QST Noise Report - us
Contents
1.2. 1632 RWA WITE322 v 2 56 software and a Philips TDA5360 preamplifier with a high frequency bandwidth of 240 MHz All AC SNR measurements were made using a LeCroy DDA 120 capable of capturing 2GS s All BER and MSE measurements were made using custom Westlake hardware using a Rembrant channel All tests were run with a 160MHz 1T data rate All QST tests were run on an ISI 2002 QST using 2002 compatable HGA cartridges Quasi97 software version 2 0 6 QST transfer curves were run from 200 to 200 Oe with 0 2 Oe steps with no averaging or stresses The QST Noise Test setup was developed in the course of this work and is detailed in the QST Noise Test Setup appendix IPA was used in all head mounting processes to reduce tribo charging amp handling related ESD events EXPERIMENTAL RESULTS Thirty 30 as received Atlantis20 MKPI heads were screened using QST transfer curves to find five 5 kinked heads Kinked heads are heads with non uniformities in their transfer curves which have been traditionally associated with damaged or defective GMR elements Recently kinks have been sub categorized as hard kinks or soft kinks where a hard kink is an instantaneous change in resistance at given field infinite local transfer curve slope and a soft kink is a increased sensitivity over a range of fields increased transfer curve slope Hard kinks often appear with open loops which are quantified as transfer loop hysteresis
3. Bias 4 5 Part ID A20 18 Head FarHd 0 100 50 0 50 Magnetic Field Oe 100 50 200 150 100 50 0 50 Magnetic Field Oe 100 150 200 Figure 3 QST transfer and 2D noise curves for head A20 18 showing a soft kink at 0 Oe 2000 350 1500 300 E 1000 250 o 3 500 E amp 200 S E 3 0 2 E S 150 E E 500 1000 1500 2000 Transverse Transfer Curve Bias 4 5 Part ID A20 20 Head FarHd 0 100 50 0 50 Magnetic Field Oe 200 400 gt o 50 200 150 100 50 0 50 Magnetic Field Oe 100 200 Figure 4 QST transfer and 2D noise curves for head A20 20 showing a soft kink at 40 Oe 400 Forward Reverse 2000 350 1500 300 ze 1000 250 o 3 500 a 3 E 200 8 2 4 o E 0 S 150 E gt 1000 Transverse Transfer Curve Bias 4 5 Part ID A20 27 Head FarHd 0 Magnetic Field Oe 100 50 200 150 100 50 0 50 Magnetic Field Oe 100 150 200 Figure 5 QST transfer and 2D noise curves for head A20 27 showing a soft kink at 40 Oe There are many interesting things to comment on in these plots First of all it can be seen that the QST Noise Test seems to be tracking the positive going red transfer curve This is because the QST Noise Test currently can only scan from negative fields to positive fields
4. Everyone over at ISI for building the QST 2002 developing new tests and fixing software bugs as quickly as we could find them Appendix I The QST Noise Test Setup INTRODUCTION To improve MR instability characterization a new QST Noise Test has been implemented ISI manuals and software refer to this test as a popcorn test 2 but has been discovered that this test detects any transient noise spike Thus this test will be refered to as the QST Noise Test for the remainder of this document The QST Noise Test stresses the reader by passing current through the writer before capturing a short read window It then checks to see if any events exceed a given threshold For example it can count how often amplitude spikes above a given threshold The strength of this test is that it can be run many times in a short period of time so the true statistical nature of instabilities can be measured The QST Noise Test runs this routine with different applied fields levels to quantify transient events at different points on the transfer curve It should be mentioned that the write excitation is unnecessary in producing kink related transient events but it would help detect heads with true popcorn noise TEST SETUP The QST Noise Test require a 2002 or 2001 generation HGA cartridge with SMB cables connected Quasi97 software version 2 0 6 or higher released 3 01 and a high frequency popcorn board QSTs with older hardware or software are not
5. capable of running this test To setup this test you must first configure QST hardware amp software for conventional HGA testing 2 just as if you were going to capture the transfer curve for an HGA This can be done by manually creating a new setup and configuring the system menu for the correct fixture amp bias level or by opening an existing HGA test setup Once you have a setup open you should push the popcorn button in the test menu bar This should bring you to the Popcorn test window pictured on the following page The settings shown in the picture are the suggested QST Noise Test settings and are described as follows Cycles number the number of read windows sampled At least 500 cycles is suggested to ensure good statistical sampling More would be better but this would also cause the test to take longer to run Filter selection This selects which QST filter is used when reading All data in this report was collected with the high frequency filter but this was default and may require more thought Test Mode selection This selects if the popcorn test is run in a single environment set field MR bias threshold write current frequency etc in popcorn mode or if one or more parameters is swept popcorn sweep mode For a standard QST Noise Test the popcorn sweep mode is used Amp Rate Qualify buttons selects amplitude or change in amplitude rate as the metric for the threshold limit Amplitude qualification i
6. of these unstable heads Analysis is as follows Basic DET data RW offset read and write width LF MF and HF TAA TAA asymmetry PW50 and PWS50 asymmetry did not flag any of the unstable heads With the possible exception of TAA asymmetry which flagged two kinked heads A20 18 amp A20 27 but missed a very similar third head A20 20 Thus this could be random scatter in the TAA asymmetry data It should be noted that the sign of TAA asymmetry agrees 4 4 with the sign of kink field location This is because kinks at negative fields are only seen by negative pulses and kinks at positive fields are only seen by positive pulses DET instability tests were not much better Baseline popping detected one of the four 1 4 kinked heads amplitude covariance with and without write excitation caught two of the four 2 4 and the Guzik s sector amplitude stability flagged all of the unstable heads 4 4 All of these results are somewhat arguable in that it is not known if these tests were setup for optimal performance and more data collection would probably be required before a true good value is established for each test Spin stand SNR results were also quite interesting Instantaneous auto correlation signal to noise iAC SNR a metric which is considered to have a great deal more scatter then average auto correlation signal to noise aAC SNR was the better indicator of kinked heads identifying all four 4 4 while aAC SNR only ide
7. the most useful in documenting GMR instabilities Sweeping the threshold allows one to distinguish while sweeping the field tests the MR element at different locations artifacts above the noise floor on the transfer curve eglp peusem duy umodog saper s18jaumimg peg synsay L 20 piar An pioysasy du START SERENA Sor We ARG muera SlayaWwele daa co esngezybig afg ed sug ny Ol sn peay sn seq 01 5n ewm Dz tzuyi fouanbary D tenipie DS isu wueung eui sts 5u glAn ploysary eye d n An piousen dwy F3 daams wosdo4 apoyy isa pgz weg dweaig ZH 08 0 3 sey eng aey Aeng dwg Aj oos epi saang Iy MOUS s9 ea2G J0 1252H amI qns enr ure sazyibig uio2dog sa uro3dog gt The QST noise test balances two basic issues time required to run the test and the level of detail the data contains The QST noise test can be run quite quickly less then 30 sec if large amplitude threshold steps and short write excitations are used Large threshold steps seem acceptable as all instabilities recorded to date have spanned multiple threshold levels Likewise short write excitation also seems acceptable as kink related instabilities seem to be independent of write duration The following section illustrates sample QST Noise Test data for a head with a kinked transfer function Note the QST N
8. 0 16 Head FarHd 0 Amplitude uV E S 1000 1500 2000 200 150 100 50 0 50 100 Magnetic Field Oe Figure 7 QST transfer curve and Guzik roll off curve for head A20 16 showing a hard kink at 150 Oe And a soft kink at 100 OE Transverse Transfer Curve Bias 4 5 Part ID A20 18 Head FarHd 0 2000 1500 1000 Ampitude uV o S 200 150 100 50 0 50 Magnetic Field Oe Figure 8 QST transfer curve and Guzik roll off for head A20 18 showing a soft kink at 0 Oe Transverse Transfer Curve Bias 4 5 Part ID A20 20 Head FarHd 0 Amplitude uV 200 150 100 50 0 50 Magnetic Field Oe 3 0 2 5 2 0 0 5 3 0 2 5 100 200 300 Frequency kfci 400 500 600 2 0 0 5 0 0 o 3 0 25 100 200 300 Frequency kfci 400 500 2 0 0 5 100 200 300 frequency kfci 400 500 Figure 9 QST transfer curve and Guzik roll off for head A20 20 showing a soft kink at 40 Oe 600 Transverse Transfer Curve Bias 4 5 Part ID A20 27 Head FarHd 0 Forward Reverse 3 0 2 5 2 0 TAA mV 0 5 0 0 0 100 200 300 400 500 600 Frequency kfci 50 donate Teld Oe Figure 10 QST transfer curve and Guzik roll off for head A20 27 showing a soft kink at 40 Oe These roll off curves do not appear very use
9. Comparison of the QST Noise Test to Other HGA Level Tests Mark Nichols Advanced Test Engineer Advanced Heads Media Advanced Electrical Characterization Quantum Advanced Development Labs March 26 2001 INTRODUCTION Recently there has been a great deal of interest in documenting instabilities in GMR heads One of the primary methods used to investigate these instabilities has been to capture MR transfer functions showing resistance vs applied field using a quasistatic tester QST Unfortunately this method has many drawbacks including no successful 10096 correlation to decreased head performance 1 no stresses at different field levels minimization of transient events due to averaging and reduced statistical accuracy due to limited numbers of test iterations The QST Noise Test is part of the standard ISI QST test software package which has only recently been debugged to the point that it has become useful The ISI software and manuals call this a popcorn test 2 unfortunately it does not exclusively detect popcorn noise nor has it detected a true popcorn event to date It should be mentioned that this is not due to any weakness in the software just that heads with popcorn events are rare today just as other types of head instabilities were rarer in the past The basic algorithm that the QST Noise Test uses is as follows it excites the head with a write event waits for a user defined number of microseconds to avoid noise unrelat
10. d DET tests are poor at identifying kinked heads with the possible exceptions of Guzik sector amplitude stability an infrequently used standard Guzik test Roll off noise amp base line popping both which identified the one open loop kink in the study TAA asymmetry which correctly identifies the polarity Oe of a kink if one is present e iAC SNR which requires external hardware identifies these kinked heads but aAC SNR does not AAC SNR aAC SNR iAC SNR identified all kinked heads Kinked heads identified by QST tests but missed by DET show worse MSEs amp BER This does not mean they would fail self scan but that they are inferior to non kinked heads from the same wafer REFERENCES 1 J Himle R Cross M Greenwell Drive Level Instabilities Correlated to Quasi Static Field Testing MMM Intermag 2001 talk BB 01 preprint Jan 2001 2 Integral Solutions Int 1 QST 2001 HGA HSA HDA Quasi Static Tester User s Manual Rev 1 9 3 1 www isiguys com 1999 3 W Deming Some Theory of Sampling pp 70 1 Dover Publications Inc 1950 4 G Mian T Howell Determining a SNR for an Arbitrary Data Sequence by a Time Domain Analysis IEEE Trans on Mag V29 6 pp 3999 4001 Nov 1993 THANKS Special thanks to Al Wallash amp Steven Lambert for resources and guidance Hai Nguy Bruce Lairson Bill Higgins Freddie Larios and everyone else in Quantum s Advanced Head Media group for help and camaraderie
11. due to software limitations so the noise test field history at a given field is identical to the positive going red transfer curve history but different then the negative going green transfer curve history Second the QST Noise Test only indicates noise from kinks it sweeps over For example very little noise is seen in QST noise plot for the soft kink at 100 Oe in head A20 16 in Figure 2 This because the field was not swept past the kink The QST Noise Test field limits used here were set to approximate the field a head will see in a drive The exact field limits these heads would see in operation is impossible to know as they would depend on many unknowns such as local media Mt fly height pole tip recession and local transition density due to linear superposition of magnetic fields Lastly the magnitude and field range of the kinks is very clear in the QST noise plots For example the soft kink in head A20 20 in Fig 4 is clearly smaller in the QST noise plot then the soft kink in A20 27 in Fig 5 both in field range and magnitude which is not immediately obvious in the QST transfer curves or transition curve metrics such as Barkhausen jump noise or hysteresis detailed later Before moving on it should be reiterated that while transfer curve and noise plot data are both collected using a QST the actual measurements are quite different It is really quite exciting that there is such good agreement DET testing was also conducted on t
12. ed to the head digitizes a user defined duration read window before finally checking to see if any digitized point in this window exceeds a user defined threshold This process is run repetitively at different amplitude thresholds and fields to create 3D transient count vs amplitude threshold vs field plots These complicated 3D plots can be converted into a simpler 2D average maximum amplitude vs field plots Sample plots and mathematics are included in the QST Noise Test Setup appendix at the end of this document The point behind constructing average maximum amplitude vs field plots is to quantify amplitude instability associated with kinks seen on transfer curves The use of average maximum amplitude is important because many traditional amplitude measurements minimize the effect of infrequent transient events For example Guziks generally measures track average amplitude TAA which is the average amplitude seen over one disk revolution at a given track position Thus if an instability adds an additional 1096 amplitude to 146 of the pulses it will show up as an additional point one percent 0 196 TAA but could result in all affected bits being read incorrectly 1 error 1e 2 BER The main focus of the experimental portion of this report is to compare the QST Noise Test to other HGA level tests such as QST transfer curves amp DET testing EXPERIMENTAL SETUP All DET tests were run on a Guzik 1701 spin stand equipped with 961 ANA a
13. following tests were used to detect known kinky heads with correct kinky head percent in parentheses QST transfer curves 4 4 this test was used to identify kinky heads QST noise test 4 4 but media fields are assumed to be 50 to 50 Oe Guzik parametric measurements 0 4 Guzik TAA asymmetry 2 4 Guzik baseline popping test 1 4 but possibly all hard kinks with open loops Guzik amplitude variance 1 4 Guzik sector amplitude stability 4 4 Guzik roll off curve spikes 1 4 but possibly all hard kinks with open loops Guzik roll off curve slope changes 1 4 possibly indicating soft kinks External iAC SNR averaged 3x 4 4 External aAC SNR 1 4 External AAC SNR 4 4 MSE amp BER 4 4 this is somewhat redundant as BER was used to select good heads In summary the Guzik sector amplitude stability seemed the best pure Guzik solution involving no external hardware Roll off curves and base line popping seem to be tied at a distant second with more data required to judge if it catches all open loops iAC SNR seems to be the best hybrid DET test but requires external non Guzik hardware MSE amp BER which also requires extensive external hardware amp software are also excellent tests though this is somewhat redundant as it was used as a metric for determining if a head was good or not in the first place CONCLUSIONS QST transfer curves and the new QST Noise Test clearly identify kinked heads e Standar
14. ful at first but there are many interesting details First the head with the hard kink and open loop A20 02 produces a noisy roll off curve This noise is unusual in that it does not diminish with averaging and seems to be jumping between two levels one with more noise and one with less This behavior suggests that the noise is not gaussian but digital Jumping between two states This makes sense after the transfer curve is examined as there is an open loop This open loop indicates the element could be in either a high or low resistance amplitude state at fields between the open loop limits Thus a noisy roll off curve could be the signature of an open loop The effect of a soft kink on roll off curves is more subtle The last three heads A20 18 20 amp 27 all have soft kinks but only head A20 18 shows a dramatic effect The soft kink appears to increase the roll off curve slope at a frequency which might generate a maximum field equal to the kink field There is of course no roll off curve noise as there are minimal open loops in these transfer curves lt 2 transfer loop hysteresis Heads A20 18 and A20 27 do not show a clear increased slope region In the roll off curves it is assumed that the media used in this study did not emit fields greater then 40 Oe Or perhaps kinks above 40 Oe are obscured by the onset of linear superposition which causes curvature in roll off curves EXPERIMENTAL RESULTS SUMMARY The
15. hese kinky heads yielding the following data Data points in red are considered outliers where outliers are defined as data points outside the measured range of the ten control heads Average non kinked values have also been included in the GOOD column for comparison Table 1 DET amp QST test results for five kinked heads amp average of ten non kinked GOOD heads ID Rmr RW r Wwrite Wreaa LETAA MFTAA HFTAA TAA ym PWS0 PW asym pem Gram Gram Gra ev ev o jam A20 02 47 8 aoao 453 s6 20 f T0 200 149 Uf T fnaf o Eq c GOOD 428 66 237 146 1 280 0 953 0 366 0 9 142 0 5 ID BLpop ACow ACow C ACov C ACows Vigma Vsigma Vranget Vrmgo SAS mV SAS mV SAS mV SAS mV iAC SNR aAC SNR AAC SNR MSE BER QST NES OST Dm OST ve a e p Gm o9 m A20 02 14 2 18 0 194 1 6e 4 xm As 383 395 sete a 2 A20 18 100 14 8 4 8 231 gt l Te 3 101 moa e wa s2 Mir ses 3 1 31 9 1 LI L T1 1 3 H GoOD 186 199 13 106 roio 5 5 Again there are a number of interesting things in this data set First of all QST transfer and noise plots showed all but head A20 16 were unstable in estimated normal spin stand conditions had a hard or soft kink between 50 and 50 Oe which caused increased average max amplitude in the QST Noise Test Thus it was hoped one of the DET tests would flag all
16. ntified one 1 4 kinked head Hai Nguy suggested that this made sense as aAC SNR attempts to measure pure media noise by averaging while iAC SNR is a raw noise which includes head amp electronic contributions 4 By calculating AAC SNR aAC SNR iAC SNR we find another good instability flag which identifies all four 4 4 kinked heads MSE and BER acted as expected with the good heads having the lowest MSE and BERs This is somewhat redundant as BER and indirectly MSE are key metrics in identifying good heads So good heads do in fact have good BERs Two transfer curve metrics have also been included for comparison Barkhausen jump noise and hystersis Both of these metrics correctly identified the hard kink but missed the three soft kinks This was expected as the soft kinks do not have large point to point deviations Barkhausen jump noise or open loops hysteresis The last metric that was evaluated was amplitude roll off TAA vs frequency plots Once again the QST transfer curves have been included for reference Transverse Transfer Curve Bias 4 5 Part ID A20 02 Head FarHd 0 3 0 Forward Reverse 2 5 0 5 0 0 200 150 100 50 0 50 100 150 200 0 100 200 300 400 500 600 Magnetic Field Oe Frequency kfci Figure 6 QST transfer curve and Guzik roll off curve for head A20 02 showing hard kink at 0 Oe Transverse Transfer Curve Bias 4 5 Part ID A2
17. oise Test field must sweep beyond a kink to detect it so small field sweeps are not advised Transverse Transfer Curve Bias 4 5 Part ID A20 20 Head FarHd O Forward Reverse 500 Ampiitude uv 1000 1500 2000 200 150 100 50 0 50 100 150 200 Magnetic Field Oe Figure 1 Sample QST transfer plot for a head with a soft kink Popcorn Test Figure 2 Sample 3D event count vs amplitude threshold vs field QST Noise Test plot This 3D plot is rather tough to absorb but it can be converted into a simpler 2D QST Noise plot using excel or other offline tools Note the raw data must be recorded to accomplish this by selecting capture raw data in the data logging window A sample 2D average amplitude vs field graph is shown as follows 400 350 300 250 200 150 Average Amplitude uV 100 50 200 150 100 50 0 50 100 150 200 Magnetic Field Oe Figure 3 Sample average maximum amplitude vs field plot for a kinked head Note both the 3D and 2D plots contain the same information The 2D graph converts the average max amplitude cumulative hystigram into average max amplitude values using the equation Vav X PiVi where P is the percentage events counted total count and V is the value amplitude threshold Average max amplitude standard deviations can also be calculated using the equation ST DEV PiViy X Pi vi 3
18. s used in the standard QST Noise Test Amp Rate Threshold values Sets the amplitude or rate threshold depending on which is being used Note these inputs are overridden by the popcorn sweep parameters below in sweep mode Write current amp frequency values This is the current and frequency used by the QST to excite the head preceding the read window All noise documented to date has been independent of this write excitation thus changes in write excitation did not influenced QST noise results in any way even at very low current write excitations Write Delay Read Ties values These are the amounts of time the head is excited by a write event delays before starting to read to avoid switching amp thermal ramp effects and time the read window is open Non head related noise spikes have been recorded lt lusec after the write possibly due to electronic switching in the QST hardware thus a lusec delay is suggested Also as noted above all the noise measured to date has been write independent so changes in write duration do not seem to change QST noise results Multi Events amp Digitization buttons are debugging features that should be turned off when the QST noise test is run More information on these features is available in the QST manual 2 Sweep Parameters selections these entries determine what elements of the test environment are swept during the QST noise test Sweeping amplitude threshold and test field has proven
19. values This transfer curve hysteresis is the area enclosed by a positively swept transfer curve red curve in the figures below and a negatively swept transfer curve green curve in the figures below Extensive QST amp DET data was collected for the five 5 sorted kinked heads and ten 10 non kinked control heads for comparison Captured QST transfer curves and QST noise plots for the kinked heads are as follows dotted lines on QST noise plots indicate the maximum experimental data point Transverse Transfer Curve Bias 4 5 Part ID A20 02 Head FarHd 0 Average Amplitude uV 200 150 100 50 0 50 100 150 200 200 150 100 50 0 50 100 150 200 Magnetic Field Oe Magnetic Field Oe Figure 1 QST transfer and 2D noise curves for head A20 02 showing multiple hard kinks amp open loop centered at 0 Oe Transverse Transfer Curve Bias 4 5 Part ID A20 16 Head FarHd 0 400 350 300 ez 2 250 o 2 amp 200 E 4 S 150 lt 100 50 0 200 150 100 50 0 50 100 150 200 200 150 100 50 0 50 100 150 200 e Magnetic Field Oe Magnetic Field Oe Figure 2 QST transfer and 2D noise curves for head A20 16 showing a hard kink at 150 Oe and a soft kinkat 100 Oe 400 2000 350 1500 300 1000 gt S 250 500 a amp 200 0 E lt Hd 150 500 S lt 1000 1500 2000 Transverse Transfer Curve
Download Pdf Manuals
Related Search