DDNA

Disk Drive Noise Analysis Measurement Package

Main Features

• Characterize head and media combinations

• Media noise measurements

• Electronics noise measurements

• Separate media noise from electronics noise and determine the dominant noise source

• Channel independent

Trace A is a portion of a captured disk drive sector. Trace C (darker-colored peaks) is a histogram of the averaged Viterbi input samples. There are three distributions due to the peak locations, the zero crossings, and the trough locations. Trace D (lighter-colored peaks) is a histogram of the actual Viterbi input samples, or total noise. Measurements performed on Trace C are Media Signal-to-Noise (msnr), Residual Signal-to-Noise (rsnr), and the ratio of Media Noise compared to Residual Noise. In this particular example, this head and media combination is media noise dominated.

LeCroy's Disk Drive Noise Analysis Measurement Package provides the ability to perform automated noise measurements of magnetically recorded waveforms. The combination of automated noise measurements, long acquisition memory, advanced triggering features, advanced signal processing capabilities, and a large screen for waveform display provides previously unavailable magnetic recording analysis capabilities.

Three new noise specific parameter measurements are provided. Media signal-to-noise (msnr) automatically measures "repetitive" noise that has been recorded on the media. Residual signal-to-noise (rsnr) automatically measures the "non-repetitive" noise, which is often referred to as electronics noise. Media noise to residual noise (m_to_r) automatically measures the ratio of media noise to residual noise. If this ratio is greater than 1.00, then the signal is media noise dominated. If not, then the signal is electronics noise dominated.

Media Noise

Granularity in magnetic media produces zigzag transitions. The exact location of the zigzags changes from write to write, causing media noise (also known as transition noise or zigzag noise). During read, the read head effectively averages across the track the location of all of the zigzags. If the recorded track width is wide, there are many zigzags and very little variability in the averaged transition location. If the track width is narrow, there are fewer zigzags and the variability in the averaged transition location increases.

Head signal amplitude is approximately proportional to track width. Media noise is approximately proportional to the square root of the track width. Therefore, the ratio of the head signal to the media noise (msnr) is proportional to the square root of the track width. As disk drive manufacturers continue to increase disk drive capacity by reducing track width, msnr will get worse. Although advanced head technology (i.e., MR and GMR) can increase the head signal output for a given track width, the effect of media noise on the head signal is also increased so there is no improvement in msnr. As track widths decrease, msnr will continue to increase despite greatly improved head technology.

Media noise can affect a disk drive signal in a number of different ways. It can distort the pulse width and amplitude. It can cause either early or late timing transitions, and partial erasure. In today's disk drive channels, media noise can be a significant source of raw errors.

The trends are clear. Track widths keep getting smaller as disk drive capacity increases. Head technology continues to improve to gain back some of the signal lost due to shrinking track widths.

However, msnr continues to get worse. As a result, the industry is shifting from designs that were electronics noise dominated to designs that are media noise dominated. LeCroy offers new simple tools to help you determine if your system is currently electronics or media noise dominated.

One of the key attributes of media noise is that once it is recorded on the media, it is the same on every read. Thus, media noise is repeating noise. This characteristic of media noise allows a LeCroy Disk Drive Analyzer to make independent measurements of media noise level and the level of electronics noise.

Examples of how media noise can distort the actual transition. Transition A versus Transition B is an example of pulse width and amplitude distortion. Transition C and Transition D are examples of timing distortions. Transition E and Transition F are examples of partial erasure.

This figure represents a single recorded magnetic transition. An ideal transition is a straight line. However, granularity in the media produces zigzag transitions. The zigzag locations change from write to write, producing variability in the read back signal or media noise. The signal distortions are also known as transition noise or zigzag noise.

This figure offers a visual representation of how bit sizes have been decreasing over time. The horizontal length represents the bit spacing, while the vertical height represents the track width. The ratio of the signal to media noise (msnr) is approximately proportional to the square root of the track width. Media noise is becoming a dominant contributor to bit error rate as the track width is reduced to increase the storage capacity of disk drives.

Residual (Electronics) Noise

Residual noise is the random noise present on a disk drive signal from read to read. Thus, any disk drive signal that is captured will contain both media noise and residual noise. LeCroy uses its disk drive triggering capabilities to read the same sector data repeatedly, then samples and aligns the data using its patent pending software capabilities.

Residual signal-to-noise (rsnr) and Media signal-to-noise are related to the noise distributions as follows:

Using this technique, the total noise is quantified and the residual (random) noise is simply averaged and quantified. Removing the residual noise from the total noise yields the resulting non-random media noise.

Since the way sigma_m approaches a final value is known, our calculations estimate the final values of sigma_m and sigma_r. The results of the parameters are therefore always unbiased estimates that stabilize as more sweeps are acquired. This eliminates the need for large n.

By utilizing single frequency data, any hardware channel can be used. As a result, the expected samples will always be peaks, zero crossings, or troughs.

DDNA Signal Requirements

LeCroy's Disk Drive Noise Analysis Package requires SECTOR BASED data. That is, a single frequency preamble, an address mark, and the data. The data needs to be a single frequency pattern.

The ideal "data" pattern is a single 2T pattern or equivalent. By utilizing a single frequency data pattern, any hardware channel can be used -- PR4, EPR4, E2PR4, even modified PRML.

The easiest way to generate this type of signal is with a direct write. That is, disable any scramblers, encoders, and pre-coders. This accomplishes control of exactly what is written on the disk.

Ordering Information