Testing Op Amps and Comparators Daryl Hiser PA

Testing Op Amps and Comparators Daryl Hiser PA Op Amp -Test TI Information – Selective Disclosure

Op Amps, Comparators, and Pattern Based Testing……oh my! TI Information – Selective Disclosure

Outline • Basics of testing Op Amps. – The Loops – The Tests • Pattern Based Testing. • Basics of testing comparators. • Beware of flying monkeys! (common errors, issues, things to watch for) TI Information – Selective Disclosure

Op Amp Loops • If you can measure Vos well, most of the Op amp test battle is won! PSRR, CMRR, AOL, Drift… are just looking for changes in Vos as some parameter is varied. • We use “test loops” to facilitate measurement and control of op amps during test. The loop gives local gain of the offset measurement. • This presentation will evaluate two useful loop methods: - False Summing Junction - Two Amp Loop TI Information – Selective Disclosure

Eagle Op Amp HIB TI Information – Selective Disclosure

Op Amp test loops (False Summing Junction) +10 V + - 100 Vos i = Vos / 1 K Vos Offset Correction Factor = ACL + AFS Vo = -(Vi + Vos(O. C. F)) TI Information – Selective Disclosure ~ -10 V

Op Amp test loops (False Summing Junction) +10. 302 V +10 V + - 101 m. V + - i 2 = 98. 99µA i 3 = 97. 99µA 102. 01µA 101. 01µA + i 1= 1µA 1 m. V - Offset Correction Factor = ACL + ASJ Vo = -(Vi + Vos(O. C. F)) TI Information – Selective Disclosure -9. 698 V -10. 0 V ACL = 200 K/1 K + 1 = 201 ASJ = 100 K/1 K + 1 = 101 OCF = 302

Op Amp test loops (False Summing Junction) The worse the DUT’s DC precision the more significant loop errors become. Pro’s • Simple • Stable • Small (4 tiny surface mount resistors and you can test everything but IB TI Information – Selective Disclosure Con’s • Feedback R is a Load • Loop drive a function of DUT Vos.

Pro’s • No Load Current • Loop drive is independent from DUT Vos Op Amp test loops (2 Amp Loop) • 100 Vos • Requires Compensation!! 101 Vos +10 V TI Information – Selective Disclosure More complex Type 2 i = Vos / 100 Vos Con’s Type 1

Bode plot for 2 Amp Loop (uncompensated) Combined response Both Amps Closed loop Gain ACL (100 K/100) +1 = 1001 or ~ 60 d. B Null Amp ACL DUT 40 db/decade closure! Unstable !!!!! TI Information – Selective Disclosure

Type 2 compensation 80 p. F CF = 1 / (20 KHz x 100 KΩ x 2π) = 80 p. F ACL 20 db/decade closure! Stable !!!!! TI Information – Selective Disclosure

Type 2 compensation range Lowest pole frequency Optimum compensation ACL TI Information – Selective Disclosure Highest pole frequency

Bode plot for Type 1 comp. 100 p. F Combined response Both Amps ACL DUT Closed loop Gain ACL (100 K/100) +1 = 1001 or ~ 60 d. B 20 db/decade closure! Null Amp (integrator) Stable !!!!! TI Information – Selective Disclosure

Type 1 Compensation Range Intersection of Combined response with Closed loop gain (ACL) ACL Null Amp(integrator) zero frequency Optimum is intersection / 4 TI Information – Selective Disclosure

Optimizing compensation You can make bode plots to find the best compensation for either method. …Or You can use TINA spice and evaluate through simulation … … Or TI Information – Selective Disclosure

… use the easy button. Type 1 compensation Given Rf, Rin, Rc, BW 1 Type II compensation Given Rf, Rin, BW 1, BW 2: TI Information – Selective Disclosure

Op Amp testing basics Key Parameters Abbr. Continuity / Contact resistance Quiescent Current Input offset voltage Power Supply Rejection Common Mode Rejection Open Loop Gain Short circuit current Slew Rate Gain Bandwidth Input Bias current Cres IQ VOS PSRR CMRR AOL ISC SR GBW IB Noise Total Harmonic Distortion THD TI Information – Selective Disclosure

Continuity ***1 current *** Continuity is measured by sourcing a small current (100 u. A typ. ) through each IO pin through the ESD diodes to the positive supply, and then sinking current through the diodes connected between the IO pins and the negative supply. This verifies all ESD diodes are connected as well as all wire bonds intact and some contact to each pin. i 1 = 100µA 18 TI Information – Selective Disclosure

Contact Resistance ***2 currents*** CRES is measured between each supply and the output using 10 m. A and 15 m. A currents to get past the diode I/V knee. It is a bulk resistance includes 2 contact resistances and the internal diode resistance. Junction Voltage (V) 50 40 Current (m. A) i 1 = 10 m. A i 2 = 15 m. A 30 Diode I-V Curve 20 10 0 0. 2 0. 4 0. 6 0. 8 1. 0 Cres 1 = (Rp 1 + Rp 2) = [(V 1 – V 2)/ (I 1 – I 2) ] – Rd 1 Cres 2 = (Rp 2 + Rp 3) = [(V 1 – V 2)/ (I 1 – I 2) ] – Rd 2 Cres 3 = (Rp 1 + Rp 3) = [(V 1 – V 2)/ (I 1 – I 2) ] – Rd 3 Contact resistance > ~1Ω can be accounted for using this method. If 1Ω or less is a problem for measurements, Kelvin contacts should be used. 19 TI Information – Selective Disclosure

CRES and temperature ***3 currents*** Complete Diode Equations i 1 = 5µA i 2 = 35µA i 3 = 85µA CRES and Temperature accuracy ~ ± 5°C with ESD diode TI Information – Selective Disclosure

Quiescent current / Integrity / ISC Iq = Iq(meas) – empty socket null Output stage 1 m. A 1. 5 m. A ibias i. OUT = 1 m. A 0. 5 m. A One note of caution…beware of measuring load current by looking at the change from Quiescent current on the DUT supplies. 21 TI Information – Selective Disclosure

Vos / PSRR / CMRR PSRR = ΔVos / ΔVcc (μV/V) CMRR = ΔVos/ ΔVcm (μV/V) If (Ib x RF)/100 is significant to Vos it must be considered! ΔVcc ± CMV ΔVee Beware of oscillations… Op Amps “may” rectify low level oscillations because of input clamping diodes and this can appear as an erroneous offset voltage. 22 TI Information – Selective Disclosure

Example of big IB error on Vos measurement In this configuration there is a 1µV / n. A error on Vos. The 2 Amp loop with same R values would have the same error! = 150 m. V /loop gain = 1. 5 m. V 50% error!!! 1. 5µA 1µA 500 n. A 1 m. V Smaller resistors reduce error. ie 10 K and 100Ω would be 50 u. V error… TI Information – Selective Disclosure or Measure IB first and use it in Vos calculation.

CMRR considerations 0 V = -4 V • Moving supplies and output is equivalent to moving common mode voltage ! • Remember that you 0 V = 4 V can not short the inputs of an Op Amp together. Doing so will put it in the rail !!! 24 TI Information – Selective Disclosure

PSRR consideration for rail to rail amps 2. 75 V -0. 6 V 0. 9 V -0. 6 V -2. 75 V -0. 9 V p. Mos to n. Mos transition CMV 1 CMV 2 CMV 1 -0. 9 V 25 TI Information – Selective Disclosure

AOL / Swing / Slam Aol = ΔVos / ΔVout (μV/V) 26 TI Information – Selective Disclosure

Swing Aol / Slam considerations AOL Curve - Precision in drive for swing…. ”just enough” - Compare Vout to supplies…not ground - Remember RF in || with RL in FSJ TI Information – Selective Disclosure

Slew Rate 2 K • Slew of source must be faster than DUT • DUT slew should not be limited by parasitic RC’s • Some Op Amps are specified in unity gain… TI Information – Selective Disclosure

Gain Bandwidth Product GBP = test frequency * [Vout(rms) / Vos(rms)] The test frequency should be 2 x Unity Gain BW / loop gain TI Information – Selective Disclosure

Bode plot of gain vs frequency Open Loop 1 K • As frequency increases, parasitics may begin to dominate! Loop 100 Test frequency The higher bandwidth the greater the challenge! TI Information – Selective Disclosure LG = 1000 LG = 100

The dreaded i. B test… “It’s like trying to poke a cat out from under the porch with a rope” -Unknown- TI Information – Selective Disclosure

Some IB issues and challenges µA down to f. A Stability (Thermal EMFs) Leakages Humidity Environmental noise … TI Information – Selective Disclosure

nano. Amps for IB greater than ~200 p. A • Ios = Ib+ - Ib- (math not measure) • Current into the input is “positive” by convention. • identical currents will produce opposite changes in “vos” measurement… Ib+ gets multiplied by -1. IB = ΔV / Loopgain / Rib TI Information – Selective Disclosure

pico. Amps for IB 5 p. A to ~200 p. A IB = C d. V/dt / Loopgain TI Information – Selective Disclosure

femto. Amps ~6 electrons / m. Sec for IB < 5 p. A +V -V IB = C d. V/dt - null ***Facilitates nulling of leakage currents on empty socket calibration*** TI Information – Selective Disclosure

Step 1: I = C(d. V/dt) Run an open socket leakage test for each amplifier input and store values Leakage Step 2: Measure bias current of each input Step 3: Bias Current = IMEASURED – ILEAKAGE TI Information – Selective Disclosure ∆V Leakage + IB ∆T

Important considerations for low IB measurements • Low leakage Capacitor ( Teflon, glass, polypropylene polystyrene). • Low thermal EMF cancelled, mechanical latching relays w/ active high drivers. TI Information – Selective Disclosure

Thermal emfs are indistinguishable from Vos + emf Vos i = (Vos +emf)1 K Thermal emf TI Information – Selective Disclosure

Thermal emfs outside this ring are divided by the loop gain!!! Thermal emfs inside this ring cause 1: 1 errors TI Information – Selective Disclosure

Low thermal EMF cancelled, mechanical latching relays w/ active high drivers. Back to Low IB Active High vs Active Low drivers Active High Driver TI Information – Selective Disclosure Active Low Driver

Relay Drivers Open Collector Ib COIL Ib C int RELAY TI Information – Selective Disclosure 12 V COIL Active High Drive 12 V C int RELAY

Relay Drivers Open Collector Ib COIL Ib C int RELAY TI Information – Selective Disclosure 12 V COIL Active High Drive 12 V C int RELAY

Relay Drivers Open Collector Ib COIL Ib C int RELAY TI Information – Selective Disclosure 12 V COIL Active High Drive 12 V C int RELAY

Relay Drivers • Actively High driven Latching relays reduce: – Leakage (Coil Voltage) during test. (Good for Low IB) – Heating (Coil current) during test (Good for Low Vos) – EMI (Coil B field) during test (Good for both) Guard traces / Shielding • Exposed guard traces reduce surface leakage! Active guarding can further reduce leakage but adds complexity TI Information – Selective Disclosure

Guard traces / Shielding High Z trace Exposed Guard traces TI Information – Selective Disclosure

Guard traces / Shielding • Internal Guarding and Shielding (cross section ) Protected trace FR 4 material Ground planes grounded guard traces TI Information – Selective Disclosure

“Two for one” ~One integration time -> 2 ib measurements Ibn = C [(M 4 – M 1) / T 1] Ibp = C [(M 3 – M 2) / T 2] TI Information – Selective Disclosure

Current Ib Test Investigation • We are designing an octal site board with a different Ib measurement experiment on each site for the purpose of improving our understanding and techniques. • We plan to contact our National counterparts to see if they have any helpful practices or techniques that we could leverage or adopt. • If you have any suggestions or ideas, I would love to hear them!! TI Information – Selective Disclosure

Getting close… TI Information – Selective Disclosure

Pattern Based Testing Ø What is pattern based testing? Ø Why bother? Ø How can you implement it? TI Information – Selective Disclosure

PBT: Overview Post-process data in DSP memories! Avoid *slow* data transfer! TI Information – Selective Disclosure

Digging a bit deeper for Op Amp Testing • Predefined waveforms are generated. • One circuit configuration is selected that is most suitable for all tests • The output is digitized and stored • Digitized data is analyzed TI Information – Selective Disclosure

Pattern Based Testing: Analysis Typical Segment • Each test segment is composed of 2 elements: – Settling time – Measuring time • Analysis of the raw data will allow for correct time selection • This alone can reduce test times! Settle 12/7/2020 Measure “TI Proprietary Information - Strictly Private” or similar placed here if applicable TI Information – Selective Disclosure

Raw data Analysis • Raw data analysis provides many of benefits • The current testing methodology would not show all of the raw data. Only 1 averaged point would be returned. • Debugging and verification can be improved by viewing the raw data! Slight oscillation on some units. Adjust compensation. 12/7/2020 TI Information – Selective Disclosure Perfect! 54

PBT Benefits • Reduced test time 1. 5 seconds to 0. 3 seconds • Reduced Test % of COB • More thorough coverage at probe and FT Where do the time savings come from? TI Information – Selective Disclosure

Quiz On the Eagle tester, how long does it take 8 QMS’s grouped together to make a measurement if the number of samples is 100 and the sample time is 10 u. S? qmsmv( QMS_GP, QMS_PREC, 100, 10 ); Answer: ~ 9 m. S TI Information – Selective Disclosure

A comparison: Discrete vs PBT on ETS 364’s Action Group time cost Action Group time cost Control APU Group (Vcc) 8 m. S Control APU Group (Vee) 8 m. S Control APU Group (Vcm) 16 m. S Control APU Group (Vout) 16 m. S Control SPU Group (Vout) 16 m. S Control Dpins ? Control Cbits ? Settling time ? Control QMS Group 8 m. S ~60 m. S x discrete setups ( 11 ) Total “Setup” Time 660 m. S TI Information – Selective Disclosure ~60 m. S x patterns (1 -4) Total “Setup” Time 60 - 240 m. S

Future Work • Continue to refine the SW and analysis tools • Final Test implementation • Hardware improvements to support method • Expanding test coverage … Noise, Settling time, Drift? Yazdi Contractor has done some outstanding work to provide standardized tools for PBT on the Eagle Test platform!!! TI Information – Selective Disclosure

Comparators TI Information – Selective Disclosure

Comparator Test • Continuity • Output Swing • Propagation delay • Quiescent Current • Output current • Hysteresis • Vos • Bias Current • Psrr • Slew Rate aka rise time/ fall time • Cmrr Ø Similar to Op amps, if you can measure Vos, you are most of the way there on comparators… and there are many ways to do this. Ø Most Comparators are not “stable” in a loop like an op amp. Ø No loop gain means precision Vos measurements need a different strategy. TI Information – Selective Disclosure

Vos / PSRR / CMRR Synchronous Attenuator reduces AWG error (drift, need offset, lsb) *May to buffer andand potentially or filter. noise. TI Information – Selective Disclosure

Vos / PSRR / CMRR Asynchronous Just arm timer and run AWG TI Information – Selective Disclosure

t 2 t 1 ΔV = 2 x propagation delay* Vin ΔV V 1 Vos V 2 *Propagation delay of “system” Vos = (V 1 + V 2 ) /2 or Vos = Vin(max) - (t 2 – t 1) /2 * dv/dt Consider ramp speed and propagation delay vs test time! TI Information – Selective Disclosure

Hysteresis Vin V 1 Hysteresis V 2 A Vos measurement on a comparator with hysteresis requires a synchronous method or a one shot that is coupled to a sample and hold or triggers a measurement. TI Information – Selective Disclosure

Propagation delay Compare to shunt to make measurement “delta” • Limited by accuracy and or resolution of the timer / counter. ie QTMU accuracy is ~100 p. S with ~10 p. S resolution. • Trace lengths, attention to parasitics in layout become a big factor fast prop times. Speed of the “fast edge” is also a factor. Many comparators have asymetric prop. delays TI Information – Selective Disclosure

Propagation delay timing PDH TI Information – Selective Disclosure PDL

Propagation delay A cool circuit that might work… a ring oscillator shunt • • RC thenumber oscillation frequency. Candetermines measure the of counts for fixed time period t. PD = (time)/ Δcounts/2 or the time for fixed number of counts. t. PD = Δtime/counts/2 TI Information – Selective Disclosure

TINA says “looks good” • Without Comparator in the loop: • Number of count (Nc): 2 ^ 8 (8 -bit binary count) • Total Time (Tc): 40. 99 us • T 1 (=Nc/Tc) 160. 1 ns (Calculated) 160. 4 ns (Measured) • With Comparator in the loop: • Number of count (Nc): 2 ^ 8 (8 -bit binary count) • Total Time (Tc): 42. 88 us • T 2 (=Nc/Tc) 167. 5 ns (Calculated) 170. 8 ns (Measured) • Prop Delay (T 2 – T 1)/2 = 3. 7 ns Might be handy multisite testing where timing measurements need to be done serially! TI Information – Selective Disclosure

Special Thanks: Matt Picket and Yazdi Contractor For their work on PBT hiser_daryl@ti. com TI Information – Selective Disclosure

TI Information – Selective Disclosure

Needs • Our implementation of Pattern Based testing utilizes added functionality available only in the new ETS 364 APU-12 resource. More AWGS and more DSP’s and digitizers. • Fortunately, APU-12’s are the roadmap for all new ETS systems purchased by TI. • Unfortunately, there are only two APU-12 systems available in TLAB for development. * * 5 more systems have been approved but are in negotiation as to the final configurations. (sans MPU’s possibly) TI Information – Selective Disclosure

PSRR consideration for rail to rail amps 3. 35 V 1. 5 V 0. 6 V 0 V 0 V 0. 6 V -0. 3 V -2. 15 V CMV 1 CMV 2 CMV 1 72 TI Information – Selective Disclosure

TI Information – Selective Disclosure