Technical modifications on Alesis MultiMix 4 USB - or how to turn an ultimate piece of crap into an usable USB ADC/DAC
Alesis MultiMix 4 USB is an affordable mixer with interesting properties. It has strong metal casing. Two microphone inputs with phantom powers. Four line inputs. Two line outputs. Stereo 16 bit 44100 Hz ADC/DAC through the USB.
If only .....
The distortion performance of Alesis Multimix 4 USB in it's factory condition is pathetic.
If you're an audiophile, close your eyes now ! You cannot bare this.
Below is FFT response on a 36 Hz sinusoidal signal at 0 dBu level. Signal is looped from Left output to Input 4.
2nd harmonic is at -58 dB and 3rd harmonic -64 dB below the fundamental. Also there are higher order harmonics as many as you can count. Even more serious are the strong spurious signal modulations at 1 kHz and its harmonics. Terrible.
Below is distortion frequency responses measured with exponential sweep at 0 dBu level. Signal is looped from Left output to Input 4.
Due to the 1 kHz spurious responses the distortion products are mixed all over the place.
Ok, things look much better now.
Below is FFT response on a 36 Hz sinusoidal signal at 0 dBu level.
2nd harmonic is lowered to -81 dB and 3rd harmonic to -99 dB below the fundamental. Higher order harmonics are below noise level. Also the 1 kHz spurious responses have been removed.
Below is distortion frequency responses measured with exponential sweep at 0 dBu level.
Responses are clean. 2nd harmonic distortion is below -80 dB at low frequencies, and getting below -90 dB at higher frequencies.
As it can be seen from the measurements there are two main problems
The root cause is bad electronic design.
Fortunately it turns out both of these problems are relatively easy to remove by simple circuit modifications.
The source for high frequency spurious signals can be traced to the USB ADC/DAC and to the implementation of the biasing the chip PCM2900B. The 1 kHz spurious signals correspond to the USB packet intervals, and they are leaking to the internal reference voltage of the ADC/DAC.
In the initial implementation the power is taken from the +5 V USB line. According to the data sheet of PCM2900B this is a bad solution [1]. (One can only wonder why this is done)
An improved high performance implementation involves adding an external voltage regulator to the pin VCCCI. By forcing pin VCCCI from outside disables the internal bias reference generated from +5 V USB line and allowes to use a high quality regulated voltage. See the picture below. The voltage should be between +3.6...3.9 V. The current drawn by the pin VCCCI is about 10 mA (I measured).
Below is the voltage regulator I used. It's a shunt regulator. The 'Neosid 39k' is actually 39mH coil, but it could be replaced by 330ohm resistor.
Another modification is to increase the value of VCOM filter cap (C2 in the data sheet), which according to the data sheet must be 10uF but initially is only 1uF (and why is this). This is easy to add.
Below can be seen the regulator connected to the circuit board. The regulator is wrapped in insulation tape and placed on the top side where there is enough of space.
(Click the pictures for higher resolution)
It turns out the main cause of nonlinear distortion is how the LED bar is connected to the signal line. The rectifier diode is connected directly to the signal line without any buffer. This distorts the signal line as nonsense.
The easiest remedy is to bypass the LED bar completely. (The more advanced solution would be adding a external buffer between the signal line and the rectifier, but I don't care about the LED bar so much to implement it)
Below is a reverse engineered schematic of the problem area. This is the right channel. The left channel is identical.
The solution is to remove the capacitor C512 and to take the signal from the output of IC503 to the board connector pin 6 directly. For left channel remove the cap C511 and take the signal directly to board connector pin 5. The board connector pins 5 and 6 must be cut to insert the signal.
(Click the pictures for higher resolution)
Initially there is an internal loop from analog inputs to the analog outputs. Ok, this is a mixer. But removing this feature allows the device to be used as full duplex synchronous ADC/DAC via USB. After this modification signals inserted into any of the inputs 1, 2, 3 or 4 do not appear at the main outputs.
The internal loop can be removed by removing the resistors R503 and R504. See the picture below.
(Click the pictures for higher resolution)
Initially there is variable gain for channels 1 and 2 for both the microphone and line inputs. For the line inputs 1 and 2 the gain can be varied between about -14 dB ... +42 dB. For the microphone inputs 1 and 2 the gain can be varied between about +1 dB ... +57 dB. This is too much of gain (and why there is). It is impossible to set the gain pot accurately.
Lowering the gain to reasonable value is in order. I find suitable values for gain to be between 0 dB ... +15 dB for the line inputs and between +15 dB ... +30 dB for the Microphone inputs.
These values of gain can be achieved by changing the following resistor values:
R109: Initially 3.3 kohm -> new value 47 kohm
R115: Initially 3.3 kohm -> new value 47 kohm
R116: Initially 4.7 Mohm -> new value 470 kohm
R117: Initially 3.3 kohm -> new value 47 kohm
R209: Initially 3.3 kohm -> new value 47 kohm
R215: Initially 3.3 kohm -> new value 47 kohm
R216: Initially 4.7 Mohm -> new value 470 kohm
R217: Initially 3.3 kohm -> new value 47 kohm
See the schematic and the corresponding part of the board for details.
(Click the pictures for higher resolution)
I didn't have 0805 size resistor at hand, so I put 0603 as you may notice ☺
This is truly strange. There appears to be several electrolytic capacitors in the signal chain where there is no need for them. Capacitors placed in series with the signal line between two points both having potential at the reference ground.
Let's replace them with wire ☺
The following caps are unnecessary: C503, C504, C507 and C508. There can be others, but I'll leave this for now.
(Click the picture for higher resolution)
Here's the reverse engineered schematics of almost all of the parts.
(Click the pictures for higher resolution)
(Click the pictures for higher resolution)
The performance of Alesis MultiMix 4 USB can be improved significantly by some simple circuit modifications.
After the modifications this device has turned into a usable audio USB interface.
[1] Data sheet PCM2900B STEREO AUDIO CODEC WITH USB INTERFACE, Burr-Brown
© Elias Pekonen