Digital Filtering for Smoothing Chromatograms

[M Farooq]

Does anyone know the name of digital filters used in modern HPLCs esp. in Shimadzu and Dionex HPLCs. The reason for asking is that the "time constant" chosen in HPLCs can have different meanings. They serve the same purpose to reduce the noise but there are subtle nuances. So far only, Waters manual clearly defines the name of the digital filter (Hamming filter) in the HPLCs and the "time constant" shown in the software is associated with this filter (as per their manual). What about others?

Thanks.

[Markus Laeubli, Metrohm]

Metrohm's MagIC Net uses Savitzky-Golay filter.

[M Farooq]

Metrohm's MagIC Net uses Savitzy-Golay filter.

Thanks. More specifically, I would be interested in knowing that in your software you must have a "time constant" feature along with the data sampling frequency for the user. Could Metrohm mention what is this "time constant" associated with i.e. it is a hardware RC time constant (like old school time constant knob on 80's instruments) or mathematical time constant of a digital filter?

Regards,

[Markus Laeubli, Metrohm]

The mentioned Savitzky_Golay filter is digital filter as the signal appears digital only. The way digital data points are taken is not disclosed.

[tom jupille]

If it's a Savitsky-Golay filter, then it is digital (essentially a more sophisticated version of the old analog "time constant"). Check Norman Dyson's book "Chromatographic Integration Methods" (pages 150 -155 in the second edition) for a more accurate explanation than I can give you.

By the way, good luck on getting a detailed answer; Dyson has a pertinent comment at the bottom of page 153: ". . . little is published as the details have commercial value.".

Here's the ISBN number for the book:
ISBN-13: 978-0854045105

[M Farooq]

If it's a Savitsky-Golay filter, then it is digital (essentially a more sophisticated version of the old analog "time constant"). Check Norman Dyson's book "Chromatographic Integration Methods" (pages 150 -155 in the second edition) for a more accurate explanation than I can give you.

By the way, good luck on getting a detailed answer; Dyson has a pertinent comment at the bottom of page 153: ". . . little is published as the details have commercial value.".

I will try to get hold of it; Even if we assume we have a Savitsky-Golay filter on a given HPLC instrument. Mathematically, it is easier to understand as to what is being done e.g. polynomial fit for a given number of points. How is the "time constant" associated with a mathematical process such as Savitsky-Golay smoothing?

In old school (when instruments were not black boxes), time constant was a Resistance x Capacitance = Time Constant of the detector circuit. Now, it is apparently associated with "digital" filters which are pure mathematical processes. I am somehow missing the connection the old school RC and the modern instruments "time constant". I would be grateful if someone can clarify this aspect or give a pertinent reference.

[tom jupille]

The essential difference between "hardware" and "software" filtering is that the latter can be reversed. With a hardware RC filter, you have thrown away the information.

My simple-minded interpretation is that the time constant represents an exponentially-weighted moving average of trailing data whereas the S-G is a polynomial-weighted moving average of both trailing and leading data. Note that this is *not* my area of expertise! If you really want to wallow in the maths, get a copy of Dyson's book.