NEWS: a page with sound samples - link below.

    The Behringer ADA8000 is an 8-channel in, 8-channel out Analog/Digital/Analog converter which costs under US$200. It has already a rather ok sound from factory, but its simple architecture allows some modifications to be easily done that can make it sound like a US$2K converter, and with a little harder work, an interface that sounds like no other. There are many reports in forums and pages. A simple example is to bypass its input preamps. If you feed a line signal to its inputs, it takes a small part of it and passes through its microphone preamplifiers. No matter how good a preamplifier is, and this is not the case in a budget unit, this signal-to-noise relationship gets much worse, as the signal is amplified back to the same line level, but now with an amplified input noise acompaining it. Another big issue commented by everyone is the power supply, with regulators running so hot you cannot touch them.    

    This is the second Behringer ADA8000 I am modifying, and I created this page to put some documents and information that may be useful to anyone interested in doing similar things. My motivation was to beat the sound of my expensive Metric Halo ULN-2. When I first compared the sound of my synthesizers connected directly to the power amp to the sound passing through the computer I found that the AD/DA loop completely kills the beauty of the music. The ULN-2 is a good unit, have nice Linear Technology opamps inside, but perhaps too many opamps and a lot of cheap coupling caps. The sound of the first modified Behringer recovers at least half of the original sound.

    The modifications I did in the first unit, whose owner is Marcello Sfoggia, to record classical concerts in my city, were:
    - replacement of ceramic caps at the power supply for high isolation voltage ones;
    - installation of a fan on the top cover to cool down the regulators that run very hot (Sfoggia did that);
    - bypass the input mic preamps, going directly to the main board with appropriate resistors;
    - opamps replacement: I put LT1359 at the inputs and OPA4134 at the outputs;
    - uncrossing the outputs, because the positive passed through two opamps, and the negative, only one.

    With the mods the unit sounds much much better. Then I started working on a second unit for myself. My goals were to avoid all the problems with the built in power supply again, to compare OPA4227, OPA4134 and LT1359, and avoid the signal coupling caps. From my experience, every time you can get rid of a coupling capacitor the sound gets to a new dimension. But it is of course a big challenge to make some circuits DC-coupled without getting into trouble. 

    Below you can find:

    - Link to the schematics by Gyraf Audio. The input coupling capacitors are inverted in their schematic. They have the right polarity in the actual unit, with the positive side at the AL1101 chip, which has 2.5V DC bias at its inputs.

   
file:Behringer_ada8000_analouge.PDF

    - My annotated picture of the power supply (click on int for a larger size). I have not figured everything out, and you are encouraged to help me find out why there are two pins with +50V going to the front panel, for example. As you see the bottom 7815/7915 regulators serve the microphone preamps in the front panel, which I disconnected, and they are responsible for the largest power consumption on these regulators. If you just cut the +-15V supply wires that go to the front panel these stop running hot.

   


    A little bit of history

    This page started one year ago, but it took me some time to resume working on these circuits due to my other activities. I got involved with the DC-inputs again just a few months ago, out of necessity, and now I can include new information and some important conclusions below. But let me first highlight how I am testing them. After the first measurements I did of frequency response, phase, distortion, I observed that this data alone could not tell me how the circuit would sound. Check for yourself in the graphics below, but as I see now they are kind of complementary, in the sense that the measurements tell me what I cannot easily determine by ear, while listening tells me how the circuit sound, which cannot be understood by looking at the data. And as the specifications come very nice from the factory, yet sound bad, most of my attention was given to listening tests.

Frequency Response 100db  
Frequency Response 3sb    FFT 1KHz

    I also have a relatively uncommon condition of testing the recording (in fact, always the complete AD/DA loop) with my keyboards. This is important because it allows me to compare it to the sound going straight from the keyboards to the power amplifier, which I take as the gold standard, for it is actually quite beautiful, full of natural colors, good depth of field, good stereo image, with uncompromised bass, and so on. That is why I stated at the beginning that the AD/DA loop of stock interfaces killed the beauty of the sound - it is beautiful before recording. It is of course possible for anyone in a studio's control room to compare the analog original sound to the AD/DA loop with a source like vocals, orchestra, or someone playing guitar, but it is very unlikely that this person could be making modifications to a circuit, changing resistors and voltages with those live sources at his disposal! On the other hand, I am in some sense very limited as I do not know how the circuit sounds with vocals, mainly, as I have not tested recording them yet. In fact, some artifacts do not compromise instrumental sounds as much as they do to solo vocals or choirs. But in the end I would still say that as I am not only appreciating the sound but comparing to the original, I am mainly in the right direction.
   
    You can see parts of my small studio in the pictures below. I am listening with a pair of home made speaker cabinets that contain a 10" weak bass driver, a 5" bass-mid driver from an Event SP5 speaker and some soft dome and paper cone tweeters each, and a pair of passive Alesis Monitor One mkII. Each system is driven by a brazillian Nashville Power-250 amplifier - this is an old hi-fi model that does not have any harsh common to other transistor amps. The big cabinets give me more bass and body, transforming the precise yet irritating sound from the Monitor Ones into a full and pleasant sound adequate for playing with satisfaction.

    The new input circuits

    My goal was to feed the purest line-level signal to the AD chips, but it is not possible to connect anything directly to the chips as they operate from single 5V supply, thus requiring the signal to be centered at 2.5V. The first SMD board that I made to perform the level shifting looked very pretty, but had at least one big flaw: it could not accept a floating signal connected to it.  When connecting an output isolated by a capacitor to that circuit, the 2.5V level at the opamp's output developed the same voltage at the inputs because of the feedback, messing up with the voltage addition. So, I simulated a new circuit using this applet simulator from Paul Falstad and implemented the circuit below, that has +1.25 and -1.25V references that bias a possible floting input correctly to 0V. This input does not work with a DC-coupled circuit that has a DC level other than 0V with low impedance, of course. The second purpose of using the opamps is to invert the signal to feed the negative input on the AD chip, and I did this in parallel, so as to keep both polarities going through the same number of stages. If I only worked with 24bits I would perhaps try using just one polarity, but as I have a very good 16-bit DAC that I want to try when playing back, loosing the most significant bit is a serious concern.

    Below are the schematics
for my new DC-coupled inputs as they are working now (click for the pdf). Both circuits take an unbalanced input to generate positive and negative polarities. The left version has unity gain but does not limit the input voltage and I suspect that one could damage the ADC inputs with very high levels connected to them. Of course you should identify hard clipping in your software meters and speakers long before any damage occurs, but that might still be a concern for someone. The version at the right uses a little adaptation trying to avoid that. It takes the input signal, multiplies it by 2, than divides the outcome by 4, having a gain of 0.5. This final division would take the maximum opamp swing down to a level of +-3.5V, which would add to the 2.5V bias resulting in voltages from -1 to +6V at the AD chip at most, and I don't think these can cause any harm. I still think the circuit at left provides a better sound, but this is hard to tell as I need to test them with different input levels, different volume levels at the power amp, and even the output LP filter is not the same. For myself I am going to stick with the first, and I would recommend the safer one to anyone working outside a studio and that would connect unknown sources to record.
 

    schematic    safer

    You can use the PCB layout I designed for the circuit at the left, if you like, for non commercial purposes. Yet for the right one you have to find where to put the 660ohm resistors. Click on the pictures below to get the postscript files for the single cooper layer and silk screen, and a pdf with a print of both sides, easier to look at when placing the components. You would need to place two small SMD resistors of 0 ohms or make a short for the unused opamps at the center. There are two wires needed to run the ground to the center parts, and a resistor marked as 500ohms also corresponds to a short in the schematic, making that opamp a voltage follower. Do not place any capacitance at bottom-left and top right. There is also a spare ground pin for each output, which I don't think will be ever used, as we have positive and negative outs. There are a few places to put small bypass capacitors, which I am currently not using. The ones at center-leftmost and center-rightmost would be used to correct opamps from overshooting or ringing.

    pcb    
print     silk


    RESULTS

    Now let me describe the exciting part. This board sounds awsome! Have to listen to believe! The DC connection, by eliminating the coupling capacitors, makes the sound free, uncompromised, as if it was set free from a can (the cap can) in which it was confined before. Remember also that I am listening with a DC coupled output as well. As for opamps, I compared the circuit with LT1359 and BB OPA4227 in, and they sound terribly different. I like the LT1359 most by far. They are cleaner, with deep and precise bass notes, preserving the color and richness of instrument tones at mid frequencies. On the other hand, the OPA4227 sound a bit ressonant, making the bass bigger but more difused, giving the impression that the sound is duller. That's a similar impression that the OPA4134 gave me when comparing them on the ADA board (with coupling capacitors). It is also a bit close to what Jensen transformers give me when connected directly between the source and ADC chip (more below), yet worse in the case of the opamp. I must say that the LT1359 are not perfect. They do sound incredibly pleasant, almost better than the original sound, but I would say that they are a bit too flat, and loose just a hint of concreteness and stereo image compared to the source signal, perhaps highlighting the highs. That is why I am looking forward to test John Hardy's 990 or other discrete opamps. Not that the sound now is bad, I insist it is really nice and I could live forever with it, but I can understand the differences from the original and other opamps can perhaps surpass the 1359. And that's also why I will keep a pair of inputs with just the Jensen JT-11P-1 transformers to isolate and shift levels, and no opamps. They also make the bass fatter (in the worse sense) and color the signal, as if I was listening to something that ressonates (again, not a a good coloring), but they do preserve better some sound properties in the mid frequencies and sound coherence, and in this way some instruments from the keyboard sound closer to the originals.

    Next there are a few pictures of my test enviro
nment, tentative circuits in the protoboard, PCB fabrication (for it is fun) and experiments. 
   
pic pic pic pic pic
pic pic pic pic pic
pic pic pic pic pic
pic pic pic pic


    Finally, I could manage to provide the first files with sound samples comparing different units and the first mods:

    - SOUND SAMPLES PAGE

    More to come - this is what I expect to include here at some time:

    - basic measurements showing the slightly increase in bass response and low noise/distortion

    What I did not like in this circuit and board:

    - the board has room for big capacitors at reference voltages, but you should not use them;
    - the board does not have a place for the division resistors that go from positive to negative outputs;
    - the 0 ohm resistors can be omitted once I decided to stick to the LTs to generate references, as they are good for it.

    Things I will probably work on next:

    - an SMD board with the same circuit
    - testing 990 opamps and Holco/Rikken resistors
    - a circuit tha accepts differential inputs
    - more tests and changes in the PSU: shunt regulator, bigger caps (and soft start), ...
    - other circuits: outputs (DAC), other AD/DA chips, my tracking mixer, I need a simple microphone preamp, ...


    More information and results will be posted in a near future.
   
Porto Alegre, Brazil. johann at inf dot ufrgs dot br 
    Created on April 24, 2010.
    Updated on April 12, 2011.
    Graphics on September, 2013.