... you get impatient and drive HEPI in the same band as driving a suspension transfer function with the same type of excitation. 

[J. Kissel, H. Paris, A. Pele]

The full story:
Hugo was finishing up the extremely long (over-night and into-the-day) measurements of HPI-HAM2, and I'd asked him in what frequency band he was currently running. Since he'd said 10 [mHz]-100 [mHz] when I'd initially asked him (and [I didn't realize / he assumed I knew] the measurement would continue on to higher frequency after this band was done), I launched the 5 hour PR3 M1-to-M1 acceptance transfer functions, to set-and-forget and just be done with it.

Once my measurement was finished, I processed it as normal (see 2013-09-04_1062347744_H1SUSPR3_M1_ALL_TFs.pdf), and found that most degrees of freedom looked spectacular as expected. However, the Vertical and Roll DOFs, only in the 0.3 to 4 [Hz] band, showed a frequency comb of exactly 0.5 [Hz]. What?! To DTT!
(The thought-process narration goes through 2013-09-04-1062347744_H1SUSPR3_M1_0p01to50Hz_DTT.pdf, and the panel ordering is "1" = UL, "2" = LL, "3" = UR, "4" = LR)
After checking to see whether 

- Did awg have a problem from poorly cleared test points? No. (See pg 1, panel 1: Looks like the standard Schroeder-phased transfer function excitation)
- Did my euler basis drive saturate the DAC? No. (See pg 1, panel 2: DAC counts are at most +/-1000 [ct], with T2 and T3 having a DC (pitch) offset in place)
- Did the frequency band or amplitude (driving from 0.3 to 4 [Hz] at 0.01 [Hz] resolution) get screwed up some how? No. (See pg 1, panel 3: an expected 150 [uN/rtHz] from DC up to 4 [Hz] in the Euler Basis, where the OSEM Basis, in DAC counts, the drive is split up evenly between the three OSEMs, and "anti-dewhitened" to compensate for the analog dewhitening.)
- Did the matlab analysis tool some how screw up the data analysis, like with data corruption from frame builder or something? No. (See pg 1, panel 4: I've gathered the transfer function soley from DTT infrastructure, replotted and calibrated the TF. Same exact result in that band. The sky-rocketing magnitude above 4 [Hz] is expected -- it's just noisy incoherent signal.) 
- Did the OSEM sensors saturate, i.e. were we swinging wildy out of the linear range from too large a drive? No. (See pg 2, panel 1: time series of sensor response shows no clipping or anything close to a saturation.)
- Is this craziness in the sensor or the actuator? The sensor (See pg 2, panel 2: ASD of the calibrated OSEM and Euler basis signals)

Well shoot.

Then, I went on to "trust but verify" Hugo's statement that he was only driving up to 100 [mHz]. I find that his excitation has *finished* the 10 to 100 [mHz] band, and the script automatically moved on to working it's way through the 700 [mHz] to 10 [Hz] band during the V and R suspension excitations, and in particular, a 0.5 [Hz] comb between 0.5 and 10 [Hz]. This is shown by the HPI excitation signals in Pg 2, panels 3 and 4. And get this, since suspensions still uses a flat drive instead a comb for their Schroeder phased excitation, the drive from HPI is only coherent at the comb frequencies of HEPI. But it *is* coherent (see pg 3)!

ANYWAYS, of course, had I known the frequencies that Hugo was driving, of course I wouldn't have started the measurement. OH WELL. Lesson learned. And a good commissioning exercise.