpass filter, you can almost-perfectly reconstruct the original smooth signal.

People who don't understand this post elaborate web pages proving that DVDs grossly distort music, which of course they don't.

John

got it. thanks

Hah! Excuse for a math digression! The vast majority of 'analog' signals are time-varying functions which are smooth (continuous and with continuous and well-defined finite time derivatives). That description excludes some kinds of phenomena, like truly uncorrelated 'white' noise (thus, we extremely pedantic types like to refer to 'pink' noise).

One measurement of a signal isn't enough to tell its development in time, so a series of measurements is made; there are some kinds of signals (periodic repeating ones) where that series can be finite and still not 'miss' aspects of the underlying analog signal, thus we speak of the series as 'oversampled' if it suffices to determine the interesting part of the signal, and 'undersampled' if it does not.

A nonperiodic signal can be packaged like a string of sausages into a periodic one.

Then there are mathematical treatments (like Fourier analysis) that can represent a signal in multiple ways, either a series of timed voltage measurements or a spectrum of amplitude-and-phase of frequencies. There's an inversion theorem that says your 1000-measurements time series and the 500-frequencies-and- phases spectrum are interchangeable representations (holding the same information content).

Then it gets complicated; the picture on your TV/video screen is a large array of changing hue/brightness/saturation triple quantities, and it too can be re-represented in lots of ways so a digital description can be squirted through the RF airwaves and reconstructed. The screen can be made into a mosaic of patches, each with some information that changes (and is retransmitted often) as well as other information that doesn't change (and is retransmitted less often). Patches with high spatial frequency (lots of detail) can be rendered with extra accuracy, patches with high time-variance can be given faster updates (but color accuracy becomes less important). Background stationary objects get the most color refinement but the slowest re-transmission of edge positions.

The important thing to remember, is that this is all manipulation of measurements, of the kind of information that has an error estimate attached to it: as long as your re-doing of the info doesn't amplify the errors, it's 'legitimate' for whatever purpose drives you. The use of a logarithm scale for plotting a function, or a frequency scale for representing music, it's all just a mathematical rethinking of the same information, in possibly more useful forms.

Sorry, it's one of those "maybe some day" kinds of things. Most of the coding effort in a big project goes into the user interface, which would mean a major effort to convert from the Windows API to something for a Linux GUI.

You might want to check out my "Gut-Level Fourier Transforms" series at .

That should provide you with the basics, though not the complete math to actually write a full FFT. You *could* do a Discrete Fourier Transform (DFT) with no more than this, but DFTs are pretty slow by comparison.

I don't have a good Web link for basic FFT code, but I haven't looked. I learned about FFT code back in the days of paper, when dinosaurs walked the Earth. A good source, if it's still available, is Hal Chamberlin's "Musical Applications of Microprocessors", which contains explanations and complete working code for FFTs. The code is in old-style BASIC, but you can easily convert to your language of choice.

Best regards,

Bob Masta DAQARTA v5.00 Data AcQuisition And Real-Time Analysis

Scope, Spectrum, Spectrogram, Sound Level Meter Frequency Counter, FREE Signal Generator Pitch Track, Pitch-to-MIDI DaqMusic - FREE MUSIC, Forever! (Some assembly required) Science (and fun!) with your sound card!