Tests and Scope pictures of the 20 Amp Power Supply with External LC Filter ----------------------------------------------- Original Rev. 20-Apr-2017 Current Rev. 23-May-2017 These are the notes from the tests of the 20 Amp DCDC Converter operating with an External LC Filter for use as MGT supplies. The Initial Setup: ------------------ 1x 1000 uFd Al and 3x 100 uFd Tant at the Converter Input 4x 150 uFd Tant at Converter Output per L 680 nH filter choke 4x 150 uFd and 5x 470 nFd at Converter Output post L dcdc converter set to operate with 1.8 Volt output 10k Ohm VSet --> 1.8 Vout 30k Ohm VSet --> 1.0 Vout The static load consists of a variable number of 0.5 Ohm resistors across the external LC filter output. The pulsed load consists of one or two 0.5 Ohm resistors in the drain of a P30N06LE power FET. The ON resistance of this FET is ?? mOhm. I have 1k Ohm resistors in series with the gates and drive the gates to +- 3.7V with the HP generator and that appears to get the FET well into full saturation. Start with the converter setup for 1.8 Volts. Scope Picture Discriptions: --------------------------- 20-Apr-2017 -1 680 nH inductor 50 mV 200 usec vs+ pre L With vs+ post L see a very similar result. -2 Short across the L with a copper strap. All caps still in circuit. 200 usec vs+ pre L. Spike about 35 usec duration or about 8 cycles of the supply. Using a 470 nH inductor (vs 680 nH) makes very little difference. During both -1 and -2 the RC fb was 220 Ohm and 15 nFd. Increasing R to 700 Ohm gives more high frequency noise. With RC of 220 Ohm and 10 nFd connected pre L and 470 nH L and no pulsing see an oscillation of 3 or 4 mV and period of about 120 usec. With this setup an pulsing see bad ringing of about 6 cycles. Back to the 680 nH L and 220 Ohm 15 nFd RC with both RC and vs+ pre L. Try decreasing the R. There is a critical point when R reaches about 170 Ohm or less. The pulsing for all above have been 3.6/7.2 Amps at 500 Hz. -3 50 mV 200 usec avg 16 pulse 3.6/7.2 A 500 Hz vs+ pre L RC NOT connected 680 nH with filter caps as described at top of page 21-Apr-2017 -4 All as in 3 but: 2 mV 20 usec single shot with a constant 7.2 Amp load "best of 10" -5 All as in 3 but: 2 mV 2 usec single shot with a constant 7.2 Amp load "best of 10" 21-Apr-2017 -6 50 mV 200 usec avg 16 pulse 3.6/7.2 A 500 Hz RC of 255 Ohm and 22 nFd RC and vs+ both pre L 680 nH with filter caps as described at top of page -7 All as in 6 but: 2 mV 20 usec single shot with a constant 7.2 Amp load "best of 10" -8 All as in 6 but: 2 mV 2 usec single shot with a constant 7.2 Amp load "best of 10" 22-Apr-2017 -9 50 mV 200 usec avg 16 pulse 3.6/7.2 A 500 Hz RC of 255 Ohm and 15 nFd RC and vs+ both pre L 680 nH with filter caps as described at top of page 25-Apr-2017 -10 All as in 9 but: 2 mV 20 usec single shot with a constant 7.2 Amp load "best of 10" -11 All as in 9 but: 2 mV 2 usec single shot with a constant 7.2 Amp load "best of 10" 25-Apr-2017 -12 50 mV 200 usec avg 16 pulse 3.6/7.2 A 500 Hz RC of 255 Ohm and 10 nFd RC and vs+ both pre L 680 nH with filter caps as described at top of page -13 All as in 12 but: 2 mV 20 usec single shot with a constant 7.2 Amp load "best of 10" -14 All as in 12 but: 2 mV 2 usec single shot with a constant 7.2 Amp load "best of 10" 25-Apr-2017 -15 50 mV 200 usec avg 16 pulse 3.6/7.2 A 500 Hz RC of 255 Ohm and 100 nFd RC and vs+ both pre L 680 nH with filter caps as described at top of page -16 All as in 15 but: 2 mV 20 usec single shot with a constant 7.2 Amp load "best of 10" Note: oscillates with a 3.6A load about 20 mVpp -17 All as in 15 but: 2 mV 2 usec single shot with a constant 7.2 Amp load "best of 10" Note: oscillates with a 3.6A load 26-Apr-2017 With the FB RC replaced with just an air variable cap from pre L to Trim, anything more than 50 to 100 pFd just makes the ringing worse at the points of 3.6/7.2 Amp steps up or down. 26-Apr-2017 -18 50 mV 200 usec avg 16 pulse 3.6/7.2 A 500 Hz RC of 1.1k Ohm and 100 nFd RC and vs+ both pre L 680 nH with filter caps as described at top of page -19 All as in 18 but: 2 mV 20 usec single shot with a constant 7.2 Amp load "best of 10" -20 All as in 18 but: 2 mV 2 usec single shot with a constant 7.2 Amp load "best of 10" -21 50 mV 200 usec avg 16 pulse 3.6/7.2 A 500 Hz RC of 2.0k Ohm and 100 nFd RC and vs+ both pre L 680 nH with filter caps as described at top of page -22 All as in 21 but: 2 mV 20 usec single shot with a constant 7.2 Amp load "best of 10" -23 All as in 21 but: 2 mV 2 usec single shot with a constant 7.2 Amp load "best of 10" Also tried 100 nFd with 2.55k Ohm and 5.0k Ohm. Basically the same result as with 2.0k Ohm. 27-Apr-2017 -24 50 mV 200 usec avg 16 pulse 3.6/7.2 A 500 Hz RC of 2.0k Ohm and 100 nFd RC pre L vs+ post L 680 nH with filter caps as described at top of page -25 All as in 24 but: 2 mV 20 usec single shot with a constant 7.2 Amp load "best of 10" -26 All as in 24 but: 2 mV 2 usec single shot with a constant 7.2 Amp load "best of 10" Review: This has been a rough study of RC values with ------- all/most other conditions held constant: 50 mV 2 mV 2 mV 200 usec 20 usec 2 usec RC Values Pluse Static Static pre L 3.6/7.2A 7.2A 7.2A connection Avg 16 Single Single Notes ----------------- -------- ------- ------- ------- none 3 4 5 255 Ohm 22 nFd 6 7 8 255 Ohm 15 nFd 9 10 11 255 Ohm 10 nFd 12 13 14 255 Ohm 100 nFd 15 16 17 1.1k Ohm 100 nFd 18 19 20 2.0k Ohm 100 nFd 21 22 23 2.0k Ohm 100 nFd 24 25 26 vs+ post L 2.0k Ohm 100 nFd 27 28 29 vs+ post L plus 3x 470 uFd at LC Filter out see 1-May-2017 29-Apr-2017 - Adding 350 pFd from Trim to Gnd just makes the ringing worse (i.e. larger amplitude and about same frequency) at the 3.6/7.2 Amp transitions but still stable. - Increasing the Cap in the RC from 100 nFd to 500 nFd or 1 uFd or up to 10 uFd makes almost no difference at the 3.6/7.2 Amp transitions. Going from 100 nFd to 500 nFd makes perhaps a 5% decrease in amplitude of the ringing and then above 500 nFd this is no change. - Adding about 1500 uFd of mid quality Al electrolytic to the output of the External LC Filter helps a lot at the 3.6/7.2 Amp transitions and almost stops the ringing. This 1500 uFd is 3x 470 uFd 16V Al cap about 15 years old EEVKF1C471P PCE3402TR with an ESR of 160 mOhm. 1-May-2017 -27 50 mV 200 usec avg 16 pulse 3.6/7.2 A 500 Hz RC of 2.0k Ohm and 100 nFd RC pre L vs+ post L 680 nH with filter caps as above plus 3x 470 uFd Al -28 All as in 27 but: 2 mV 20 usec single shot with a constant 7.2 Amp load "best of 10" -29 All as in 27 but: 2 mV 2 usec single shot with a constant 7.2 Amp load "best of 10" 3-May-2017 -30 50 mV 200 usec avg 16 pulse 0.0/3.6 A 500 Hz RC of 2.0k Ohm and 100 nFd RC pre L vs+ post L 680 nH with filter caps as above plus 3x 470 uFd Al -31 All as in 30 but: 2 mV 20 usec single shot with a constant 0.0 Amp load "best of 10" -32 All as in 30 but: 2 mV 2 usec single shot with a constant 0.0 Amp load "best of 10" -33 50 mV 200 usec avg 16 pulse 10.8/14.4 A 500 Hz RC of 2.0k Ohm and 100 nFd RC pre L vs+ post L 680 nH with filter caps as above plus 3x 470 uFd Al -34 All as in 33 but: 2 mV 20 usec single shot with a constant 14.4 Amp load "best of 10" -35 All as in 33 but: 2 mV 2 usec single shot with a constant 14.4 Amp load "best of 10" Review: A rough study of various current steps and static ------- loads all with the setup of: RC of 2.0k Ohm and 100 nFd RC pre L vs+ post L 680 nH with filter caps as above plus 3x 470 uFd Al 50 mV 2 mV 2 mV 200 usec 20 usec 2 usec Pulse Pluse Static Static Current Static Step Load Load Step Load Avg 16 Single Single ------------------- -------- ------- ------- 3.6/7.2 A 7.2A 27 28 29 0.0/3.6 A 0.0A 30 31 32 10.8/14.4A 14.4A 33 34 35 4-May-2017 With the same setup, i.e. RC of 2.0k Ohm and 100 nFd RC pre L vs+ post L 680 nH with filter caps as above plus 3x 470 uFd Al Check the cold start startup waveform into the following static loads. These are all 500 mV DC 2 msec single shot -36 Startup into a 14.4 Amp static resistive load -37 Startup into a 3.6 Amp static resistive load -38 Startup into a 0.0 Amp static resistive load 5-May-2017 Now start test where the mid quality old 3x 470 uFd Aluminum capacitors at the output of the External LC Filter have been replaced with a fancy Wurth 1500 uFd 6.3V Polymer capacitors with 7 mOhm ESR. -39 50 mV 200 usec avg 16 pulse 3.6/7.2 A 500 Hz RC of 2.0k Ohm and 100 nFd RC pre L vs+ post L 680 nH with filter caps as at top plus Wurth 1500 uFd Polymer and 40 mOhm series --> Compare to #39 to #27 to see the difference between mid quality Al and fancy Wurth with 40 mOhm. 8-May-2017 -40 20 mV 200 usec avg 16 pulse 7.2/10.8 A 500 Hz RC of 2.0k Ohm and 100 nFd RC pre L vs+ post L 680 nH with filter caps as at top plus Wurth 1500 uFd Polymer and 20 mOhm series -41 20 mV 200 usec avg 16 pulse 7.2/10.8 A 500 Hz RC of 2.0k Ohm and 100 nFd RC pre L vs+ post L 680 nH with filter caps as at top plus Wurth 1500 uFd Polymer and 40 mOhm series 9-May-2017 -42 20 mV 200 usec avg 16 pulse 7.2/10.8 A 500 Hz RC of 2.0k Ohm and 100 nFd RC pre L vs+ post L 680 nH with filter caps as at top plus Wurth 1500 uFd Polymer and 60 mOhm series -43 20 mV 200 usec avg 16 pulse 7.2/10.8 A 500 Hz RC of 2.0k Ohm and 100 nFd RC pre L vs+ post L 680 nH with filter caps as at top plus Wurth 1500 uFd Polymer and 80 mOhm series -44 20 mV 200 usec avg 16 pulse 7.2/10.8 A 500 Hz RC of 2.0k Ohm and 100 nFd RC pre L vs+ post L 680 nH with filter caps as at top plus Wurth 1500 uFd Polymer and 10 mOhm series -45 20 mV 200 usec avg 16 pulse 7.2/10.8 A 500 Hz RC of 2.0k Ohm and 100 nFd RC pre L vs+ post L 680 nH with filter caps as at top plus Wurth 1500 uFd Polymer and Zero mOhm series Review: A rough study of various series resistors with the ------- fancy Wurth polymer output capacitor with all other conditions constant during a 7.2/10.8 Amp step. All: 20 mV 200 usec avg 16 pulse 7.2/10.8 A 500 Hz RC of 2.0k Ohm and 100 nFd RC pre L vs+ post L 680 nH with filter caps as at top plus Wurth and Rx Series Scope Resistance Picture ---------- ------- Zero mOhms 45 10 mOhms 44 20 mOhms 40 40 mOhms 41 60 mOhms 42 80 mOhms 43 Conclusion: 40 mOhms is probably best but either 20 or even 60 mOhm are OK. How much with the polymer capacitors internal 7 mOhm ESR increase over time ? 10-May-2017 -46 20 mV 200 usec avg 16 pulse 14.4/18.0 A 500 Hz RC of 2.0k Ohm and 100 nFd RC pre L vs+ post L 680 nH with filter caps as at top plus Wurth 1500 uFd Polymer and 40 mOhm series Observation: With operation up to about 15 Amps the converter remains fairly cool - not really warm to the touch - with only a small forced air flow over it. With an 18 Amp load it is distinctly warm to the touch with the same forced air flow. 11-May-2017 Now check operation with the PMBus connected and polling the converter once every 2 seconds and the SYNC pin grounded as it always should have been (but I don't think that this made any difference in the test so far). With "normal" pin numbering on the I2C Pod the pinout of its 0.1" x 0.1" 2x5 connector is: Open: 1:5 and 7 Ground: 6 I think this is the only Ground pin SMB_Alert_B: 8 Clock: 9 Data: 10 I2C and SMB_Alert_B Pull-Ups are in the I2C Pod. SYNC is connected to Signal_Ground Adrs_0 54.9k Ohm Adrs_1 36.5k (used 33.3K) Adrs = 28 So far usings just the DPI_GUI software. -47 20 mV 200 usec avg 16 Pulse 10.8/14.4 A 500 Hz RC of 2.0k Ohm and 100 nFd RC pre L vs+ post L 680 nH with filter caps as at top plus Wurth 1500 uFd Polymer and 40 mOhm series I2C connected as described above and running -48 All as in 47 but: 2 mV 200 usec single shot with a constant 14.4 Amp load "best of 10" -49 All as in 47 but: 2 mV 20 usec single shot with a constant 14.4 Amp load "best of 10" -50 All as in 47 but: 2 mV 2 usec single shot with a constant 14.4 Amp load "best of 10" Recall: AWG 16 is about 4.1 mOhm/ft AWG 18 is about 6.5 mOhm/ft 12-May-2017 Clean up the bread board by finally installing a nice well dressed twisted pair from the VS+ VS- pins to the sense points. If the sense points are at the far output end of the capacitor strips then there seems to be more high frequency noise than if the sense points are near the juntion of the external L and the post L output capacitors - why ? -51 20 mV 200 usec avg 16 Pulse 10.8/14.4 A 500 Hz RC of 2.0k Ohm and 100 nFd RC pre L vs+ post L 680 nH with filter caps as at top plus Wurth 1500 uFd Polymer and 40 mOhm series I2C connected as described above and running New clean twisted pair VS+ VS- otherwise same as #47. Now change to 1.000 Volt output, i.e. 30k Ohm R Vset. Running with 5x 0.5 Ohm static load and 2x 0.5 Ohm Pulsed load, i.e. 10 Amp / 14 Amp Static load 14 Amps -52 Now 1.000 Volt output. 20 mV 200 usec avg 16 Pulse 10.0/14.0 A 500 Hz RC of 2.0k Ohm and 100 nFd RC pre L vs+ post L 680 nH with filter caps as at top plus Wurth 1500 uFd Polymer and 40 mOhm series I2C connected as described above and running New clean twisted pair VS+ VS- otherwise same as #47 and #51. -53 All as in 52 but: 2 mV 200 usec single shot with a constant 14.0 Amp load "best of 10" -54 All as in 52 but: 2 mV 20 usec single shot with a constant 14.0 Amp load "best of 10" -55 All as in 52 but: 2 mV 2 usec single shot with a constant 14.0 Amp load "best of 10" Review: Compare 1.800 Volt and 1.000 Volt operation. ------- 1.800 Volt Pulse 10.8/14.4A Static 14.4A 1.000 Volt Pulse 10.0/14.0A Static 14.0A 20 mV 2 mV 2 mV 2 mV 200 usec 200 usec 20 usec 2 usec Pluse Static Static Static Output Step Load Load Load Voltage Avg 16 Single Single Single ------------ -------- -------- ------- ------- 1.800 Volt 47 48 49 50 1.000 Volt 52 53 54 55 Observations: The "DPI" I2C software is nice but it seems very touchy about what order things are started up: gui, 12V power to the POL converter, plug in USB, plug in the I2C bus to the converter. This is just as the instructions say one sould expect. Of the various "settings" that one can change, the Vin Startup and Vin Shutdown seem like ones that need to be adjusted for the Hub 12 Volt Vin operation. The current 20 mOhm resistors are 0.375 Watt size. 100 nFd in 0805 does not appear to be available in CG0 / NP0. Can get them in X7R and X8R. 13-May-2017 With everything setup for 1.000 Volt operation as described above in #52, look at some cold start power ups to verify that there is no oscillaton. -56 1.000 V setup as in #52 but cold start power up into a 14.0 Amp static resistive load 200 mV 2 msec -57 1.000 V setup as in #52 but cold start power up into an Open Circuit "load" 200 mV 2 msec 15-May-2017 With everything setup for 1.000 Volt operation as described above in #52, look at the power down waveform to verify that there is no oscillaton. -58 1.000 V setup as in #52 but record the power down rail collapse when running into a static 14.0 A load. Note that the 40400 buck controller was setup to turn off when its Vin fell below 5.5 Volts. Using the DPI Gui I had edited Vin Turn ON to be 6.0 V and Vin Turn OFF to be 5.5 V, then done a "Write Settings" to copy these gui values into the "working memory" of the 40400 controller and then done a, "Store Defaults" to copy these values from the controller's working memory into its non-volatile memory. scope picture 200 mV 1 msec. As noticed before, sometimes when I turn off the input power to the DCDC Converter module, the DPI Gui hangs up and will not work when the input power to the DCDC Converter is turned back on - but sometimes there is no problem. When it does hang up it appears that killing the Gui, then un-plug and re-plugin the USB connector (while the DCDC Converter is powered) will typically get the DPI Gui running gain. 17-May-2017 -59 This is basically a redo of scope picture #52. The only known change is that I have had the "Load Box" disconnected from the supply so that I could clean up the Load Box and add a second FET switch so that now each of its pulsed 0.5 Ohm resistors has its own switch. Stay at 1.000 Volt. 20 mV 200 usec avg 16 Pulse 10.0/14.0 A 500 Hz RC of 2.0k Ohm and 100 nFd RC pre L vs+ post L 680 nH with filter caps as at top plus Wurth 1500 uFd Polymer and 40 mOhm series I2C connected and running. Compare with #52 Now I want to see if things are at all rational using a smaller capacitor after the LC filter, so instead of the standard 1500 uFd 40 mOhm setup try a 750 uFd 40 mOhm combination. The 750 uFd is really just two of the Wurth 1500 uFd tried in series. -60 This is the same as #59 but now with 750 uFd 40 mOhm after the LC filter. 20 mV 200 usec avg 16 Pulse 10.0/14.0 A 500 Hz RC of 2.0k Ohm and 100 nFd RC pre L vs+ post L 680 nH with filter caps as at top plus Wurth 750 uFd Polymer and 40 mOhm series after the LC filter. I2C connected and running. Compare with #59 and #52 Yes, you can see a little more ringing but 750 uFd is not a completely ridiculous choice but it does shown one more cycle of ringing when pulsed.. -61 All as in 60 but: 2 mV 20 usec single shot with a constant 14.0 Amp load "best of 10" -62 All as in 60 but: 2 mV 2 usec single shot with a constant 14.0 Amp load "best of 10" Review: Compare now standard 1500 uFd 40 mOhm post LC network ------- with 750 uFd 40 mOhm 20 mV 2 mV 2 mV 200 usec 20 usec 2 usec Pulse Static Static Post LC Step Load Load Network Avg 16 Single Single ------------------- -------- ------- ------- 1500 uFd 40 mOhm 52 & 59 54 55 750 uFd 40 mOhm 60 61 62 I also looked at 750 uFd 60 mOhm with the pulsed load and it looked about the same. Both show a hint of the third cycle. For tests so far I have ordered: 820 uFd 2.5V 5 mOhm 5k Hr 8mm x 9mm Panisonic 1000 uFd 2.5V 7 mOhm 5k Hr 8mm x 9mm Panisonic 1500 uFd 2.5V 7 mOhm 2k Hr 8mm x 13mm United Chemi-Con 1500 uFd 6.3V 7 mOhm 5k Hr 10mm x 14mm Wurth if 1000 uFd works then Wurth also makes a 1000 uFd 6.3V 2k Hr @ 125 deg 8mm x 13mm The issue is long life and fitting under the FPGA heat sink. Milling slots in the FPGA heat sink also needs to be studied. Observation: DPI GUI Software gets hung up ------------ If you turn OFF the power supply then the DPI software can get screwed up and not work when you turn the supply back ON. The software is hung up in some way that it can not communicate with the PMBus power supply once it is turn back ON. I don't know if the problem is in the DPI software or in the USB hardware/software connection. To get things working again requires stopping the software, unplugging the USB, plug the USB back in, restart the software and set it up again, e.g. find the modules, set their Vset resistors,... A work around seems to be to always stop the polling of the modules before you turn OFF the power supplies. 19-May-2017 Review how closely the servo loop compensator currently matches the rest of the 20 Amp External LC Filter design. For this section I will use the reference designators from drawing #55 which mostly match the TI TMS40400 data sheet page 15 figure 17. Note that none of this is going to be exactly correct because of the 350 pFd Cux from the output of the error opamp to ground. - The integrator goes as one over R1 (C2 + C3) = 1.05 kHz or if you model this and not including the cap from the series RC feedback pair you have one over R1 x C3 = 31.8 kHz - The first high frequency zero is 1/ R4 x C2 = 2.54 kHz - The 2nd high freq zero is 1/ C1 (R1 + R3) = 15.2 kHz or if based on the external RC_Tune = 72 Hz - The first high freq pole is 1/ R3 x C1 = 612 kHz or if based on the external RC_Tune = 795 Hz - The 2nd high freq pole is 1/ R4 (C2 series C3) = 77 kHz - The first (built in L) LC resonance is at 470 nH with 604 uFd = 9.45 kHz - The external LC resonates at 680 nH with 2513 uFd = 3.85 kHz or if you model it as the series of the caps before and after the external L then it is 680 nH with 487 uFd = 8.75 kHz - The zero from the first set of output capacitors (i.e. before the external L) is at about 470 uFd and 12 mOhm = 28.2 kHz - The zero from the second set of output capacitors is about 1500 uFd and 32 mOhm = 3.32 kHz 23-May-2017 -63 This is basically the same as: #52, #59, #60 but now the RC network post LC is 1000 uFd and 40 mOhm. The rest is the same: 100 nFd 2k Ohm RC pre L, vs+ post L, 10/14 Amp pulse at 500 Hz, 1.000 Volt output -64 This is the same as #63 but now the RC network post LC is 1620 uFd and 40 mOhm, i.e. two 820 uFd in parallel. Review: Compare now standard 1500 uFd 40 mOhm post LC network ------- with 750 uFd 40 mOhm 1000 uFd 40 mOhm 1620 uFD 40 mOhm 20 mV 200 usec 10/14 Amp Post LC Step Network Avg 16 ------------------- -------- 1500 uFd 40 mOhm 52 & 59 750 uFd 40 mOhm 60 1000 uFd 40 mOhm 63 1620 uFd 40 mOhm 64