Thank you very much for your exceptionally clear answer. I apologise for not acknowledging you as first clarifying that aliasing was not an issue (looking back many people said this in at least a roundabout way) I missed that as I was overloaded with information. Prob. use an Analog Devices anlog MUX but shouldn't be vastly diff. from standard 4051 (an I would hope if it was diff. it would be for the better). I am prob. using an Analog Devices ADC with inbuilt 8ch MUX so filtering before the MUX's would still require 8 filters. So filtering at the source for 60Hz sounds best. Looking at ADC's with appropriate specs in other senses (like package, supply voltage, cost etc.) about the only options are a 2MSPS 8-bit ADC or a ~200kSPS ADC, I havn't been able to find many ADCs between 200kSPS (a little too slow for 32+ inputs) and 1MSPS (a bit of an overkill). Hence the high sample rate came about more due to the possibility than the need. Hence it could be brought down if this simplified design. One of the other advantages of a high speed ADC is large oversampling. While I had considered sampling at high rates the maximum rate for sending MIDI messages I would consider would be about 1500 msg\s as this is all MIDI can handle (A 7-bit knob message requires 2 bytes which at 31250bps (1 start, 8 data, 1 stop) is ~1500msg\s. And that's if only one knob is active. If you have say 5 knobs (try tweaking 5 knobs at once:) ), allows 300msg\s or 1 message each 3ms. Hence I would be looking at sending each knob every 1-3ms (assuming it's changed). So I would prefer that each knob was updated every 1ms max. This would require 32kSPS (32knobs), with no oversampling. The other advantage of a high speed ADC is the ability to do high oversampling. If, I'm sending every knob whenever it changes I can't afford to have knobs change values when they shouldn't or I shall quickly run out of MIDI bandwidth. With say 5x oversampling, I need 150kSPS. So, theoretically that would work on a 200kSPS ADC, but you could neither add more knobs (would like that option) nor afford any wasted time. The other advantage of a high speed ADC is the ability to do high oversampling. If, I'm sending every knob whenever it changes I can't afford to have knobs change values when they shouldn't or I shall quickly run out of MIDI bandwidth. Oh and finally I would like to apologise if I in any way contributed to the degeneration of this topic through repeated long questions and little understanding of the topic. BTW: I would appreciate anybodies experience with the Analog Devices Microconverters (ADC and DAC with embedded Flash MCU (yep, they call the MCU embedded not the ADC)) as they could be very good for this solution. Thanks all, Tom. ----- Original Message ----- From: Mario Thomaidis Subject: Re: A/D Challenge Summary > Thomas, > > I had indicated the same thing as below several messages ago. > Anti-aliasing is only a concern when trying to reconstruct a signal. > Since you have no intent on reconstructing the signal, you have no need > for an anti-aliasing filter. > > To eliminate the 60 Hz hum you would need to either filter *each* pot > *before* it enters the multiplexed (4051, correct?) or at the voltage > source. > > I will admit that a 1 kHz sampling rate seems high to me, but it is your > project and you know it's requirements better than anyone else. I will > also say that sampling at less than 60 Hz seems much too slow (musicians > can be very exacting and fickle), so that although you may be able to > inexpensively filter out the 60 Hz hum in this case, you are doing it at > the expense of your original intent for the circuit. > > To eliminate digital ground signal noise, your filtering needs are very > much less - a filter of ~10 kHz will probably work just fine. This one > can be located after the multiplexed since the signal at that point will > be lower in frequency content. In this case you will require only 2-3 > low pass filters (at each multiplexed) of minimal order (first should be > more than fine). Cost 2-3 resistor and 2-3 capacitors. > > Of course, you may just take Harold's advice, and make sure your ground > plane is laid out properly - this, for my case, may be the simplest > solution. And if you do get a bouncy signal, just implement a simple > digital filter or do some averaging. Cost, a well laid out PCB or some > filtering code. > > One last thing. I have to agree with Lance. The etiquette in the > conversations seems to be degenerating for no apparent reason. > > Mario