Home  Patents  Binary Sampler  Articles  Contact 

---

PST Patents & Innovations

---

Description	Patent Number	Benefit
Binary Sampler	In process	New sampling method offers dramatic improvements in speed and resolution.
Data Recovery Phase-Locked Loop	3,810,234	First LC oscillator pll for hard disk data recovery. Solved internal crosstalk problem at high data rates. Improved product yields and reduced manufacturing costs.
Phase Margin Analysis	--	This was a major breakthrough in hard disk manufacturing and R&D. As an illustration, it improved manufacturing yields and reduced test time from 48 hours to five minutes, and probably saved the industry hundreds of million if not billions of dollars. Everyone benefited from this innovation. Invented in 1970, it helped start the amazing productivity improvements in magnetic recording since the 1980's, and it became an industry standard test method from 1980 to around 1994. IBM used it in at least two patents: 4789598, Dec 6, 1988 and 5227212, Jul 13, 1993.
Rapid Start VCO	4,533,881	Provides instant single-cycle pll lock capability.
Constant Duty Cycle Peak Detector	4,603,299	The unique noise rejection property provides exceptional accuracy and increased dynamic range at RF frequencies.
Disk Defect Detection	4,929,894	This is the only method available to locate hard disk micro-defects. It solved major yield problems and doubled the test throughput.
Improved Disk Test	5,050,169	The key discovery is thermal drift is low enough to test a written track without a servo. The parallel approach increases test throughput
Video Pointer	3,674,930	First handheld cursor for Closed-Circuit Television. Multiplexed analog bidirectional signals on single coax.

Data Recovery Phase-Locked Loop

Early hard disk read channels used astable multivibrators as vco's in the phase-locked loop. This type of oscillator is sensitive to switching noise and crosstalk on the pcb. The resulting timing jitter reduced the timing margins and increased the disk error rate. This placed the burden on other components such as the read/write head and channel electronics, which resulted in poor yields and increased manufacturing costs.

LC oscillators are much less sensitive to transient noise due to the Q of the oscillator tank. However, this also makes it difficult to synchronize the oscillator phase to the incoming data and permit data recovery. A novel method was used to rapidly synchronize the phase of the oscillator tank to the incoming data.

This patent describes the first LC oscillator pll for data recovery in hard disk drives, and includes the original solution for the deadband problem in the phase detector. This has now become the standard method used in phase-locked loops. The new design approach greatly reduced the static and dynamic timing errors, which improved the channel error rate, increased the product yield, reduced the troubleshooting and repair time in manufacturing, and reduced the cost of heads and other components needed to meet the product specification.

Read the complete patent here (pdf file): United States Patent 3,810,234: Data Recovery Circuit

Fig. 1. Bit Error Rate Curves

Phase Margin Analysis

In the early days, production testing of hard disk drives was very slow and uncertain. The drive had to meet stringent error rate specifications, which meant writing different data patterns on each head and cylinder and reading the data to check for errors. This test took several days, and any data errors gave little indication why they occurred.

For example, intersymbol interference caused some pattern sequences to be more error-prone than others. Random errors could be caused by Gaussian noise in the channel or stray noise transients from the surrounding equipment. Disk defects could be on the ragged edge of failure, and be error-free on one pass or fail on the next. Troubleshooting these problems was largely guesswork and luck, and production lines were frequently shut down due to failure to meet the error rate specifications.

A solution was found by a simple modification to the Data Recovery patent (3,810,234). This allowed measuring the actual bit error rate curve of the channel. Typical results are shown in Fig. 1, where the red curve indicates a channel that meets specification, and the black curve fails.

Simple curve-fitting routines extracted the intersymbol interference and Gaussian noise components, and explained why a channel failed. For example, the signal to noise ratios in curves "A" and B" are comparable since the slope of the curves are similar. However, channel "B" has more intersymbol interference as shown by the slight knee in the curve. This places the curve closer to the edge of the data window and causes errors due to Gaussian noise.

The diagnostic information provided by these curves was invaluable. It made it possible to distinguish between multiple sources of problems, and solve them one at a time.

The wealth of diagnostic information contained in these curves changed the disk industry. It dramatically reduced development and production costs, and was a key contribution to the swift rise in hard disk performance. For example, the test time for one model disk drive was reduced from 48 hours to 5 minutes, and the improvement in cost per megabyte for hard disk drives outpaced every other competing technology, such as magnetic bubbles.

The technique came to be known as "Marginalized VFO", or "Phase Margin Analysis", and it became a worldwide standard for R&D and production test from 1980 until about 1994, when PRML encoding became common.

The theory is described in "Effect of Bitshift Distribution on Error Rate in Magnetic Recording", Eric R. Katz and Thomas G. Campbell, IEEE Transaction on Magnetics, Vol. MAG-15, No. 3, May 1979, pp 1050-1053.

IBM used it in at least two patents: 4789598, Dec 6, 1988 and 5227212, Jul 13, 1993.

You can read about the invention that changed the entire industry here: Effect of Bitshift Distribution on Error Rate in Magnetic Recording (2.9MB pdf)

Rapid Start VCO

In Phase-Locked Loops for data recovery, it is desirable to synchronize the pll by starting the Voltage-Controlled Oscillator (VCO) in phase with the incoming data. This minimizes the time needed for the loop to settle before data recovery can begin. As data rates increase, it becomes harder and harder to do so without disrupting the performance of the VCO.

This design is suitable for narrow-band applications where good noise performance is required. It is capable of operation into the low VHF region.

Read the complete patent here (pdf file): United States Patent 4,533,881: Rapid starting variable frequency oscillator with minimum startup perturbations.

Constant Duty Cycle Peak Detector

The readback signal in a hard disk drive is one of the most difficult to analyze. It contains various forms of amplitude noise, dropouts, and variations in amplitude as the head moves across the tracks.

Obtaining an accurate measure of the peak amplitude has been extremely difficult. Standard methods of peak detection cannot cope with the noise and wide variations in amplitude, frequency, and waveforms that are present in the read signal.

The new method represents a breakthrough in performance and accuracy, and provides reliable measurements that were impossible with previous methods.

Due to the excellent noise-rejection properties, this invention can be considered a precursor to the Binary Sampler

You can read the complete patent here (pdf file): United States Patent 4,603,299: Constant duty cycle peak detector

A LTspice analysis showing 0.032% Total Harmonic Distortion (THD) at 95% modulation is available here

Disk Defect Detection

Disk defects come in all sizes, but the hardest ones to find are tiny micro-defects caused by microscopic scratches on the surface of the media. These produce defects that are smaller than a bit cell, and they distort the readback signal in different ways depending on the timing relationship between the defect and the bit cell. An example is shown below:

The standard methods of detecting disk defects used in production could find this defect, providing it always kept this timing relationship in the bit cell. Unfortunately, small timing changes will cause the defect to move with respect to the bit cell, which can increase the amplitude so the defect looks like the surrounding signals. When this happens, the standard methods become very unreliable, and these defects can pass unnoticed.

The invention described below solved this problem by comparing a signal to its surrounding neighbors. Any distortion caused by a defect makes one cycle a slightly different shape than the others, thus locating the defect.

An additional benefit of this invention is it allowed two separate tests to be combined into a single test, which doubled the throughput of the production line.

Read the complete patent here (pdf file): United States Patent 4,929,894: Method and apparatus for increasing throughput on disk drive testing

Improved Disk Test

A typical hard disk uses several disks with one Read/Write head per surface. At one time, production test methods were obliged to write on a single head at a time because of safety features designed into the Read/Write controller. But if a method could be found to bypass the safety feature, all the heads could be written simultaneously. This saves disk revolutions and accelerate the test process.

The key is to recognize that if the tests are performed rapidly enough, the heads will not drift off-track between the time they are written and the time they are tested. This is important since there is no servo track at this stage in the manufacturing process to keep them on track.

The following table shows the improvement for a disk drive using 6 heads:

This illustrates a savings of n - 1, or 5 disk revolutions for each test, which almost doubles the throughput.

Patent 5,050,169 describes several methods to write on all heads simultaneously. The benefit to the user is increased productivity and reduced cost.

Read the complete patent here (pdf file): United States Patent 5,050,169: Method and apparatus for testing magnetic disks

Video Pointer

This was the first portable, battery-powered video cursor for closed-circuit television displays. Instead of using separate cables for Video In and Video Out, both signals were multiplexed on a single coax. A rapid-start variable frequency oscillator adjusted the angle of the line to facilitate pointing to the desired object on the display.

Read the complete patent here (pdf file): United States Patent 3,674,930: Video Pointer