James Long Company
Lab Tips

1. James Long Company distributes updates in the ".ZIP" compressed form. Windows has built-in support for decompressing ".ZIP" files.
2. The magnitude of 60 Hz (or other environmental noise) present at the bioamplifier output is influenced by the following considerations.
   • First the magnitude is proportional to the 60 Hz (or other noise) field strength present both "common mode" and "differential mode". Second, fields may be either induced into loops or dipoles, the effects of each type of field is different as are the remedies. Common mode v. differential mode and loop v. dipole are discussed below.
   • Common mode voltage is that induced voltage present at both the plus and minus inputs of a channel referenced to ground. Common mode 60 Hz is present because all instrumentation (differential) amplifiers require a ground reference connection for the amplifiers to work correctly. (This is because instrumentation amplifiers can tolerate common mode voltages of only up to the internal power supply voltage, usually a few volts; without a ground connection the common mode voltage might rise to tens or hundreds of volts.) A mathematically ideal amplifier would not pass any common mode voltages because the amplifier is supposed to amplify only differences between the plus and minus input; the perfect amplifier should not pass voltages that are differences between both inputs and ground. Real amplifiers do pass common mode voltages. Common mode 60 Hz at the inputs can be minimized by dressing the ground lead close by the other electrode leads and choosing a ground site that is nearby all the other sites, e.g., AFz. (If only ECG were being recorded then the best ground site would be midway between the two active electrodes.) The effects of 60 Hz (or other frequency) common mode noise vary greatly between amplifiers. The deleterious effects of common mode noise are specified by the amplifier manufacturer as a logarithmic ratio. Our bioamps have a Common Mode Rejection Ratio (CMRR) of roughly 112 dB measured at 60 Hz. Older amplifiers have a CMRR in the 80 dB to 90 dB range; this means the effect on spectral power of common mode noise is 150 to 1500 times worse on an older bioamplifier; hence, the required but distorting 60 Hz notch filter. Note: some amplifiers specifications report CMRR at alpha frequencies; this greatly overstates CMRR because CMRR deteriorates with increasing frequency. CMRR also deteriorates with greater impedance mismatch between the plus and minus inputs, a specification rarely reported by bioamplifier manufacturers.
   • Differential mode 60 Hz is that induced voltage that presents a potential difference between the plus and minus inputs of a channel. Amplifier design can not change this voltage. This is the very voltage the amplifier is supposed to amplify.
   • Other factors that also directly affect 60 Hz noise magnitude are:
     • Electrode lead dressing and induced loop voltages.
       • Including all low-level signals in one ribbon cable greatly reduces the area of any electrical loops in space. This argues for use of electrode drops from the cap for facial EMG and EOG signals.
       • Further, the effective area of any loop can be pushed towards zero by adding a few twists to the ribbon cable as it runs from the subject to the headbox and from the headbox to the bioamp. Each twist results in induced voltages 180 degrees out of phase with the adjacent twist, so induced signals tend to cancel out.
       • Note that induced loop voltages do not directly depend on electrode impedance.
     • Electrode impedance
       • Impedance mismatch between the plus and minus input of a channel (as stated above) diminishes CMRR. So, while an amplifier may have a published CMRR of 112 dB, the actual CMRR may be far below that if the impedance mismatch is large enough. In labs with high noise levels this argues for neither collecting referenced to A1/A2 nor for re-referencing to A1/A2 since it is harder to achieve as low impedance in A1 and A2 as scalp sites.
       • Dipole voltages do not depend on closing of a loop. They can appear as a voltage difference between two wires unconnected to each other. Since the ideal bioamplifier has infinite input impedance, the bioamplifier would see the full amount of this induced voltage. In practice, the two wires would be connected to electrodes attached to the subject. Thus, most induced voltage would be shunted through the subject instead of appearing at the bioamplifier inputs. The degree to which any induced voltage is shunted through the subject is inversely proportional to electrode impedance. This alone is the major reason for attempting to reduce electrode impedance. Recall that induced loop voltages will not be affected by electrode impedance. Indeed, if a small coil, e.g., a transformer, were near the electrode leads inducing a voltage into the lead loop, no amount of electrode impedance reduction would reduce the induced voltage.
       • Electrode impedance used to be an important factor in obtaining accurate voltage measurements. Ancient amplifiers had input impedances of one megaohm or ten megaohm. Our amplifier has an input impedance well in excess of one gigaohm. In the old days, if one electrode, say F3 had an impedance of 1 kiloohm and the homologous electrode, F4, had an impedance of 11 kiloohms and the bioamp had an input impedance of 1 megaohm then the bioamp's load on the scalp signal would attenuate the F4 site by 1 percent voltage (2 percent in power) relative to the F3 site. This would tend to confound any hemispheric laterality study (indeed it could confound any study). If the bioamplifier input impedance is very large relative to the electrode impedance, e.g., one gigaohm v. ten kiloohm (or even one hundred kiloohm) then electrode impedance no longer impacts measured voltage. Such is the case with our amplifiers.
     • Bioamplifier intrinsic noise.
       All bioamplifiers add a noise voltage to the signal. This noise is primarily thermal (3/2 kT noise) and, thus, is spectrally white or flat. This noise is usually specified by the bioamplifier manufacturer. The amplitude units would be in volts per radical hertz. (Less commonly, the noise would be specified in power units, e.g., square volts per hertz or watt ohms per hertz.)
       Care must be taken to ensure amplitude specifications are comparable. First, specifications must be adjusted so that the measurement bands are comparable. Second, amplitudes must be adjusted if one specification is in peak-to-peak amplitude and another is in rms (root mean square) amplitude. Amplifiers have other noise sources, e.g., 1/f noise; this noise tends only to dominate at very low frequencies and, thus, only affects protocols such as CNV studies.
3. There is no amplitude measurement reason for matching impedances between electrodes. Also, electrically linking A1 with A2 has largely been abandoned; that practice required nearly perfect impedance matching. Instead, researchers are deriving A1+A2/2 algebraically, e.g., with EEGTRANS. Finally, if the reason for matching impedance is in some hope of matching induced noise levels, this hope is for naught because, as stated above, the induced noise voltage has very much to do with lead dressing and orientation in space, so the homologous sites will never match in noise. Thus, a better rule is: if the noise power is essentially insignificant in the band of interest (either directly or via aliasing), it does not matter what the impedance differences are; the goal should be to establish impedance limits that ensure insignificant noise levels.