Measure a Disk-Drive's Read Channel Signals

Analog and digital measurements in the channel reflect a drive's storage capacity and data throughput.

Martin Rowe, Senior Technical Editor -- Test & Measurement World, 8/1/1999

Disk-drive engineers use several types of test equipment to measure the write-channel and read-channel signals. Typically, engineers use oscilloscopes, logic analyzers, vector analyzers, arbitrary waveform generators, bit-error-rate (BER) analyzers, and, of course, computers to measure a read-channel’s performance.

Start with a Computer
Disk-drive engineers start diagnostic testing and failure analysis with a computer. According to Lewis Cronis, design engineer at Quantum (Shrewsbury, MA), disk-drive manufacturers develop their own test software. These proprietary test programs communicate with the disk-drive controller IC through the drive’s I/O port. The drive controller controls the read-channel IC, which processes the read-channel signals and makes the results available to the host computer. The test program stores the test results and plots the data.

The test program typically tests for

• bit-error rate—the average rate at which bit errors occur;
• seek time—the time required for the drive to move the read/write head to a specified location on a platter; these measurements include track-to-track seek time, average seek time, and full-track seek time;
• frequency response—the response of the read channel’s head and preamp; and
• data-transfer rates—the rate at which the drive can transfer data to and from the platter as well as to and from the host computer.

Should a production or prototype disk drive fail or should production drives begin to drift toward the limits of their specs, engineers must look deeper to find the cause of the problem. Here’s where the test equipment comes in. Disk-drive engineers need to know if the failure occurred in the write head, the media (platter), the read head, the preamp, or the read-channel IC. Figure 1 shows the components of a disk drive’s read channel and where engineers connect test equipment.

While the drive’s read-channel IC provides a BER test, the test may not go far enough to help locate the problem. Analyzing the bit errors, however, can help. Tom Waschura, principal engineer at SyntheSys Research (Menlo Park, CA) says bit-error analysis involves calculating the probability that an errored bit will occur. To perform the analysis, engineers use BER analyzers to compare data read from a drive to data sent to the drive.

Find Bit Errors
The intervals at which errors occur can reveal the source of the problem. A histogram of error occurrences (Fig. 2) can reveal bit errors at regular intervals. To produce the histogram, engineers program a computer to read bits from the drive under test at a constant rate. When errored bits occur at regular intervals in time, then the error source could be oscillations such as those from a switching power supply. The servo controller that moves the head actuator across the platters of a drive also contains a clock, which can cause bit errors at regular intervals. Waschura says savvy engineers look for frequencies that correspond to regularly occurring errors.

If the errors appear random, then the problem could be caused by noise anywhere in the read channel or the write channel. After all, the signals written to and read by the heads are low-level analog voltage and currents, although they represent digital bits. When engineers can’t locate the error source through digital measurements, they must resort to analog measurements. Engineers most often turn to oscilloscopes to view the read signals and the write signals.

In the write channel, an inductive head magnetizes the platter’s tracks to represent a series of data bits. Engineers must measure the parametrics of the current in the inductive head as it writes data onto the platters. Quantum’s Cronis uses a 1-GHz current probe and a 1-Gsample/s DSO to measure the risetime, symmetry, overshoot, and peak-to-peak values of the current through the write head. He measures the time between 10% and 90% points of amplitude on the current’s rising edge.

If the write-current signal appears correct, engineers turn their attention to the media and the read channel. First, they connect a scope, through a differential probe, to the preamp output in the read-channel. The probe should have a bandwidth of at least 1 GHz and the scope’s bandwidth should be at least 500 MHz for data rates of 300 Mbps.

Figure 3 shows a typical signal measured at the output of a disk drive’s preamp, and the figure indicates some of the important measurement parameters. These measurements help engineers verify that the read head and preamp function properly. PW50+ and PW50– refer to the width of the pulse at the points where the pulses amplitude is 50% of its peak value and trough value, respectively. Because the magnetic-flux transitions on the platter appear as pulses, narrower is better. If the pulses are too wide, then adjacent bits on a track may interfere with each other. What constitutes a pulse that’s too wide depends on the drive’s design and the manufacturer’s specs.

Pulse Those Bits
Pulse amplitude is another important measurement. Measurements TAA+ and TAA– refer to the track-average amplitude—the average amplitude of the pulses in a given track on the platter. Here again, the optimal value of this measurement depends on the density of the drive’s platter and the sensitivity of the read-channel. If the TAA is too high, the magnetic flux from the platter will cause the magnetoresistive (MR) head to saturate, which distorts the waveform that the head sends to the preamp. If the TAA is too low, the channel’s SNR will be too low, causing random bit errors.

Baseline measurements, says Chuck Nielsen, chief technologist at Fujitsu (San Jose, CA), refer to the height of the “knees” that occur on the pulses near the baseline, the average between the peaks of the pulses. If the height of the knees (a pulse’s local baseline) is too high, it causes the read-channel IC to see a pulse whose width is too wide. When that happens, the read-channel IC misinterprets the pulse.

The read-channel IC needs a clock to ensure it reads bits at the proper times. Through decoding of the data, the read-channel IC extracts a clock signal. A phase-locked-loop (PLL) synchronizes the clock signal extracted from the data to the constant spin rate of the platter.

Because of the synchronous nature of the data, jitter in the clock pulses can cause read errors when a pulse occurs early or late relative to a clock pulse. Engineers often use oscilloscopes to measure pulse jitter. Figure 4 shows a histogram of the jitter in a transition between two pulses. The 3.14-ns standard deviation shown in the figure doesn’t necessarily indicate an acceptable or unacceptable measurement; the drive’s design dictates what is acceptable and unacceptable.

Bit errors can occur not only because of electrical issues but also because of contamination on the platter. If bit errors occur in physical proximity to each other on the platter, then the platter is most likely damaged.

One form of contamination is a particle on the platter. If the read head hits a particle, friction rapidly heats the head and changes its resistance, producing a voltage spike in the output waveform. To find this type of problem—called thermal asperity—engineers perform an asperity test with an oscilloscope. Because these spikes have larger amplitudes than proper signals, engineers can find them by setting a scope trigger level that’s higher than the amplitude of good signals.

So far, the measurements I’ve discussed are all accessible at the read-channel’s preamp output. But several measurements are available as test points on the read-channel IC. These points include servo signals and the read-channel data signals as they pass through the function blocks in the read-channel IC.

Once such test signal is a disk drive’s position error signal (PES), an analog voltage that represents the position of the actuator holding the heads relative to each track on the platter. Tracks are organized as concentric circles around the platter’s center. Disk drives are also divided into sectors that cut across tracks and radiate from the platter’s center. A servo controller controls the actuator’s position as it moves across the platter.

Platter manufacturers place timing and positioning signals on each track at the beginning of each sector. The positioning signals—called “bursts”—actually sit just inside and just outside of each track. A burst signal consisting of 10 pulses appears between the cursors in Figure 5. The first five pulses come from a burst just outside a track (burst 1), while the second five pulses come from a burst just inside the track (burst 2). If the head rides equidistant to the burst signals—the head is in the center of the track—then the mean amplitudes of each set of five pulses will be equal and the PES will be 0. If the mean of the peaks in burst 1 is greater than that in burst 2, then the head is running off the track’s center.

Measurements and Aircraft
Off-track capability (OTC) measurements also relate to the center of a track. Engineers such as Randy Rannow, senior development engineer at Seagate (Longmont, CO), use OTC measurements to assess the error rate when the head position deviates from the center of the track. Engineers plot OTC on what’s called a “747 curve” (Fig. 6) by measuring BER relative to a head’s position on a track. (The curve looks like the profile of a 747 aircraft.) A 747 curve shows OTC on a track of interest as the drive writes to adjacent tracks at incremental track spacings, or track pitch. The graph in Figure 6 plots the distance that a head can move from a track’s center (the vertical axis) as a function of track pitch while maintaining an acceptable BER.

To generate a 747 curve, engineers fill the areas alongside the track of interest with noise, or old information (red area). Next, they write signals on the track of interest (green area) over the old information. Each written track has a magnetic width and erase bands (yellow areas) on each side. Then, adjacent tracks are written at a large track pitch (blue areas), and engineers measure OTC by moving the read element across the track of interest and measuring BER at the track pitch at point 1 in the graph. Adjacent tracks are written again at a smaller track pitch and OTC is measured again at point 2. Engineers repeat this process as they reduce the distance between adjacent tracks.

As track pitch decreases, the erase bands in the track of interest and those belonging to the adjacent tracks begin to interact, working together to erase the old information. The read-head “sees” the noise up to point 3 on the graph. The track signals work together, effectively decreasing noise and improving error rate and therefore OTC, which accounts for the hump in the curve at point 4. Here, the peak of the hump is where the two erase bands meet before they overlap. The old information has been completely overwritten. At this point, the head can deviate to its maximum distance (8 min.) from the center of the track while still maintaining an acceptable BER.

As track pitch continues to decrease, the OTC distance will return to the level where the old information was dominating the read signal (point 5). Point 5 is known as the average track spacing (ATS) point. Here, track spacing is equal to the width of the write head plus 1 erase band. At point 6, OTC all but vanishes because adjacent tracks are written too close to each other, or they encroach upon the track of interest, which increases BER.

So far, I’ve discussed measurements that engineers perform with scopes and BER testers, but those instruments aren’t enough to fully measure a drive’s performance. Besides using a scope to look at the read signals from the preamp, engineers also use a spectrum analyzer. The spectrum analyzer lets these engineers look at the head and preamp frequency response and noise. By lifting the read head away from the disk, engineers can measure the noise generated by the head and the preamp.¹  

With hundreds of megabits per second passing through a read channel, the wires between the preamp and the channel IC take on the characteristics of transmission lines. The connections between a read-channel’s components can cause impedance mismatches. To measure those impedances, disk-drive engineers use a network analyzer.

Once a head signal enters the read-channel IC, it goes through several processing steps before the read-channel IC outputs bytes of data to the disk-drive controller. The first stage in the IC is typically a low-pass filter that shapes the pulse before a 6-bit ADC digitizes the signal. Not all read-channel ICs use an ADC. For those that do, the IC provides access to the ADC’s outputs where engineers can attach a logic analyzer. With the logic analyzer, engineers can see a digital representation of the analog signals read from the platter. They can process the digital information to see how accurately the read-channel IC digitizes the preamp signal.

To test the preamp and read-channel IC’s ADC, engineers use an arbitrary waveform generator (AWG) to simulate the head signals and view the ADC’s output with a logic analyzer. To simulate the head signal, engineers connect the AWG’s output to either a current probe or to a voltage transformer. By placing a wire through the current probe and driving it with the AWG (shown in Figure 1), engineers create a drive current that follows the AWG’s voltage output. With a voltage transformer, the AWG drives the preamp with voltage instead of current. Using the logic analyzer, engineers verify that the ADC properly detects the peaks in the signal.

The logic analyzer can also connect to the 8-bit digital output of the read-channel IC, which connects to the disk drive’s controller IC. With the AWG, engineers can alter the characteristics—timing, amplitude, width, shape, and position—of the simulated head pulses until the read-channel IC produces bit errors. That lets disk-drive designers test the limits of the drive’s ability to produce error-free reads. T&MW

FOOTNOTE
1. Ashar, Kanu G., Magnetic Disk Drive Technology: Heads, Media, Channel, Interfaces, and Integration, IEEE Press, Piscataway, NJ, ISBN 0-7803-1083-7, 1997, p. 254.

FOR FURTHER READING
Application Note: Measurement Solutions for Disk Drive Design, Tektronix, Beaverton, OR, document number 49W-1276-0, 1998.

“Making PRML Measurements with Digital Oscilloscopes,” Data Storage, June 1995. The article is available at www.lecroy.com/applications/PRMLMeas/default.asp