157:
93:
289:(e.g. PR4 forms a ternary eye). In general, however, the target can just as readily have non-integer values. The classical approach to maximum-likelihood detection on a channel with intersymbol interference (ISI) is to equalize to a minimum-phase, whitened, matched-filter target. The complexity of the subsequent Viterbi detector increases exponentially with the target length - the number of states doubling for each 1-sample increase in target length.
207:. The PRML channel outperformed a competing implementation based on "Null-Zone Detection". A prototype PRML channel was implemented earlier at 20 Mbit/s on a prototype 8-inch HDD, but Ampex exited the HDD business in 1985. These implementations and their mode of operation are best described in a paper by Wood and Petersen. Petersen was granted a patent on the PRML channel but it was never leveraged by Ampex.
307:
Viterbi detector was modified such that it recognized the expected signal-level and expected noise variance associated with each bit-pattern. As a final step, the detectors were modified to include a 'noise predictor filter' thus allowing each pattern to have a different noise-spectrum. Such detectors are referred to as
Pattern-Dependent Noise-Prediction (PDNP) detectors or
297:
Given the rapid increase in complexity with longer targets, a post-processor architecture was proposed, firstly for EPRML. With this approach a relatively simple detector (e.g. PRML) is followed by a post-processor which examines the residual waveform error and looks for the occurrence of likely bit
306:
As data detectors became more sophisticated, it was found important to deal with any residual signal nonlinearities as well as pattern-dependent noise (noise tends to be largest when there is a magnetic transition between bits) including changes in noise-spectrum with data-pattern. To this end, the
172:
focused on a high level of integration and low power consumption for a mass-market HDD. In both cases, the initial equalization to PR4 response was done with analog circuitry but the
Viterbi algorithm was performed with digital logic. In the tape application, PRML superseded 'flat equalization'. In
103:
was first proposed by Adam Lender in 1963. The method was generalized by
Kretzmer in 1966. Kretzmer also classified the several different possible responses, for example, PR1 is duobinary and PR4 is the response used in the classical PRML. In 1970, Kobayashi and Tang recognized the value of PR4 for
139:
had recognized that the
Viterbi algorithm could be applied to analog channels with inter-symbol interference and particularly to the use of PR4 in the context of Magnetic Recording (later called PRML). (The wide range of applications of the Viterbi algorithm is well described in a review paper by
284:
PR4 is characterized by an equalization target (+1, 0, -1) in bit-response sample values or (1-D)(1+D) in polynomial notation (here, D is the delay operator referring to a one sample delay). The target (+1, +1, -1, -1) or (1-D)(1+D)^2 is called
Extended PRML (or EPRML). The entire family,
331:
uses a 2-input, 1-output equalizer.) The detector uses the PDNP/NPML approach but the hard-decision
Viterbi algorithm is replaced with a detector providing soft-outputs (additional information about the reliability of each bit). Such detectors using a soft Viterbi algorithm or
285:(1-D)(1+D)^n, was investigated by Thapar and Patel. The targets with larger n value tend to be more suited to channels with poor high-frequency response. This series of targets all have integer sample values and form an open
340:
used in modern HDDs. A single integrated circuit contains the entire read and write channels (including the iterative decoder) as well as all the disk control and interface functions. There are currently two suppliers:
66:
Ampex introduced PRML in a tape drive in 1984. IBM introduced PRML in a disk drive in 1990 and also coined the acronym PRML. Many advances have taken place since the initial introduction. Recent
70:
operate at much higher data-rates, are fully adaptive, and, in particular, include the ability to handle nonlinear signal distortion and non-stationary, colored, data-dependent noise (
323:, correction for the nonlinear read-element response, and a low-pass filter with control over the high-frequency boost or cut. Equalization is done after the ADC with a digital
319:
Although the PRML acronym is still occasionally used, advanced detectors are more complex than PRML and operate at higher data rates. The analog front-end typically includes
267:
for PRML NRZ recording. 'Precomp.' largely cancels the effect of NLTS. Precompensation is viewed as a necessity for a PRML system and is important enough to appear in the
234:. A parallel R&D effort at IBM San Jose did not lead directly to a product. A competing technology at the time was 17ML an example of Finite-Depth Tree-Search (FDTS).
193:. DCRS was a cassette-based, digital, instrumentation recorder capable of extended play times at very high data-rate. It became Ampex' most successful digital product.
260:
recording at high density and/or high data-rate was recognized in 1979. The magnitude and sources of NLTS can be identified using the 'extracted dipulse' technique.
196:
The heads and the read/write channel ran at the (then) remarkably high data-rate of 117 Mbits/s. The PRML electronics were implemented with four 4-bit,
237:
The IBM 0681 read/write channel ran at a data-rate of 24 Mbits/s but was more highly integrated with the entire channel contained in a single 68-pin
900:
80:
refers to the fact that part of the response to an individual bit may occur at one sample instant while other parts fall in other sample instants.
51:
than earlier simpler schemes such as peak-detection. These advances are important because most of the digital data in the world is stored using
164:
The first two implementations were in Tape (Ampex - 1984) and then in hard disk drives (IBM - 1990). Both are significant milestones with the
308:
71:
185:
The first implementation of PRML was shipped in 1984 in the Ampex
Digital Cassette Recording System (DCRS). The chief engineer on DCRS was
216:
737:
144:.) A simplified algorithm, based upon a difference metric, was used in the early implementations. This is due to Ferguson at
513:
328:
298:
pattern errors. This approach was found to be valuable when it was extended to systems employing a simple parity check
958:
219:
It was full-height 5¼-inch form-factor with up to 12 of 130 mm disks and had a maximum capacity of 857 MB.
244:
operating off a 5 volt supply. As well as the fixed analog equalizer, the channel boasted a simple adaptive digital
48:
84:
refers to the detector finding the bit-pattern most likely to have been responsible for the read-back waveform.
907:
337:
32:
238:
200:
777:
Maximum
Likelihood Sequence Estimation of Digital Sequences in the Presence of Intersymbol Interference
186:
516:", J. IERE, Vol., 55, No. 6, pp. 229-236, June 1985. (invited) (Charles Babbage Award for Best Paper)
888:
Adaptive Noise-Predictive
Maximum-Likelihood (NPML) Data Detection for Magnetic Tape Storage Systems
461:
346:
156:
36:
526:
500:
320:
264:
248:
after the A/D to compensate for changes in radius and/or changes in the magnetic components.
20:
399:", Trans. AIEE, Part I: Communication and Electronics, Vol. 82, No. 2, pp. 214-218, May 1963
8:
190:
168:
implementation focused on very high data-rate for a digital instrumentation recorder and
125:
764:
A Class of
Partial Response Systems for Increasing Storage Density in Magnetic Recording
396:
861:
804:"A new target response with parity coding for high density magnetic recording channels"
358:
263:
Ampex was the first to recognize the impact of NLTS on PR4. and was first to implement
257:
241:
174:
112:
105:
92:
67:
847:
633:
436:
Error bounds for convolutional codes and an asymptotically optimum decoding algorithm
363:
132:
117:
741:
815:
621:
Error Rate Performance of Experimental Gigabit per Square Inch Recording Components
311:(NPML). Such techniques have been more recently applied to digital tape recorders.
52:
934:
646:
581:
56:
40:
711:
487:
887:
776:
763:
724:
698:
685:
672:
568:
555:
539:
474:
448:
435:
422:
409:
333:
121:
939:
874:
833:
803:
789:
620:
607:
594:
569:
Viterbi Detection of Class IV Partial Response on a Magnetic Recording Channel
383:
952:
659:
223:
714:", IEEE Trans. Magn., Vol. MAG-31, No. 2, pp. 1071-1076, Mar. 1995 (invited)
610:", IEEE Journal on Selected Areas in Comms, vol.10, No.1, pp.38-56, Jan 1992
556:
An Experimental Eight-inch Disc Drive with One-hundred Megabytes Per Surface
423:
Application of Partial-response Channel Coding to Magnetic Recording Systems
558:", IEEE Trans. Mag., vol. MAG-20, No. 5, pp. 698-702, Sept. 1984. (invited)
686:
The Helical-Scan Magnetic Tape Recorder as a Digital Communication Channel
571:", IEEE Trans. Comm., Vol., COM-34, No. 5, pp. 454-461, May 1986 (invited)
96:
Continuous-time Partial-Response (class 4) and corresponding 'eye pattern'
286:
231:
204:
141:
699:
Identification of Nonlinear Write Effects Using Pseudo-Random Sequences
540:
Error Control in Duobinary Data Systems by Means of Null Zone Detection
324:
227:
60:
44:
819:
301:
145:
944:
Characterization of Magnetic Recording Systems: A Practical Approach
864:" IEEE J. Sel. Areas Commun., vol. 19, no. 4, pp. 730–743, Apr. 2001
47:. PRML was introduced to recover data more reliably or at a greater
727:", IEEE Trans. Magn., Vol. MAG-22, No. 5, pp. 1203-1205, Sept. 1986
701:", IEEE Trans. Magn., Vol. MAG-23, no. 5, pp. 2377-2379, Sept. 1987
662:", IEEE Trans. Magn., Vol. MAG-25, No. 5, pp. 4075-4077, Sept. 1989
634:
Performance Data for a Six-Sample Look-Ahead 17ML Detection Channel
438:", IEEE Trans. Info. Theory, Vol. 13, No. 2, pp. 260-269, Apr. 1967
342:
792:", IEEE Trans. Magn., Vol. MAG-29, No. 6, pp. 4018-4020, Nov. 1993
451:", IEEE Trans. Inform. Theory, vol. IT-17, PP. 586-594, Sept. 1971
712:
Characterization of the Read/Write Process for Magnetic Recording
660:
New Detector for 1,k Codes Equalized to Class II Partial Response
582:
Digital maximum likelihood detector for class IV partial response
197:
725:
The Effects of Nonlinear Distortion on Class IV Partial Response
256:
The presence of nonlinear transition-shift (NLTS) distortion on
779:", IEEE Trans. Info. Theory, vol. IT-18, pp. 363-378, May 1972.
688:", IEEE Trans. Mag. vol. MAG-15, no. 2, pp. 935-943, March 1979
271:
HDD setup although it is now handled automatically by the HDD.
877:". IEEE Trans. Magn. Vol. 32, No. 5, pp. 3968–3970, Sept. 1996
636:", IEEE Trans. Magn., Vol. 29, No. 6, pp. 4012-4014, Dec. 1993
501:
Overview of the prototype of the first commercial PRML channel
189:. The machine evolved from a 6-head, transverse-scan, digital
766:", IEEE Trans. Magn., vol. 23, No. 5, pp.3666-3668 Sept. 1987
673:
Improvement of recording density by means of cosine equalizer
215:
In 1990, IBM shipped the first PRML channel in an HDD in the
165:
890:", IBM J. Res. Dev. Vol. 54, No. 2, pp. 7.1-7.10, March 2010
875:
Improving Performance of PRML/EPRML through Noise Prediction
862:
Pattern-dependent noise prediction in signal dependent noise
836:", IEEE Trans. Magn. Vol. 37, No. 2, pp. 714–720, March 2001
675:", IEEE Trans. Magn., Vol. 12, No. 6, pp. 746-748, Nov. 1976
597:", IEEE Trans. Magn., Vol. 27, No. 6, pp. 4538-43, Nov. 1991
542:", IEEE Trans. Comm., Vil 16, No. 6, pp. 825-830, Dec., 1968
848:
Coding and Signal Processing for Magnetic Recording Systems
593:
J. Coker, R. Galbraith, G. Kerwin, J. Rae, P. Ziperovich, "
425:", IBM J. Res. Dev., Vol, 14, No. 4, pp. 368-375, July 1970
410:
Generalization of a Technique for Binary Data Communication
268:
738:"Kursk: BIOS Settings - Standard CMOS Setup, Feb 12, 2000"
623:", IEEE Trans. Magn., Vol. 26, No. 5, pp. 2298-2302, 1990
412:", IEEE Trans. Comm., Vol. 14, No. 1, pp. 67-68 Feb. 1966
169:
136:
845:
M. Despotovic, V. Senk, "Data Detection", Chapter 32 in
834:
NPML Detection Combined with Parity-Based Postprocessing
449:
Correlative level coding and maximum-likelihood decoding
397:
The duobinary technique for high-speed data transmission
832:
R. Cideciyan, J. Coker; E. Eleftheriou; R. Galbraith, "
490:", THIC meeting, Ellicott City, MD, 16 Oct., 1996 (PDF)
386:", IEEE Spectrum, Vol. 33, No. 11, pp. 70-76, Nov. 1996
35:
from the weak analog read-back signal picked up by the
16:
Method for interpreting data in digital storage systems
477:” Bell Syst. Tech. J., vol. 51, pp. 493-505, Feb. 1972
475:
Optimal reception for binary partial response channels
606:
R.Cidecyan, F.Dolvio, R. Hermann, W.Hirt, W. Schott "
514:
High Data Rate Magnetic Recording in a Single Channel
503:", Computer History Museum, #102788145, Mar. 26, 2009
464:”, Proc. IEEE, Vol. 61, No. 3, pp. 268-278, Mar. 1973
512:
C. Coleman, D. Lindholm, D. Petersen, and R. Wood, "
554:
R. Wood, S. Ahlgrim, K. Hallamasek, R. Stenerson, "
302:
PRML with nonlinearities and signal-dependent noise
222:The PRML channel for the IBM 0681 was developed in
950:
647:Apparatus and method for fixed delay tree search
697:D. Palmer, P. Ziperovich, R. Wood, T. Howell, "
488:Ampex Digital Cassette Recording System (DCRS)
384:PRML detection boosts hard-disk drive capacity
292:
851:edited by B. Vasic, E. Kurtas, CRC Press 2004
382:G. Fisher, W. Abbott, J. Sonntag, R. Nesin, "
309:noise-predictive maximum-likelihood detectors
151:
901:"Marvell 88i9422 Soleil SATA HDD Controller"
608:A PRML System for Digital Magnetic Recording
595:Implementation of PRML in a rigid disk drive
160:Early PRML chronology (created around 1994)
336:are essential in iteratively decoding the
87:
710:D. Palmer, J. Hong, D. Stanek, R. Wood, "
886:E. Eleftheriou, S. Ölçer, R. Hutchins, "
584:", US Patent 4504872, filed Feb. 8, 1983
550:
548:
251:
226:lab. in Minnesota with support from the
155:
91:
790:Turbo-PRML, A Compromise EPRML Detector
274:
951:
801:
525:Computer History Museum, #102741157, "
940:Online Chapter "Introduction to PRML"
545:
314:
173:the HDD application, PRML superseded
671:T. Kameyama, S. Takanami, R. Arai, "
279:
210:
25:partial-response maximum-likelihood
13:
928:
14:
970:
180:
31:) is a method for recovering the
906:. September 2015. Archived from
893:
880:
867:
854:
839:
826:
795:
782:
769:
756:
730:
717:
704:
691:
678:
665:
652:
639:
626:
613:
600:
587:
574:
561:
532:
519:
506:
493:
808:IEEE Transactions on Magnetics
480:
467:
454:
441:
428:
415:
402:
389:
376:
1:
942:, from Alex Taratorin's book
369:
338:low-density parity-check code
177:codes with 'peak detection'.
116:decoding using the eponymous
527:Ampex PRML Prototype Circuit
201:analog-to-digital converters
7:
421:H. Kobayashi and D. Tang, "
352:
293:Post-processor architecture
10:
975:
873:E. Eleftheriou, W. Hirt, "
567:R. Wood and D. Petersen, "
152:Implementation in products
499:R. Wood, K. Hallamasek, "
959:Computer storage devices
802:Conway, T. (July 1998).
684:R. Wood, R. Donaldson, "
649:", filed Oct. 30th, 1989
120:was proposed in 1967 by
124:as a means of decoding
88:Theoretical development
161:
97:
645:R. Carley, J. Moon, "
462:The Viterbi Algorithm
265:Write precompensation
252:Write precompensation
159:
95:
21:computer data storage
762:H.Thapar, A.Patel, "
723:P. Newby, R. Wood, "
275:Further developments
860:J. Moon, J. Park, "
619:T. Howell, et al. "
191:video tape recorder
126:convolutional codes
68:read/write channels
935:The PC Guide: PRML
744:on October 4, 2018
359:Maximum likelihood
315:Modern electronics
242:integrated circuit
162:
113:Maximum-likelihood
106:magnetic recording
98:
82:Maximum-likelihood
820:10.1109/20.703887
364:Viterbi algorithm
230:Research lab. in
133:Hisashi Kobayashi
118:Viterbi algorithm
966:
922:
921:
919:
918:
912:
905:
897:
891:
884:
878:
871:
865:
858:
852:
843:
837:
830:
824:
823:
814:(4): 2382–2386.
799:
793:
786:
780:
773:
767:
760:
754:
753:
751:
749:
740:. Archived from
734:
728:
721:
715:
708:
702:
695:
689:
682:
676:
669:
663:
656:
650:
643:
637:
630:
624:
617:
611:
604:
598:
591:
585:
578:
572:
565:
559:
552:
543:
536:
530:
523:
517:
510:
504:
497:
491:
484:
478:
471:
465:
458:
452:
445:
439:
432:
426:
419:
413:
406:
400:
393:
387:
380:
280:Generalized PRML
246:cosine equalizer
211:Hard disk drives
101:Partial-response
78:Partial response
53:magnetic storage
974:
973:
969:
968:
967:
965:
964:
963:
949:
948:
931:
929:Further reading
926:
925:
916:
914:
910:
903:
899:
898:
894:
885:
881:
872:
868:
859:
855:
844:
840:
831:
827:
800:
796:
787:
783:
774:
770:
761:
757:
747:
745:
736:
735:
731:
722:
718:
709:
705:
696:
692:
683:
679:
670:
666:
657:
653:
644:
640:
631:
627:
618:
614:
605:
601:
592:
588:
579:
575:
566:
562:
553:
546:
537:
533:
524:
520:
511:
507:
498:
494:
485:
481:
472:
468:
459:
455:
447:H. Kobayashi, "
446:
442:
433:
429:
420:
416:
407:
403:
394:
390:
381:
377:
372:
355:
317:
304:
295:
282:
277:
254:
213:
187:Charles Coleman
183:
154:
90:
17:
12:
11:
5:
972:
962:
961:
947:
946:
937:
930:
927:
924:
923:
892:
879:
866:
853:
838:
825:
794:
781:
768:
755:
729:
716:
703:
690:
677:
664:
651:
638:
625:
612:
599:
586:
580:D. Petersen, "
573:
560:
544:
531:
518:
505:
492:
479:
473:M. Ferguson, ”
466:
453:
440:
427:
414:
408:E. Kretzmer, "
401:
388:
374:
373:
371:
368:
367:
366:
361:
354:
351:
334:BCJR algorithm
316:
313:
303:
300:
294:
291:
281:
278:
276:
273:
253:
250:
212:
209:
205:100k ECL logic
182:
181:Tape recording
179:
153:
150:
122:Andrew Viterbi
89:
86:
39:of a magnetic
15:
9:
6:
4:
3:
2:
971:
960:
957:
956:
954:
945:
941:
938:
936:
933:
932:
913:on 2016-12-13
909:
902:
896:
889:
883:
876:
870:
863:
857:
850:
849:
842:
835:
829:
821:
817:
813:
809:
805:
798:
791:
785:
778:
772:
765:
759:
743:
739:
733:
726:
720:
713:
707:
700:
694:
687:
681:
674:
668:
661:
655:
648:
642:
635:
629:
622:
616:
609:
603:
596:
590:
583:
577:
570:
564:
557:
551:
549:
541:
535:
529:", circa 1982
528:
522:
515:
509:
502:
496:
489:
483:
476:
470:
463:
457:
450:
444:
437:
434:A. Viterbi, "
431:
424:
418:
411:
405:
398:
392:
385:
379:
375:
365:
362:
360:
357:
356:
350:
348:
344:
339:
335:
330:
326:
322:
312:
310:
299:
290:
288:
272:
270:
266:
261:
259:
249:
247:
243:
240:
235:
233:
229:
225:
224:IBM Rochester
220:
218:
208:
206:
202:
199:
194:
192:
188:
178:
176:
171:
167:
158:
149:
147:
143:
138:
134:
129:
127:
123:
119:
115:
114:
109:
107:
102:
94:
85:
83:
79:
75:
73:
69:
64:
62:
58:
54:
50:
49:areal-density
46:
42:
38:
34:
30:
26:
22:
943:
915:. Retrieved
908:the original
895:
882:
869:
856:
846:
841:
828:
811:
807:
797:
784:
775:D. Forney, "
771:
758:
746:. Retrieved
742:the original
732:
719:
706:
693:
680:
667:
654:
641:
628:
615:
602:
589:
576:
563:
534:
521:
508:
495:
482:
469:
460:D. Forney, “
456:
443:
430:
417:
404:
395:A. Lender, "
391:
378:
318:
305:
296:
283:
262:
255:
245:
236:
221:
214:
195:
184:
163:
130:
111:
110:
100:
99:
81:
77:
76:
72:PDNP or NPML
65:
33:digital data
28:
24:
18:
632:A. Patel, "
538:J. Smith, "
287:eye-pattern
232:Switzerland
142:Dave Forney
61:tape drives
917:2019-10-09
788:R. Wood, "
748:October 8,
658:R. Wood, "
486:T. Wood, "
370:References
325:FIR filter
228:IBM Zurich
203:(A/D) and
45:tape drive
41:disk drive
146:Bell Labs
131:By 1971,
108:channel.
57:hard disk
953:Category
353:See also
343:Broadcom
217:IBM 0681
347:Marvell
198:Plessey
911:(PDF)
904:(PDF)
166:Ampex
750:2019
345:and
329:TDMR
269:BIOS
239:PLCC
104:the
37:head
29:PRML
816:doi
327:. (
321:AGC
258:NRZ
175:RLL
170:IBM
137:IBM
135:at
74:).
59:or
55:on
43:or
19:In
955::
812:34
810:.
806:.
547:^
349:.
148:.
128:.
63:.
23:,
920:.
822:.
818::
752:.
27:(
Text is available under the Creative Commons Attribution-ShareAlike License. Additional terms may apply.
