This is probably the least understood point in any car stereo-- Its CRUCIAL, its a link in the chain that actually occurs right at the hook, but almost no-one can explain how it is that a D/A convertor takes a series of on and off signals and makes them analog waves. Also, the couple of people that can, always are stumped by the question "How can a 1 bit or 0 bit DAC work on the same program material that a 20 bit DAC works on"
How about a quick explanation of the ins and outs of DAC, be scientific, confusing and very VERY detailed.
Thanks!
ShadowStar
------------------ HEY! I'm looking for a CHEAP used Clarion Thunderdome.. Hook me up!
If the sub moves, use it. If the tweet's metal, lose it. If you car doesn't have Oz, I just can't excuse it.
Posts: 2578 | From: Somewhere In the Northeast | Registered: May 1999
| IP: Logged |
posted
You're making a guy work hard on his day off.
I can give a basic (not real confusing)explanation now, but I will wait and look over all the info I had gathered on D/A's. Then I will compile that into a technophile's dream of incomprehensible jargon and acronyms that no one will be able to understand, not even me.
First I will go over a couple things for background info
CD audio data is 16 bits long and is transferred one bit at a time at 44.1 thousand samples (full 16 bits) per second from the device that reads the data, you cannot get around these facts, although audio formats with longer bit lengths and higher sampling rates are out there. Oversampling is the process of taking two consecutive samples and interpolating more samples between them. This provides the D/A converter with a higher sample rate, and often makes the samples a longer bit length than the 16 bits. For example the Analog Devices AD 1893 will take the 16 bit 44.1kHz data, and output a 24 bit higher frequency signal.
OK now I will start from the beginning, the laser beam reads the pits from the CD and sends that info to the D/A converter. First the D/A will take the data that is coming in one bit at a time and store it into a shift register. This takes the serial data and converts it to parallel data. So now all the relevant audio data is shown at the output of the shift register. At this point all the non audio data is taken out and diverted to whatever devices need it. If there is any oversampling to be done it will be done at this point. And in the same manner it will be presented in a parallel format.
Now for the actual conversion to analog. If you know the basics of binary numbers you will understand this. Each of the bits will either turn on or off a transistor, if the bit is one it will turn on a transistor, a zero will turn it off. Depending on which bit is controlling it the transistor will allow a certain current to flow out of it, the current will be different for each bit. The MSB (most significant bit) is worth half of the maximum current, the next is worth 1/4, the next 1/8, down to 1/32,000 in the case of 16 bit data. Also since the analog data has to have positive and negative voltages there will actually be two different 15 bit D/A converters, one for the positive and one for the negative. Which one is used will depend on the 16th bit that controls the sign of the binary number. So now you will have a current that is the sum of all the currents coming out of the transistors. This current is then turned into voltage. This process happens 44 thousand times per second or more.
So now we have a voltage that is one of 65,536 voltages every 1/44,000 seconds this voltage will stay at that level until the next voltage comes along. So if you think about it the output waveform will have steps in it, it won’t be a nice smooth waveform. So we add a reconstruction filter to filter out the high frequency component from the output. If oversampling is used the demands on the filter are reduced since the high frequency component is much higher than the audible spectrum. Now we have a analog reproduction of the original signal.
Now to try to answer some of your questions I don’t know exactly what the companies are talking about since I do not know what chipsets they are using to do their D/A conversion. But all converters receive their data one bit at a time so I guess they could all be called 1 bit D/As. I haven’t heard of a 0 bit D/A, I think if you had 0 bits you wouldn’t have any information. I think the 20 bit D/As refer to the number of bits being processed in the final D/A stage, if oversampling is used and the bit length is raised to 20 before the final stage I would consider it a 20 bit D/A.
posted
Ron, Do know more about the 1 bit vs. 20 bit claims? I know some manufacturers claim 20 bit D/As and some say 1 bit D/A. Quite a few people claim the 20 bit is superior, but is there a difference in technology other than what I mentioned.
Once again I know the technical portion but not the practical.
posted
I have the info at home somewhere. I haven’t touched the DSP stuff in a while. I know the one bit D/A was a little confusing when I read about it. If I remember correctly there are different trade offs for different D/A conversions. I think and I’m not sure I remember correctly but the 1 bit D/A invented by ??I forget, I think was slightly superior. Nothing that sounded like it would be noticeable. I think it was all about getting the signal to error ratio higher. Sorry I dont remember more. My notes are on my computer at home.
------------------
Posts: 1259 | From: Fullerton. CA ,USA | Registered: Oct 1999
| IP: Logged |
posted
Could you send me whatever you have on the subject? Everything I have is mainly centered around the chips I used for the design of a D/A converter. So other than what I have, and the knowledge of what goes on in any D/A chip, I'm kinda clueless.
posted
Ok, now that we have the basics (did you really understand all that??)....
One bit DAC:
2^1 Resolution, at 44.1Khz. Unless a HUGE switching speed can be acheived (upwards of 3.2ghz), combined with lots of oversampling, not great resolution, can be very linear. As Jeremy stated, 65,536 different voltages every 1/44,000 seconds. Like taking a picture with a high quality camera, and printing the pic on a paper towel....
20 bit DAC:
2^20 Resolution, at 44.1 Khz. This basically allows the switching speed to be reduced while still allowing 1,048,576 points in space to resolve the analog signal. Great detail, but, unless a really high quality filter is in parallel with the DAC, Digital jitter can be a problem.
Higher order DAC's definatley can add resolution, the higher the order, the easier a job the interpolater has, mainly 'cuz there are more points in space to "smooth" out of the analog side.
Flame away!!!!
------------------ Ron Hawkins
Sometimes the majority only means that all the fools are on the same side.
Use your ears to judge components, NOT your wallet! K.I.S.S.
posted
Sorry I have been busy for the last couple days and have neglected my Advanced Topics duties. Anyway I don't know anything about the delta sigma stuff. I tried to get some info from Burr-Brown's site but they do not elaborate on what it does. They have a 5 stage sigma delta but I don't know what that is, but it has improved jitter reduction over 1 bit sigma delta. Maybe that is what is meant by 1 bit D/A??? Anyway I will try to find some more info and post anything I find out.
posted
I've found a load of info on DA converters during some recent research. If anyone wants any info let me know and I can email you some .pdf files.
posted
It's been awhile but here's what I remember... 1 bit DAC converts 1 bit at a time, very quickly. 16 bit DAC converts 16 bits at a time, doesn't need to be as quick. 20 bit DAC uses the extra 4 bits for error correction. I could be wrong, like I said it's been awhile. I too am interested in more information, bring it on.
posted
Have a friend who has a pretty good ear for music bought an ID 24 bit D/A converter for his car a few years ago. Ran it, liked it, thought that it was doing something nice, called Autosound 2000 and talked to them. They asked if had REALY heard any difference in the system. He said that he would do some critial listening with it and without it (using 1 bit in the deck), and asked they why. Their reply was that in ALL of the tests that they have done no one could tell the difference between thier 24 bit Apogee converter and a 1 bit deck converter- and they quite selling D/A converters a short while before that. Nice huh?
------------------ RF, Veritas, Image Dynamics
IDMCTHS
Posts: 7866 | From: Charlotte,NC USA | Registered: May 1999
| IP: Logged |
posted
I think like with most everything, it is really the quality of the components and not necessarily the type used. 1 bit and multibit DACs all have their advantages and disadvantages, and there are two types of multibit DACs also. If you have a DAC with more than 16 bits two things can happen, they can auto fill the extra bits with 0's, or there can be a process in the conversion that converts to a higher bit count. Every DAC has dual DACs, because there are two channels encoded in the datastream. Most companies use a DAC that does both channels, some use a separate DAC IC for each channel. Given equally good components I think it would be a wash on which is better.
Would any of you guys have info on the Alpine 3900 D/A Converter. I picked up the entire Alpine Digimax system and I'm interested to see how it compares with today's gear...
-------------------- Knobs Rule! Keep it simple, clean, and loud!
