|
|
|
|
|
|
BMW Garage | BMW Meets | Register | Today's Posts | Search |
|
BMW 3-Series (E90 E92) Forum
>
DIY - Server Power Supply for Coding
|
|
07-13-2023, 08:22 AM | #24 |
New Member
2
Rep 12
Posts |
mate, very small flat screwdriver will do, thats not an issue, if it is, buy a one or set flat screwdrivers for the watches and a lupe so you can see what you are doing.
|
Appreciate
0
|
10-08-2023, 01:32 PM | #26 | ||
Private First Class
62
Rep 108
Posts |
Thank you for sharing this great info! I've done similar PSU mods and have some specific questions regarding your comments:
Quote:
Quote:
3. But what is the function of the 75k(82k) resistor(i.e. what's connected at the left/right ends of the 75k(82k) resistor)? Is this resistor reducing the voltage sensed by the OVP circuit(compared with actual output voltage effectively "tricking it") so the PSU is not shutdown with increased voltage?* * I ask as you seem to have not lifted/disconnected any pins related to an OVP circuit's IC. 4. You did not switch out the output capacitors to say 25V caps. Wouldn't the higher voltage caps be safer? 5. Do you have / can you share a schematic of the PSU? Thank you for your clarifications! Last edited by tinkerman; 10-08-2023 at 02:36 PM.. |
||
Appreciate
0
|
10-08-2023, 04:24 PM | #27 |
Private First Class
471
Rep 170
Posts |
1. Pin 36 is wired to a pull-down resistor, which means it isn't true ground. The PSU isn't connected to a real server socket, so the board won't fight you in trying to pull up the connection, but if you have the option between true ground on pin 32 and a pull-down on 36...Might as well choose true ground.
2/3. Exactly, the added trim pot resistor changes the voltage divider to allow the voltage to be set higher than designed (We need 14+V instead of the ~12V desired by a PSU), and adjusts the over-voltage protection. 4. I didn't look at changing out the other parts, since server PSUs are typically over-designed. Besides, this was done on the cheap to avoid paying $700 for a true power supply. 5. I don't have a schematic, this was based on similar online modifications, and brief look at the circuitboard. (It was also tested on the bench with a dummy load before connecting it to the car) 대한민국 축구팀 아시안게임 잘했어! |
Appreciate
1
tinkerman62.00 |
10-08-2023, 04:45 PM | #28 | |
Private First Class
62
Rep 108
Posts |
Quote:
You mentioned "Solder a 22k Resister between Pins 33 and 32.", but the photo shows pins 36 & 37. I'd appreciate your clarification on which is correct and what is with this discrepancy. THANK YOU! Last edited by tinkerman; 10-08-2023 at 05:44 PM.. |
|
Appreciate
0
|
10-08-2023, 05:46 PM | #30 | |
Private First Class
62
Rep 108
Posts |
Quote:
There is a very cool mod video of a slightly different but similar model DPS-800GB in which the gent lifts the OVP circuit's sense pin, then attaches various resistors (including a 2k variable resistor @ 09:50 into the video) between pin 32 and ground to achieve a similar result of raising the output voltage. I especially like the variable method, which is how I've done my own previous mods on various PSU units. Is this possible because pin 32 at least in this different model is associated with a 'voltage divider' connected to voltage regulation or is this related to an entirely different concept/function? Do you think this would work with the DPS-1200FB? Last edited by tinkerman; 10-08-2023 at 06:13 PM.. |
|
Appreciate
0
|
10-09-2023, 10:25 AM | #31 | |
Private First Class
62
Rep 108
Posts |
Quote:
He states essentially that pin 36("Present") is likely intended for the PSU to sense being "connected to the server" by a 'voltage pull up' when inserted into a server's slot. This 'pull up' of its voltage can be mimicked by connection with pin 37's "Standby 12V" or pin 33's "Enable" voltage ("pulled up to the PSU internal 3V3 bus via a 10K resistor"). But what is really revealing is that he goes on to suggest that the voltages, resulting from the range of usable resistors, between pin33 and pin36 are "unconventional." Apparently the "splitting of 33 & 36, into a pulled up / clamped "present"(36) signal, and a simple pull to ground "enable#"(33) signal (which can also be driven over I2C) is much more consistent with other PSU" he has worked with. Further, "a fixed 22K resistor from "Present"(36) to 12VSB(37), and a separate switch between Gnd and Enable#(33), ... leaves the possibility of software control over the PMBus(Power Management Bus)." In other words, one can do more with "conventional"(as intended) connections(than the method of connecting pins 33 and 36 which 'does the trick' but limits other control features, that would otherwise be available). Pin 34 is apparently a "Current Monitor" output (60mV/A) so it's got a current sensor built-in! Therefore, a "6:1 resistive(voltage) divider to ground would give 10mV/A, and you could then connect a cheap 0-1V voltmeter module, ideally with programmable decimal place, to directly display current in A"(amps) which means adding a bulky current shunt or even an inline Allegro current sensor IC(as I'd thought) is not even necessary! Dang! This should apply to many similar model HP Common Slot Power Supplies so this is very useful info! Cheers! Last edited by tinkerman; 10-09-2023 at 07:17 PM.. |
|
Appreciate
0
|
11-08-2023, 09:43 PM | #32 |
Private
18
Rep 69
Posts
Drives: 2010 328i
Join Date: May 2012
Location: Kenmore, WA
|
Hi All,
I just finished the OVP and voltage mod on an HP 1200 watt server PSU, model HSTNS-PL11. Successfully boosted it to 14.2 volts, then tested under load to 55 Amps, worked great. Some tips for future modders: 1. Follow Jagaer's instructions for grounding out pin 33, then connecting pins 36 and 37 with a 22k resistor to turn on the supply! There are lots of videos on youtube that say to just connect pins 33 and 36 with a resistor. While that does work, Jagaer's way is the official HP common slot PSU way. This works for any HP common slot supply. 1a. If you want to add a power switch, it should be between pins 33 and ground, and must have a debounce circuit. If you don't know what switch debounce is, skip the switch and solder a wire to ground out pin 33. 2. If you have never micro soldered before, like me, prepare to spend a few hours learning. FFS this was the most frustrating thing. I had to watch youtube videos and get a large magnifying glass before I finally successfully soldered the resistors for OVP and Voltage. 3. When you pick a supply to mod, find youtube videos or instructions that explain how to do the OVP and Voltage mods on the EXACT model number. There are at least 5 different HP 1200 watt models out there and each require a different approach. This applies ONLY to the HSTNS-PL11 4. For the OVP mod, I used a 10k resistor. The lower the resistor value the higher the OVP will go. I chose 10k since I don't plan to run this supply much higher than 14v. 5. For the Voltage mod, I used a 5.2k resistor, which resulted in a range of 13.65V to 14.85V when turning the pot under a 2A load. I set mine to 14.1V under 2A load, which increased to 14.2V under no load. It took 40A to dip to 14V, and 55A to dip to 13.92V. Now comes the fun part, need to find a good case and attach power cables. Some pics of my soldering work, as I'm pretty proud of how I did, especially on the Voltage mod. Last edited by owdi; 11-09-2023 at 02:40 PM.. |
Appreciate
0
|
11-10-2023, 04:35 PM | #33 | |
Private First Class
62
Rep 108
Posts |
Quote:
Found a simple explanation of it (a de-bounce circuit): https://www.circuitbasics.com/switch-debouncing/ "Schmitt invented this circuit way back in 1937." Dang! Thanks for sharing your work... look forward to how it turns out! |
|
Appreciate
0
|
11-10-2023, 05:35 PM | #34 |
Private
18
Rep 69
Posts
Drives: 2010 328i
Join Date: May 2012
Location: Kenmore, WA
|
If you want to dive in deeper, I found great info in this reddit post:
https://www.reddit.com/r/homelab/com...n_slot_pinout/ That reddit post points to an aftermarket PSU data sheet that is compatible with HP Common Slot, which has great info about what all the pins do, and how to properly connect them. https://power.murata.com/pub/data/po...0-12-hxxdc.pdf Unfortunately I don't see remote sense pins, which would be nice to compensate for voltage drop with smaller gauge or longer cables, so the next best thing is to keep cables short and thick. Last edited by owdi; 11-10-2023 at 05:41 PM.. |
Appreciate
1
tinkerman62.00 |
11-12-2023, 03:20 AM | #35 |
Private First Class
62
Rep 108
Posts |
I've been eye-ballin' a "Flex Slot"* type 1400W HP DPS 1400CB A (HSTNS-PD43), as it would cover the full range of BMW's recommended 70~100A capacity for a PSU during programming.
* "Flex Slot", as opposed to "Common Slot" type, have different pin out assignments and are more compact. And there is at least one demonstrated stable hack on YouTube: "HP HSTNS-PD43- Hacked- OVP- 15.5 VDC" (includes OVP mod) and a simpler(not as thorough) mod of just the output voltage for just under 14V output: "HP DPS-1400CB A flex slot SMPS PSU" (without OVP manipulation) I'm hoping to find "Output And Signal Specification" like that of the "Common Slot" PSUs indicated on the muRata datasheet, but for "Flex Slot" PSUs. Reason being that it would be nice to have a clear understanding of the main pins associated with "conventional use" for "powering ON" the PSU, like that with the "Common Slot" PSUs If anyone can share a lead, I'd appreciate it! Thanks in advance! P.S. Update: I've found some related info, albeit not as "official" as a datasheet, on rcgroups.com thread: https://www.rcgroups.com/forums/show...-Part1/page315 (post #4724) Note: 1. The pin assignments were assessed from breakout boards used for crypto mining. 2. When such "Flex Slot" type PSU is "conventionally" inserted into a host server, the "PS-ON" and "PSKILL" pins get connected to ground [or connected to circuits that would bring their voltages down below a certain threshold(e.g. below 1V) with a minimum current draw(e.g. 1~5mA to induce voltage drop on the pins)] in the host server. The shorter "PSKILL" pin is a safety feature: When a PSU is operating in parallel with another PSU and then removed from the host server, the shorter "PSKILL" pin disconnects from the host server's socket conductors before the longer main DC output contacts, quickly turning off the power supply before the DC output disconnects, preventing arcing between the DC output and the host server's input contacts. 3. Pin D11 marked "CS (25mV/A)" suggests that it is a 'Current Sense' output with 25mV/A. Therefore, a 2.5 to 1 ratio voltage divider could be used to make it 10mV/A, and a cheap 1.000V meter could be used as a current meter (as long as the decimal can be reassigned to 100.0 position, or manually manipulated). Last edited by tinkerman; 11-18-2023 at 05:27 AM.. |
Appreciate
1
eastsidaz213.50 |
11-13-2023, 11:27 AM | #36 |
Private
18
Rep 69
Posts
Drives: 2010 328i
Join Date: May 2012
Location: Kenmore, WA
|
Any particular reason you're looking at newer flex slot instead of older common slot? Per my limited recent exeprience, older supplies are much easier to work with. There is more info available online and the pins and soldering points are larger.
The 1500 watt HP Common Slot PSUs actually sell for less than the 1200 watt version since they require 200V-240V. HSTNS-PL33 is $15 on ebay while HSTNS-PL11 is $20. Here's a video showing both voltage mod and OVP mod for the HSTNS-PL33, which is rated for 125A. Also, consider the 1200 watt supplies are built to output their rated current down to 100V / 200V. With typical 115V / 230V voltage at a home wall outlet you will have some buffer. Per the below post, the 1200 watt HSTNS-PL11 surged to 150A @14.4V output for a few seconds on UK power (240V) https://www.eevblog.com/forum/projec...78/#msg3381878 |
Appreciate
0
|
11-14-2023, 10:22 AM | #37 | |
Private First Class
62
Rep 108
Posts |
Quote:
But that vid you linked by "Tao LI" hacking the HSTNS-PL33 shows an OVP hack via a firmware patch method(without the patched coding details), and in response to a question in the comments under the video(on the YouTube site), asking for a resistor based method "Tao LI" refers the inquirer to another hack video of the HSTNS-PL30/PL42* in the comments of which "Tao LI" mentions that "adding a resistor between the 14th pin of the MCU(dsPIC33FJ64GS606) and ground can increase the OVP voltage." * Refer to the YouTube video titled "HP HSTNS-PL30/PL42 PSU OVP & Output voltage increase" or rather its comments section. I've, asked in a related thread in EEVBlog for clarification: https://www.eevblog.com/forum/projec...47/#msg5169747 If there is any update, I'll certainly pass it on here. A few reasons why I noted the 1400W HSTNS-PD43: 1. The hack is so simple. 2. The hack video uploader "Hopper" mentions towards the end of the video that it "don't make any noise... some of these (other server PSUs) once you get past like 14.7 or 15 volts they start gettin' trashy in the HF bands... this one here... no noise whatsoever... very impressive..." Though he's referring to its advantage for radio use, a "noise-free" output is also preferable for sensitive ECUs as well. 3. I'm impressed that such a small unit can pump out so much current! 4. Gen9 servers have been around at least since 2014~2015(? if not before) so they ain't that new either (but perhaps at least it's less likely that they will have been passed around by multiple users). Now Gen11 servers are the 'thing', and they seem to use the same compact "Flex Slot" PSUs. I'm thinkin', going forwards it may be easier to find parts if necessary for more recent PSUs. Overall though, I'm sure the older PSUs would be fine as long as their key components, such as(but not limited to) output capacitors are "healthy". Update: In the EEVBlog thread you linked, post #15 by "eblc1388": https://www.eevblog.com/forum/projec...78/#msg3381878 It is mentioned that the HSTNS-PL11 1200W unit consumes 42.2W even without a 12V load! Not sure if this is common among all server PSUs, but I now see why "Hopper" mentioned that he prefers an AC input switch instead of a DC output switch. Last edited by tinkerman; 11-15-2023 at 11:34 AM.. |
|
Appreciate
0
|
11-16-2023, 11:49 AM | #38 |
Private First Class
62
Rep 108
Posts |
This is an update regarding the HSTNS-PL33 1500W unit discussed above. Very useful info below...! (All credit to Tao LI)
"Tao LI" had apparently uploaded a link*(below) to a site with clear photos(attached reduced image below) of the OVP access point (pin14 of the Microchip dsPIC33FJ64GS606) and the output voltage access points. "In the description section of the video (for the HSTNS-PL33 above), there's a linked image* with specific areas highlighted in red font. By paralleling a resistor to the designated points for both ground and over-voltage protection (OVP), you can fine-tune the output voltage and OVP threshold." (Hi res image link) * https://github.com/darwinbeing/HPSer...L33_MOD_HW.png NOTE: 1. The link above not only shows the high res image of the OVP & output voltage access points(attached at bottom), but also seems to have the firmware "patch" used in the videos of "Tao LI"(refer to the repository's menu on the left). Also, at the bottom of the left-side menu, there is a "README.md"* which explains the HP HSTNS-PL11/PL30/PL42 circuitry in detail with relevant schematics! (A few examples attached as screenshots below) https://github.com/darwinbeing/HPSer...main/README.md 2. The HSTNS-PL33 and HSTNS-PL30 seem to have the same basic side circuit board, and for questions regarding resistor values for OVP/output voltage "Tao LI" refers inquirers to his earlier PL30 video, in the comments section of which the following is asked and answered: Q: (@ten3mareczkos) "... Hey I want to modify my unit to work with 13.8V ... what kind of resistor is needed there in order to have 13.8V?" A: (TaoLi) "To achieve an output of 13.8 volts, you will need to connect a 1.8 kilohm(R1*) resistor in parallel. Additionally, for OVP adjustment, it is recommended to parallel an approximately 6.63 kilohm(R2*) resistor." *Note: "R1" & "R2" as the respective resistors are indicated in the relevant schematics(attached below) from the README.md linked above in which he mentions: "R1 is used to adjust the output voltage of the power supply, while R2 is used to adjust the threshold of the Over Voltage Protection (OVP). The modification mentioned above is not meant to increase the OVP threshold, but rather to trick the dsPIC33FJ64GS606 MCU into thinking that the output voltage has not been adjusted." The link below to the online WolframAlpha Solver contains the equations used to compute the R1 & R2 values by adjusting the desired output voltage "V1"*. *V1 is the very last/far right entry in the orange input cell. If V1 = 13.8V, then R1 = 1.80k, and R2 = 6.63k If V1 = 14.4V, then R1 = 1.27k, and R2 = 4.7k If V1 = 15.2V, then R1 = 910Ohm, and R2 = 3.4k If V1 = 16.0V, then R1 = 715Ohm, and R2 = 2.67k etc. "WolframAlpha Solver Output 13.8V (example) https://www.youtube.com/redirect?eve...pgG9qLzC6Dvb9T " So soldering on one end of each resistor to its respective access point and the other end of the resistor to the nearest/convenient ground soldering point would achieve the result. Tao LI apparently chose to hack with the patched firmware because the OVP access point is a difficult soldering job. But as there is some copper trace going away from pin14, it may be possible to clear some of the green solder mask along this trace at a point where soldering might be easier. The variable potentiometer used in the video is 10K(max) according to Tao LI. So it seems both the firmware and resistor solutions have all been shared in detail all along(in the GitHub repository) , just not obviously presented(in the videos). Hopefully someone tries the resistor method and reports back here. Cheers~ P.S. The Gen8 PL33(1500W) will require a bit more care in soldering resistors and the current prices for most of the units is for some reason actually higher than the newer Gen9 PD43 (1400W) which is more compact yet easier to solder resistors due to the nature of the access points, and the PD43 seems more available with more attractive prices both on Ebay and Aliexpress (considering international shipping). The original image on the GitHub site is much higher resolution for those that need it. I've reduced its size for upload below to prevent this thread's view size from getting too wide. Last edited by tinkerman; 11-18-2023 at 09:10 AM.. |
Appreciate
0
|
11-18-2023, 11:18 PM | #39 |
Private
18
Rep 69
Posts
Drives: 2010 328i
Join Date: May 2012
Location: Kenmore, WA
|
Good stuff tinkerman, you're living up to your handle :-)
I made some chunky power cables and got a second inverter to really make this PSU sweat. I also lowered the voltage from 14.2V to 14.0V for this round of tests, since I read (somewhere) that only 13.6V is needed for aux power, while 14.4V is for charging the battery. HP 1200 watt PSU - HSTNS-PL11 10k resistor for OVP, 5.2k resistor on voltage, voltage pot set to 14.0V with no load, includes cable voltage drop Code:
Current(A) Voltage(V) Input(Watts) Output(Watts) 0.0 14.00 50 0 0.9 14.00 55 13 17.5 13.89 290 243 35.7 13.75 556 491 77.5 13.42 1215 1040 95.9 12.14 1360 1164 (tested for 5 seconds) I'm going to take more measurements. One thing I noticed is voltage sagged from 119V to 114V under load at the outlet on my workbench. I have a better outlet in my garage, right where I work on my car, and right next to my breaker. I expect it will sag less, which should buy me a few more watts of output. I also want to test a few more loads between 77A - 95A to find where it dips below 13V, just need to find something to load the inverter properly. |
Appreciate
0
|
11-19-2023, 04:38 AM | #40 |
Private First Class
62
Rep 108
Posts |
Wow! I'm truly amazed by the output capacity of these server PSUs, even if it's for a moment.
Interesting that you've been load testing through an inverter... I'd been racking my brains for a while to muster up a dummy load in a relatively cheap and easy way to load test PSUs, and realized a very simple solution for a "dummy load" is to use a cheap inverter and a heatgun(or space heater): 1. Inverter 1000W~2000W(continuous output)["12V input" (10~14V or 12~24V) to 110V~120V/220V~240V output] There are some as low as $20~$30 on some online offerings! (Capacity based on PSU capacity - check whether the watt rating indicated is for "continuous" or "peak" output as there's a tendency for marketing info to show the peak power rating, and the actual continuous power rating is half the peak rating.) and 2. Heatgun with variable output (or a small space heater) (e.g. Stanley STXH2000 2000-Watt Variable Speed Heat Gun - it's about $69 , there are cheaper ones, but I have this just sittin' around from DIY tinting. It has a variable heat control and a slide switch for 2 blower speeds, so one can make gradual fine adjustments as well as jump between 2 load settings instantly. Dang!) Heatgun's like this with a flat rear end, can stand upright on a bench top pointing upwards during operation, to free your hands for other tasks. With these hooked up to the PSU, and a clip-on amp meter on the hot(+VDC) cable, the amp(x voltage) based power(Watt) can be monitored in a relatively safe, controlled, gradually(or abruptly) variable manner (as opposed to limited control options in a crude/exposed coiled wire or water cooled copper tubing dummy load that involve touching the load with a live lead at various points on the dummy load, which can cause arcing, smoke, etc. as seen in Tao LI's videos). Hope this helps someone. I'd love to see someone try this and post their experience! Cheers~ Last edited by tinkerman; 11-19-2023 at 02:41 PM.. |
Appreciate
0
|
11-19-2023, 05:07 PM | #41 |
Private
18
Rep 69
Posts
Drives: 2010 328i
Join Date: May 2012
Location: Kenmore, WA
|
I'd caution you against using that heat gun for the load on the inverter, it's almost certainly going to use PWM with variable duty cycle, like a modern dimmer switch. It will switch between full ON and full OFF, that may lead to strange readings.
|
Appreciate
0
|
11-19-2023, 05:24 PM | #42 |
Private
18
Rep 69
Posts
Drives: 2010 328i
Join Date: May 2012
Location: Kenmore, WA
|
I had some fun this morning performing more tests, this time while connected to a power outlet right next to my circuit breaker. I measured 120.4VAC at the wall with no load, and about 119VAC under load! Much less voltage drop this time. With this higher voltage, I got 85A before things went south.
To load the PSU, I used a combination of DC to AC inverters with blow dryers for a base load, and an old class AB car audio amplifier with tone generator to tune the load. It was not pretty, my wife gave me "the look", and asked me not to burn the house down. Haha. This time I measured both voltage at the PSU and voltage at the cable clamp ends. My cables use solid copper terminals, 4awg solid copper wire, and solid copper battery clamps. Cable voltage drop was around .2V at 90A. I'd probably have to solder everything to get it any lower. |
Appreciate
0
|
11-19-2023, 08:18 PM | #43 |
Private First Class
62
Rep 108
Posts |
Very cool...!
Thanks for the cautionary heads up. The Stanley heat gun apparently has a ST BTA12 600B "Snubberless" Triac (for ON/OFF function applications such as static relays, heating regulation or induction motor starting circuits). I'll have to do some reading of its datasheet and an 'application note' on 'Snubberles and logic level TRIAC behavior at turn-off', to get a better understanding (though on quick glance of the 'current and voltage waveforms for resistive loads' I see the jaggedness of the alternating voltage and current waveforms(attached below) as opposed to a smooth sine wave form, due to the cyclic dips during the OFF stages). I realize that if my main intention were to test a PSU with an oscilloscope for noise characteristics, such jagged waveforms in the AC load make such load hardly a good choice, but for testing mainly the high current output capability over an extended time, wouldn't it suffice? If not, what particular issue(s) would be reason to not use a heat gun for load testing at all? Also, how is the blow dryer you used in your load test different from such triac based heat regulating heat gun? Is it just a pure resistive load without any switching? Thanks. Last edited by tinkerman; 11-19-2023 at 09:17 PM.. |
Appreciate
0
|
11-19-2023, 10:25 PM | #44 |
Private
18
Rep 69
Posts
Drives: 2010 328i
Join Date: May 2012
Location: Kenmore, WA
|
Yes, a blow dryer has no switching, it's 90% resistive heating and 10% reactive AC motor.
I'm no expert, so take this with a large grain of salt. Consider the server PSU takes AC and converts to DC, then the inverter converts the DC back to AC. If you make the inverter drive a switching power supply I think you'll have a harder time finding the current limits of the server PSU. Perhaps if you have a very nice sine wave inverter it will behave better. I'm making assumptions here. I don't have an oscilloscope to see what current is actually doing, just the clamp meter. |
Appreciate
0
|
Bookmarks |
|
|