毛细管电泳高压电源在环境检测与中药分离中的电压梯度
Capillary electrophoresis separates species by the combined action of electro-osmotic flow and electrophoretic migration, and both terms scale with the field applied along the capillary. The high voltage supply therefore does more than hold a number: it sets the shape of the field in time and space, and the shape decides whether two peaks that differ by a few percent in mobility resolve or collapse into one broad band. In environmental analysis the target list usually contains ions of similar charge to mass ratio, and in the separation of herbal extracts the mixture contains dozens of compounds with overlapping mobilities, so the gradient becomes the main lever for resolution.
A voltage gradient can be built in three ways. The first is a step, where the field rises from zero to the working value within a fraction of a second and stays flat until the run ends. The second is a ramp, where the field rises over tens of seconds along a defined curve. The third is a spatial gradient, produced by tapered capillaries or by segmented electrodes that impose different potentials along the length. Each approach places a different demand on the supply, and mixing the demands without matching the loop usually produces drift rather than separation.
For a step gradient the critical parameter is the settling time. The column behaves as a capacitor of a few hundred picofarads in parallel with a resistive path through the buffer, and the supply must charge that capacitance while holding the current limit above the steady-state value. If the current limit sits too close to the steady value, the rise becomes a slow exponential and the first part of the run happens at an undefined field, which shifts retention positions and ruins reproducibility between injections. A practical rule is to set the transient current limit at twice the expected steady current and to keep that window open for no more than two hundred milliseconds, so the column is not overheated while the field is still forming.
For a ramp the critical parameter is the shape, not the duration. A linear ramp applied to a system with a first-order response produces an output that lags the command by one time constant, and the lag is visible as a systematic offset in the calculated mobility. The correction is either to pre-distort the command with an inverse first-order term or to measure the actual column voltage at the electrode and integrate that value instead of the setpoint. The second approach costs an extra divider and an isolated amplifier but removes the guesswork, and in regulated work the measured value is the only defensible record.
For a spatial gradient the critical parameter is the balance between segments. Segmented electrodes divide the column into zones, and the field in each zone is set by the potential difference across that zone. If the segment dividers draw unequal leakage currents, the zones drift apart and the gradient becomes uneven in a way that no single setpoint can correct. The remedy is to buffer each segment with its own follower and to trim the dividers against a common reference at the assembly stage, then verify the trim with a conductivity scan before the column is installed.
Environmental samples bring a complication that herbal samples do not: the matrix is often high in chloride and sulfate, and the ionic strength can vary by an order of magnitude between sampling sites. High ionic strength lowers the column resistance, and a supply sized only for clean standards will hit its current limit and fold back, which looks like a sudden loss of field. The design response is to size the current capability for the worst matrix expected, and to add a matrix detection step that lowers the field before the injection plug reaches the detection window, so the run continues at a field the supply can hold rather than stalling.
Herbal extracts bring the opposite complication. The matrix is loaded with high molecular weight compounds that adsorb on the capillary wall and change the surface charge, which changes the electro-osmotic flow and therefore the apparent field. The supply cannot correct surface chemistry, but it can keep the current trace clean enough that adsorption is visible as a slow rise in current at constant voltage. A current resolution of one part in ten thousand is enough to see the onset of adsorption, and that signal is more useful than any post-run calculation because it appears while the run can still be salvaged by a polarity reversal.
Ripple matters differently in the two applications. In trace environmental work the detector is often a conductivity cell or an amperometric cell, and ripple on the field modulates the migration speed at the ripple frequency, which smears the peak in a way that looks like poor column efficiency. A ripple of one part in a thousand is tolerable when the run lasts a few minutes, but the same ripple becomes visible when the run is stretched to twenty minutes for a difficult separation, because the smear accumulates with time. In herbal work the detector is usually ultraviolet absorbance, and the same ripple produces a modulation of the baseline that can be averaged out, so the ripple requirement can be relaxed by a factor of three without harming the result.
The hardware that satisfies both cases usually shares a common architecture. A push-pull stage drives the column from two rails, one positive and one negative, so the field can be reversed without waiting for a discharge. A current sense resistor sits in the return path with a differential amplifier that has a common mode range covering the full rail, and the amplifier output feeds both the control loop and the safety comparator. A bleed network discharges the column to a safe value within two seconds of a fault, and the bleed current is monitored so a failed bleed is reported rather than hidden.
Safety interlocks deserve more attention than they usually receive. The electrode compartment is opened often during a shift, and the supply must remove the field before the lid is fully open, not after. A reed switch on the lid with a debounce of ten milliseconds and a series relay in the primary path gives a response fast enough that the operator cannot contact a live electrode, while the debounce prevents nuisance trips from vibration. The relay should be a forced-guided type so a welded contact is detected at the next power-up test rather than at the next accident.
Calibration of the field is the last link in the chain. The divider that scales the column voltage down to the measurement range drifts with temperature and with humidity, and in a laboratory the humidity swings widely between seasons. A divider built from foil resistors in a sealed module, trimmed at the factory and checked against a reference at each annual service, keeps the scale error under one part in ten thousand, which is the level at which the mobility calculation stops being sensitive to calibration.
Records tie the whole system together. A run record that captures the command, the measured voltage, the measured current, the compartment temperature and the lid events allows a failed separation to be reconstructed weeks later. The reconstruction is what turns a complaint into a corrective action, and it is the reason the measurement chain deserves the extra cost of isolation and shielding.
The practical conclusion is that a capillary supply is specified by the gradient it can hold, not by the voltage it can reach. A unit that reaches the number but cannot hold the shape will produce separations that look acceptable on a good day and inexplicable on a bad one, and in regulated laboratories the inexplicable result costs more than the instrument.
The buffer system adds one more variable that the supply cannot ignore. Borate buffers change conductivity with temperature far more than phosphate buffers, and a column that warms by three degrees shifts its resistance by nearly ten percent. A supply with a fixed current limit reacts to that shift by folding back, while a supply that tracks the column temperature and adjusts the limit proportionally keeps the field constant and the run comparable across the working day. The cost of the tracking is a temperature sensor on the cartridge and a small lookup table, which is cheap compared with repeating a batch of samples because the field sagged in the afternoon.
Long runs also expose the interaction between the field and the detector. As the field rises, the electrokinetic injection bias changes, and the amount of sample loaded in the first second depends on the field during that second. A step gradient that settles slowly therefore loads a different amount than one that settles fast, and two instruments with the same nominal settings will disagree until the settling is specified as part of the method. Writing the settling time into the method file is the simplest fix, and it removes a disagreement that otherwise gets blamed on the chemistry.
Field programming for difficult separations usually needs a reversal at the end of the run, a short negative segment that pushes retained species off the column instead of letting them elute in the next injection. The reversal must be fast enough to clear the capillary within the rinse window, and the supply must handle the transition without overshooting the opposite rail. A symmetric push-pull stage with matched slews on both rails makes the transition predictable, and a current limit that is shared between the rails prevents the reversal from drawing more than the column can tolerate.
