It's a good point to mentioned filtered vibration which I didn't mention at all.
NEMA MG-1 requires us to meet the curve 7-6 for
both filtered (spectral) and unfiltered (overall).
NEMA MG1 – 2003, Section 7.8.1
quote:
Figure 7-6 establishes the limits for bearing housing vibration levels of machines resiliently mounted for both unfiltered and filtered measurements.
For unfiltered vibration the measured velocity level shall not exceed the limit for the appropriate curve on Figure 7-6 corresponding to the rotational frequency.
For filtered vibration the velocity level at each component frequency of the spectrum analysis shall not exceed the value for the appropriate curve in Figure 7-6 at that frequency.
Unfiltered measurements of velocity, displacement, and acceleration may be used in place of a spectrum analysis to determine that the filtered vibration levels over the frequency range do not exceed the limits of the appropriate curve in Figure 7-6. For example, for the top curve in Figure 7-6 the unfiltered velocity should not exceed 0.15 in/s peak (3.8 mm/s), the displacement should not exceed 0.0025 inch (p-p) (63.5 microns), and the acceleration should not exceed 1g (peak).
NOTE—International Standards specify vibration velocity as rms in mm/s. To obtain an approximate metric rms equivalent, multiply the peak vibration in in/s by 18.
Maybe it seems to some people a little redundant to have limits of both filtered and unfiltered. But you can of course come up with situations where we need both to cover different situations. Let's say I have a 2-pole machine, and my unfiltered limit is 0.15 ips resiliently mounted (we'll stick with resiliently mounted for simplicity, although should be rigidly mounted).
A situation that is caught by the unfiltered (overall) but not by the filtered (spectral): Let's say I have a 0.14 ips peak at 60hz and a 014 ips peak at 120 hz. It meets my spectral limits, but the overall would be around 0.22ips and would exceed the overall.
A situation that is caught by the filtered (spectral), but not by a single unfiltered (overall) check: Let's say I have a 0.1 ips peak way up at 1000hz, it meets my velocity overall limit, but doesn't meet the filtered limit at that frequency which is around 0.06.
An alternate approach mentioned above is to check the overall in all three units: mils, velocity and g's and compare to those listed on the curve. That would also catch the last scenario since g's would exceed the limit. This approach would capture any overall or spectral deviation from those limits.
Again as I mentioned there is additional discussion in section 7.8.5 specifically related to 2*LF vibration and the only time it gives any different limit than the previous section is when there is beating between 2*LF and 2*running speed on 2-pole motor. In that case they allow you to compute an RMS-type of average of max and min vibration over a slip cycle and if that is below 80% of the curve (0.15 ips), you're still ok.
It looks like NEMA has the same unit conversion error that EASA did. They give the correct conversion factor of 18 conversion from ips pk/0 to mmps rms, but then they put in parantheses mmps numbers that are 25.4 times as large as the ips number. A little misleading imo.