| Electronics > Projects, Designs, and Technical Stuff |
| Lab Power Supply - The Lost Current |
| << < (8/26) > >> |
| C:
--- Quote from: radoczi94 on February 05, 2018, 11:17:29 pm ---I really wanted to build an indestructible, foolproof power supply, that is why I used that massive transformer and heatsink. With the case I built, it's weight is around 20 kgs, and not even used the full 5 amps of the transformer. --- End quote --- No such thing exists. All have limits and is up to user to keep connected load in limits. Example a 30v power supply and idiot connects a 48v battery to output. For power supply to protect against this is very hard to do, best to think of limiting the damage. Example 1000F cap bank connected to output. The power supply would be at current limit for days. How long can power supply survive with output at 0V & Max current. Sure you can add to this list. So the best you can do is the limit damage and user has to supply remaining protection. One thing I would not pass on is using what you have to learn. A lot could be learned. Back in the 80's HP was doing a lot of internal research. A lot of things HP sold were first designed,built & used to cure a research problem. Often a power supply consisted of many modules or boards. This makes upgrades or new designs easer. So before you look for something better, would be good to know what to look for. The separate parts have to work together to make the whole thing good. You can have very good blocks in schematic, but they have to be connected together in a way that makes total good. You also have to keep in mind that nothing is perfect. You have a cap across output terminals. For CC mode you want no cap or small cap. The reason is that cap charge gets dumped into a short on output. For CC mode any cap on output slows rate of current change to hold desired current. For VC mode to hold constant output you want that cap huge. For VC mode to detect a load change on output, smaller lets circuit see same change sooner. The right cap or caps can stop supply from oscillations. There are more things So output cap is war of sizes with an option of making CC or VC better. Then it could be that expected loads will survive the dump of output cap on short such that a little larger cap is ok. Or load is not effected badly with a voltage change. For a lab power supply you need to think of all the nasty loads that could be connected. Step one for lab power supply is to protect connected load. If connected load is real bad then step two is for lab power supply to try to protect it's self. Then you have option of big load power supply & small load power supply where you pick best for load So at cost of some of your time you could learn a lot from what you have. Keep in mind the many ways to do things. One change from last schematic to first is power and reference. Last schematic used op amp's power supply rejection to lower noise. First schematic I see creating problems by coupling noise into control power. Again bad schematic hides details. Take last schematic and redraw it. Power section on left is D1-D4, R1 & C1 on Right is current sensor R7 Output control Q2,Q4 & R16 ? C3 Some protection by D10, D11 Output filter C7 What is left is Control circuit. Now if you do not have ability to easily do this, use your versions. Now the Control circuit needs inputs to function. Voltage and Current inputs should be differential. Voltage sense is connection to Circle 3 and Circle 4 Current sense is connection to each side of R7 For control circuit to control output you could need connections to circle 7, circle 3, circle 4, left side of R7 Note that control and sense should be separate wires or paths, A Kelvin connection. Time to learn Does current PC board do this separation? Does your first do this? It's not shown on any schematic posted, yet is simple and cheap to do. If you look for new design, something to look for. Note also this separates the high heat parts, parts that fail on heat. With the schematic redone, You should be able to see some problem areas and have some ideas on how to improve circuit So a great schematic is very important. Makes it easy to understand the circuit quickly. Can show problem areas & and many more things. Now for control section separate out it's parts. Reference section U1, D8, R4,R5,f6 it;s outputs are connection to R18 & circle 6 Should note a few things here Power is via U1 V+ & V- Part of differential sense is by R4 & R6 being connected to circle 4. Should note that V+ and current set pot is main source current changes. Should note that last schematic makes great use of op amp's power supply rejection to lower noise. A change to less is not an improvement. The V+ connection of U1,U2 & U3 is connection to circle 7 for output control. Adding a second line across to circle 7 could show this. So I see a good first step is to learn from what you have now & fix schematic to show details and make understanding easer. Take a huge number of measurements with your meter with different settings & loads. Save this to look back to to see a change is good or bad. Each change a new spreadsheet of measurements. Unknown is how well the last schematic functioned before changes. Could be original has problems or you created problem for CC mode. Only measuring will tell. One common thing is to have separate power for control & load supply. So you have a massive transformer, would be a shame not to use it if you see a need for the power supply it can be a part of. The way I see a single PCB is a way to mass produce something. If you want something good that can upgrade then many PCB's or circuits is better. A good circuit design also makes this easy. If you look at old HP supplies you often see it is made of many modules. A module is often used in many different supplies. Think of future some, today you build best power supply you can. Next year you could think of it as junk. Easy to change junk could become new supply better then first. |
| radoczi94:
--- Quote ---No such thing exists. All have limits and is up to user to keep connected load in limits. Example a 30v power supply and idiot connects a 48v battery to output. For power supply to protect against this is very hard to do, best to think of limiting the damage. Example 1000F cap bank connected to output. The power supply would be at current limit for days. How long can power supply survive with output at 0V & Max current. Sure you can add to this list. So the best you can do is the limit damage and user has to supply remaining protection. --- End quote --- Yes, I know that, i did not really meant it as it is, just basic protection and really good cooling. I have some desperate ideas, like having a huge parallel diode reverse biased and a fuse on the output in series (may be fast blow), it could prevent a reverse polarity battery for example. A huge truck battery will kill the whole thing anyway, so not much protection, but better than nothing. A series diode with proper cooling could help also, but the voltage drop can be an issue. I see now, I can not really build an ok-ish psu without external voltages. So I have to go to the store, and buy a pair of small transformers, build a good secondary supply for the control circuit. Even if I hate wiring work. I was not aware of the Kelvin-connection thing.I did not knew, that such thing exist. Saw in a bunch of EEVBlog vids, but never knew, if it is intended or not. Therefore my design has absolutely noting like that. Thank you for that! Otherwise my pcb design is a bit beta-ish, tried to maintain wide traces for the 3A current paths, so the contol gone into a tiny little space between huge polygons.Never again. I think I will do some further investigation on what is happening around the circuit, to see if I can make it work, but I don't really feel the motivation to stick to this schematic and make it an acceptable psu. My buddy sent me an old schematic he built two of this. It has separate power supplies, opamps controlling trough an OR gate etc. A lot of things I seen in older Tektronix schematics. The schematic itself is a perfect example of how not to draw a schematic. My only concern is the old parts used. If it works, I can replace the 741s with better opamps, like OP07, which has excellent DC characteristics ( the BW is a bit low, so certainly not the fastest opamp on the market). The other thing is the 723. It can be used as a switching regulator so it's not hard to make it oscillate. https://drive.google.com/file/d/0B7yeYY6jyIVVaHZpTlhTQVl2b0wzVHpEUXAtS1VtU1E0b2o0/view?usp=sharing Here it is.I kinda like it, have not found any better yet. The description is pretty good too, explains the principle. |
| C:
In my past I do not remember any Tektronix supplies. Used a lot of HP and a few other brands. If you spend some time looking at old, look at HP power supplies Protection As a last resort protection the simple thing is have control circuit blow a fuse. On input side you have many choices. Blow line fuse, blow a fuse on secondary On output side you want to get disconnected from load. Think the best device is the SCR or triac to do this. --- Quote ---I see now, I can not really build an ok-ish psu without external voltages. So I have to go to the store, and buy a pair of small transformers, build a good secondary supply for the control circuit. Even if I hate wiring work. --- End quote --- You can use any type of supply to supply control circuit power. If you have any electronics junk you have something you could probably use. The important part is the separation. Think of this, you have a possible large current draw on input bridge diodes. Current = voltage drop, changing current = noise A simple change is just using a second diode bridge and filter. You gain some isolation from the noise. Separate source is more noise isolation. Actual power voltages is not that critical. Some designs are easer if control power exists below or above output terminals of supply. Important thing here is op anps have a sweet spot in center between supply with problems at close to supply rails. Kelvin-connection You may have seen some current sense resistors with a Kelvin-connection. For a lab power supply you want a Kelvin-connection to output terminals for voltage sense. Better supplies add remote voltage sense terminals to extend Kelvin-connection to load. Can be important to use on a PCB as well. Think of When this changes where do the currents go/flow. Think of this, a voltage based temp sensor and a current based temp sensor. On end of a 1000 feet of wire one is crap with other still great. --- Quote ---Otherwise my pcb design is a bit beta-ish, tried to maintain wide traces for the 3A current paths, so the contol gone into a tiny little space between huge polygons.Never again. --- End quote --- Guess why you see separate boards for high current or one small area for high current. --- Quote ---I think I will do some further investigation on what is happening around the circuit, to see if I can make it work, but I don't really feel the motivation to stick to this schematic and make it an acceptable psu. --- End quote --- Investigate to learn. If you chop up the schematic like I suggest, then future is replacing blocks on schematic. The 723 is ok for a cheap simple power supply but is not great. The link you post from my quick read is not great. Keep in mind that you want the lesser value of Voltage or Current. Link Instead of simple you have a mess. In link voltage control A1 if not modified wants to turn on output more in current limit mode while current control needs to turn it off more. Any change in A1 output messes up good current control! Add a pull up resistor to output transistor causing output to turn on full blast. A2 is all ready, Too much current and it pulls down lowering current to set point. A1 needs a diode and then will work same way and reduce to set point. Just want you want for a supply that has CC & VC Lower output level win. Now you have two separate circuits controlling output. You do not have to modify one for other to work properly. Now look at range for the change. A diode has a voltage vs current curve. The curve is fairly flat but not flat. Change over goes from one output control, a current to the two outputs equal , 50% of one controlling. If it takes 20 ma to hold down output then you would look at two spots on diode curve, 20 ma & 10 ma. and get some voltage value. Now due to feed back this becomes mostly hidden and change from CC to VC is fast with few if any problems. Want something great just read a HP manual for a power supply. Every thing is in it, how it works, schematic, parts list and often PC Board layout. Gave you a hint above but you may have not caught it. Current goes the distance, You can connect low side circuit to high side circuit with current. A current change then can control output. |
| xavier60:
Because of the way the voltage adjust potentiometer is configured as a VR in the Elektor design, it is not fail safe. If the potentiometer becomes scratchy, the PSU output could spike to full voltage. I think it is better to use the potentiometer as a potentiometer to directly vary the reference voltage. Including a wiper bleed resistor would make it fail safe. The reference voltage will need to be negative. I don't like the remote sensing. It should be omitted. |
| radoczi94:
One thing I do not really understand about this old schematic. 723 makes the reference. As far as I know emitter followers have a voltage gain of 1. How is 0-30 volts on the output, if the Vref is 7,15V? I'm shure, that some resistor makes this magic. Maybe R17 is talking back? Why don't just pull it up to the collectors? I've seen in some HP and Tektronix supplies, the base were just pulled up. What I do not like about this, is the 723. Maybe a little bit too crusty. I'm shure there are much better solutions for a good reference. --- Quote from: xavier60 on February 07, 2018, 12:12:18 am ---Because of the way the voltage adjust potentiometer is configured as a VR in the Elektor design, it is not fail safe. If the potentiometer becomes scratchy, the PSU output could spike to full voltage. I think it is better to use the potentiometer as a potentiometer to directly vary the reference voltage. Including a wiper bleed resistor would make it fail safe. The reference voltage will need to be negative. I don't like the remote sensing. It should be omitted. --- End quote --- You mean, like the current set pot? Dividing the vref with the pot itself? |
| Navigation |
| Message Index |
| Next page |
| Previous page |