What are the advantages of one ADC per channel ?
 
Two principle setups are used in multichannel digital acquisition systems. See the diagram below.
   
Shared ADC setup:
the amplified signals are stored in sample-and-hold circuits, an analog multiplexer is used to scan the S/H outputs, and a single ADC is used to convert the signals sequentially to produce a serial output signal.
   
ADC per channel setup:
each channel has its own ADC and the final conversion to a serial data stream is a completely digital process of scanning the digital outputs of all these ADCs.


Until approx. 10 years ago, high performance analog-to-digital converters were relatively expensive components with a high power consumption. This prohibited using more than one ADC for a system. There were problems with the Shared-ADC-setup though. The need for sampling all the channels at exactly the same moment forced designers to include sample-an-hold circuits. Together with the limitations imposed by the analog multiplexer, many problems surfaced in the Shared-ADC-setup.

Problems arising with the Shared-ADC-setup:
1) The dynamic range of the S/H and analog multiplexer circuitry is limited to approx. 80 dB (14 bit). This means that relatively high amplifier gain is needed to lift the input noise above the noise level of the S/H and MUX. High amplifier gain decreases the input range, and thus eliminates the possibility of DC amplifiers.
2) Charge leakage currents in the S/H circuits causes gain differences between all channels.
3) Charge injection by the switching of the analog multiplexer imposes glitches on the S/H output.
4) The analog multiplexer has limited channel separation. Besides measurement errors, this has the nasty effect of high crosstalk to neighboring channels when one of the electrodes isn't properly connected. Not nice when you are measuring many channels.
5) The output of the analog MUX is a staircase like signal, with the sequential S/H voltages. This means that for every next channel, the ADC has to quickly settle on a new signal level. This imposes extreme demands on the bandwidth, frequency response and settling time of the analog input stage of the ADC (and on intermediate buffer amplifier stages which may be needed from preventing loading the MUX output). Effects like overshoot and ringing lead to further deterioration of the sampling precision.
6) On systems with many channels (say 128 and more), proper routing of all the sensitive analog signals to the single analog MUX becomes a formidable task (interference problems become huge)

During the early nineties, researchers kept asking for systems with higher dynamic range (to allow DC measurements), better channel separation, and more channels than could practically be achieved with the Shared-ADC-setup. Advancing developments in low-power, high-bit ADCs and low-power Programmable Logic Devices (PLD) allowed designers of modern multichannel acquisition systems to switch to the ADC-per-channel setup. BioSemi already introduced the new setup on the Mark-6 system in 1995, and has since then continued to improve the setup in terms of miniaturization, power consumption and reduced costs. The ADCs operate synchronously, so no sampling skew is present (all ADCs perform the conversion at the same moment). The multiplexing operation is now performed entirely digital, so any deterioration of the signal is eliminated. The dynamic range of the system can now be really equal to the dynamic range of the ADC. The 24 bit ActiveTwo with its 110 dB dynamic range and >110 dB channel separation is a good example of the advantages which can be achieved by using one ADC-per-channel. Specs like these are fully unattainable with the older setup (30-40 dB worse, a factor of nearly 100).

The important step forward that could be achieved by using an ADC-per-channel and digital multiplexing, forced all serious manufacturers to use this setup for their new designs. Consequently, you will only find the Shared-ADC-setup in older designs.

Customers not acquainted with the performance features of modern ADCs, sometimes raise the question whether the new setup does not introduce additional errors as a result of differences between the ADCs. As they see it, the use of a Shared-ADC-setup at least ensures hat any ADC error is the same for all channels, and thus cancels when the difference between channels is displayed. Although this assumption is not untrue, these customers ignore that the differences between modern ADCs in a properly designed circuit topology are much smaller than the errors introduced by the analog multiplexing circuitry.

A lot of work has gone into optimizing the circuitry used in our systems in order to minimize channel differences. One of the key design features is that we use a Zero reference setup with a common Reference voltage for all (up to 256) ADCs. The circuit board has been optimized (regardless of costs) for identical REF voltage for all ADCs. For example: REF is distributed among the ADCs by low-impedance, low inductance ground planes, a method effectively preventing small voltage drops across the motherboard. Consequently, this is one of the reasons why we insist on having all modules on one single motherboard. You can never attain this kind of precision when you start coupling subsections with for example 32 channels. Another feature of the "Zero reference setup" is that we make a final subtraction of the channels in software, effectively canceling all noise and drift effects of the ADC references. Finally, all the ADCs run on the same master clock to prevent any timing differences between the ADCs (which may effect the conversion).

In a multichannel system, there are always small gain differences between the channels. Part of these differences are caused by resistor tolerances in the amplifier stage between active electrode and ADC, and the other part is caused by tolerances in the analog sections of the ADCs. The difference between the ADCs is much less than 1% (guaranteed by the manufacturer, and checked by us). We found that the dominating source of gain errors between the channels are resistor tolerances in the amplifiers. BioSemi exclusively uses precision metal film resistors to ensure high accuracy and stability over time. We specify an overall gain accuracy of 1% (both absolute and relative). Further software calibration can be added for anyone insisting on better accuracy. It should be noted that the amplifier gain tolerance problem is exactly the same when using a Shared-ADC-setup scanning the channels.

Given the excellent accuracy of modern ADCs, and the way they can be made to perform identical in a properly designed system, the ADC-per-channel setup actually makes it far more easier to ensure equality of the channels than the obsolete Shared-ADC-setup