Spectral Peak in a noise spectrum

[badinsults]

I do a lot of work testing electronics that are used to measure and record voltage over a wide range of frequencies. A while back, I was measuring the noise spectrum of the instrument by grounding the inputs. Anyways, at low frequencies (<0.5 Hz), the noise is dominated by pink noise, as might be expected from any electronics. Above that, the noise is white. The way this instrument is set up, there is a digital-to-analogue (DAC) converter that used to manipulate the gain of the input voltage. When the gain is increased, the amplitude of the white noise is decreased, while the pink noise stays at the same magnitude (and as a result, and high gain values, pink noise is dominant at a much higher frequency). This means that the white noise is caused before the amplifier circuitry. However, one odd thing I have noticed in the white noise spectrum is that there is a broad peak (generally 4-8 Hz wide) that exceeds the amplitude of the white noise, and it occurs in the range of 4-20 Hz. The amplitude of this peak decreases when the gain is increased, so therefore it is caused something before the DAC amplifier. However, upon repeated measurements, even with the same sampling rate, recording time, etc, this peak in the white noise spectrum will shift frequencies.

I was wondering if anyone has any idea what could cause this? My first though is that it could be caused by changes in temperature within the instrument, which is affecting the high cut and DC-offset filters, which are located before the DAC.