ADC multiplexing

lts

heck out

e is a

tifacts.

nce,

rential

g van Rijn

I=92m a =93johnny come lately=94, but this job isn=92t difficult.

Put the effort into the mux/preamp and use a single a/d.

It=92s not like this is a blazing fast acquisition system with high precision. This is practically DC. These days a good calibration routine can help solve a lot of hardware issues.

Like others have said, the op=92s is looking for parts and doesn=92t have the system knowledge to make component choices. Her real challenge will be when she doesn=92t get her expected results.

Then look at Analog's multiplexers. Those cheap CMOS mux are real crap. We had them in one product at the pre production level, but they gave us fits. We switched to some low end video mux from ADI, even though the specs were overkill and had no more problems.

Very true, the ADC part is easy once you've got a good amplified signal without CMRR effects etc.

When doing calibration or working otherwise with EEG-level singnals, I've had no major problems with simple 74HC405x-series multiplexers. If using these in the electrode path, take into account that different manufacturers (or processes..) have different leakage (ESD diodes?), which might degrade your input impedance.

If your concern is partly about the errors in the multiplexers, you also need to specify the impedance (of the input signals (and to some extent the loading by the ADC also). It's usual to buffer small signals like this with some kind of differential preamplifier, which you don't mention.

Real-time simultaneity is a hard task for a micro; I'd consider a dual-port RAM or FIFO if this part of the scheme becomes a bottleneck.

Maybe not a great choice. Battery power isn't accurate (that'll be 3.3V plus .1V or minus 0.3V), and the low voltage will preclude some interesting hardware options. DC/DC conversion is easier if you start with 10V or so.

That implies a transmitter right next to the microvolt input signals? A micro can easily handle the bluetooth without any memory hardware support, I suspect, because it will be separated short bursts (no need to interrupt-handle every bit).

the

How do the ON-Semi 74HC405x stack up? I did the latest redesign of those parts when ON-Semi went to using Chartered for processing. Still have those files on this laptop ;-) ...Jim Thompson

without

e had

se in the

sses..)

se

those

We don't disqualify it just because you designed it. However, we were looking more in the range of 500MHz, rather than 100MHz frequency responses. For the OP, it should be fine.

The first thing you need to do is be realistic. The ECG signal changes a fair bit with minor errors/alterations/changes in the placement of the electrodes/skin conditions/weather. Therefore 16-bits is a total waste of time. 8 or 10 bits is the maximum you will ever be able to properly utilise. Adjust.

t the

lts

And the patient is not going to be running around. Why hook it up wirelessly? Anyway, FDA, FCC and XXX will never approve such scheme.

Actually for long term monitoring the patient will be moving around. When my son was doing this years ago he had to drag a cable around behind him and was limited to about 20 feet of movement from the cart they had the equipment on. They had the rooms arranged around a central lounge so people could get together and visit. A wireless solution would be much better.

I did not measure ON Semi parts, only the NXP 4053 because I wanted to use their tiny DHVQFN packages. They passed the I_leakage

They're generally the body diodes of NMOS and PMOS respectively, often with an additional R-C-biased NMOS device that "snaps" on upon an ESD event. So you can't count on balanced leakage. ...Jim Thompson

Thanks, I'll keep that in mind!

Hi,

Option 1:

If I choose to work with AD7699 than I need preamp stages before the ADC. Am I correct? If yes, than according to the data sheet the input voltage range will be Vref/2 to -Vref/2.

Now , the thing is that the input EEG signal is 10 to 100uV. I need to design the ampifier that can amplify the input signal (10 to 100uV) and match it with the input voltage range of the ADC. So, If Vref is

-1.25 volts. So, I need an amplifier with a gain of 12000 to get the input signal to match the ADC input range. Am I right?

Option 2:

If I choose AD7606 than i do not need to worry about the preamp problem plus the ADC can work with 3.3 volts and has a parallel interface. The data sheet says input voltage range is +/- 10 volts. But my supply is 5.0 Volts. So, I suspect that the input voltage range will be +/- 2.5 volts if the Vref is 2.5 Volts. I do not see any amp gain information in the data sheet. Any adivce!

thanks

jess

Right order of magnitude, but ... "signal is 10 to 100 uV" could mean PEAK values, or RMS values, of an AC fluctuation; if it's RMS, then you need some allowance for the crest of a wave being higher than its average (and 5x is a typical crest factor). If it's PEAK, then you must deal with the fact that the differential gain becomes zero when the ADC is saturated, so it's a little like dancing on the edge of a cliff: the closer to the edge you get, the higher you are. Until you drop off the edge one day.

In theory, white noise has the most stringent 'crest factor' requirement, being essentially infinite (if you can limit bandpass and accept a small dead time, you can make do with less). Best practice would be to quantify the falloff of useful information as a function of too-low and of too-high gain, and park at or near the peak of those intersecting asymptotes.

Maybe, too, the different inputs will have DC offsets; if DC offsets are presented to the multiplexer inputs, the amplified signal will have amplified DC offsets as well. You can't use a high-pass filter to reject those offsets, unless you use one filter per input.

Why not? You don't want to feed uVs or even mV to ADC, with or without mux.

Yes, you'll need a preamp with almost any ADC. As a first amplifier stage yo could use INA126 with 5-10x gain so that electrode offsets will not rail the output. After that a suitable highpass (0.1Hz-1Hz?) to remove the offset, more gain (10x+1000x gets you to 1Vp-p with 100uVp-p input) with a suitable lowpass (35Hz ?).

After the preamp+lowpass you're working with band-limited signal with reasonable amplitude and you can use simple MUX+ADC combination with reasonable specs. Using 4 74HC4051 MUXes and 4 single-channel ADCs should be ok.

After the MUX, use a successive approximation (SAR) ADC. Most of the sigma-delta converter do not settle to the next channel value on single cycle.

Also, MUX+sample fast enough to avoid aliasing of unwanted signals; the highpass filter is not a brick wall and will pass some higher frequency signals.

At these signal levels, it will be almost required that you use individual instrumentation amplifiers on each channel, and only multiplex the amplified signal. This will greatly reduce the headaches of making a multiplexer that doesn't obliterate the signal with crosstalk and switching transients.

Linear Tech has a new LTC1864/1865 series of serial 16-bit ADCs that are incredibly small and low power. We are using the most significant 13 bits in our present application, and they are doing very well. Later units may use more of the bits, we will see how good they look then. These ADCs are about $8 each.

Jon