Linear power supply project

[Therm Mr.]

A 2x14 V transformer with the center tap at ground would give about +-18 V at the filter capacitors. With a small transformer and low load the voltage can be some 20 % higher. So about OK (with plenty of reserve for a lower mains) for the +-12 V regulators.

For T1 one one would have to choice of something like 24 V / 2 A or 12 V up to about 3.2 A  DC (assuming a transformer rated for some 3.3 A AC).

Thanks for clarifying, I understand it better now. andrewtaylor and his original advice was actually pretty good. Now with your help I fully understand why it was good advice.

Thanks

Therm

[Therm Mr.]

I had several questions I was going to ask and decided to re-read the whole thread from the beginning first. All my questions had already been answered, I just did not absorb it. Some were answered twice.   I don't know if it were info overload or just reading to fast. This thread is a great reference for me. By re-reading I did come up with a new question though see the quoted post below.

If you will be testing the design on a breadboard, consider connecting R6 to the ORing node instead of directly to the output of the CV opamp. This idea suggested to me by Kleinstein greatly reduced the voltage overshoot in my PSU design whenever it transitioned from CC to CV mode.

What does this mean? What is the ORing node xavier60 is referring to?

I received a box of components from Mouser yesterday (Sat). I have another due from DigiKey tomorrow (Mon). I will be able to begin mocking up the circuit on the breadboard. I am going to take pqass's advice and work in sections getting each section working before moving on to the next. Rather than just building the whole thing from the start. I ordered the display posted by rolycat. It should be here late this week or early next week. I chose that one because it had both amps and volts. We shall see if it can work, it is cheap enough to experiment with.

I am still pretty much dedicated to the original HP design. Although some changes seem likely for the display to work properly. As this project has progressed I have become a bit more open to some design changes based on testing. But as of now I plan on sticking to the plan.

Thanks to all for your help so far. with a special mention to pqass and Kleinstein.

Therm

[xavier60]

During initial testing, the values of R6 and C6 might need to be altered if the CV regulation shows signs of instability.
Another test involves momentarily overloading the PSU's  output while monitoring the output for voltage overshoot during the CC to CV transition. At this stage, connection of R6 can be moved from the CV opamp's output to the anodes of CR4 and CR5 to see if there is a worthwhile reduction in overshoot.

[Therm Mr.]

Thanks @xavier60

I have another (maybe dumb) question about mounting the heat sinks. 

I thought in the past I just mounted V regulators with thermal paste to the heat sink. But a TO-3 transistor I remembered had a mica insulator sheet.  I noticed at the suppliers they have mounting kits with either mica or thermal pads for the TO-220 and some others.

Do the Darlingtons (TIP142) need an insulator or is thermal paste sufficient? Also whats needed for 7805, 7812, 7912 V regs?

I will be mounting heat sinks on the diode bridges also, anything needed with them?

My DigiKey order arrived early (yesterday) so I am ready to start. I want to be sure I don't get ahead of myself and make stupid mistakes.

Thanks

Therm

[Kleinstein]

With TO3 and essentially all TO220 transistors the case metal is connected to the collector of the transistor. If one needs isolation depends. Thermal past alone is not a vaild isolation.

For the 7812/7912/7805 it depends on the power loss if and how much heat sink is needed. With a low consumption for the regulator part there is a chance to get away without. One could still reserve space for a small heat sink. I would consider some 200 mW OK for a 78xx without a heat sink, so maybe 10 V and 20 mA. Chances are the LED display would need more power.
The diode bridges for the auxiliary power likely can get away without a heat sink.

[Therm Mr.]

With TO3 and essentially all TO220 transistors the case metal is connected to the collector of the transistor. If one needs isolation depends. Thermal past alone is not a vaild isolation.

For the 7812/7912/7805 it depends on the power loss if and how much heat sink is needed. With a low consumption for the regulator part there is a chance to get away without. One could still reserve space for a small heat sink. I would consider some 200 mW OK for a 78xx without a heat sink, so maybe 10 V and 20 mA. Chances are the LED display would need more power.
The diode bridges for the auxiliary power likely can get away without a heat sink.

The TIP142 is actually a TO-247 I mis-quoted that. But I would assume the same is true for that package. What would the need for isolation depend on? Would providing such isolation as a precaution have any negative effects if it were not needed?

I am going to provide heat sinks for the regulators if for no other reason that the pictures I have seen inside of the power supply have all had them.

Thanks

Therm.

[CaptainBucko]

Insulators add thermal resistance, which means the junction temperature of the transistor (Tj) will be hotter. Hotter means less reliability, or lower current supply capability for the same level of reliability.

In linear power supplies, you try and avoid the need for insulators where you can, but sometimes you can't, or you are not using the transistor near its limits, so the extra thermal resistance does not matter.

In high current linear power supplies, this can be a problem, and you can actually regulate the negative rail, so the pass transistor collector is at GND potential. This is what I did for my 50 Amp 13.8v supply.

[Wallace Gasiewicz]

Listen to Kleinstein     

As far as I know almost all transistor packages have the collector or drain connected to the tab or the metal on the back of the package. This needs to be electrically isolated from the Gnd.  I think your are building an Isolated PS with pos and neg terminals.   These are not referenced to Gnd (until you want them to be).  I think Mica is the best along with heat sink compound. Silicone pad is also used. The Tab or back of the package needs to be thermally attached to the heat sink which is almost always grounded. It needs to be electrically isolated from the Gnd. The metal screw that attaches the transistor to the heat sink is also electrically isolated, sometimes this screw is a non conductive Nylon screw.  Heat sink compound helps the thermal conduction by establishing a better thermal connection, filling in most mechanical deformities. It will not electrically isolate the transistor by itself,     
There are a few RF transistors that I am aware of that the Emitter is attached to the Tab, this is unusual and there is no electrical isolation used with these transistors, but heat sink compound is used without the Mica. I am not aware of any Power Transistors that are configured this way.     
This goes for ALL flat packages and also the metal can of the TO3 transistors (and other can transistor size)   
 
Sometimes when the transistor or volt regulator IC does not handle much power the transistor is left "flying" or unattached to heat sink.  Not very often though.     
I think it is safer to heat sink all the devices that are designed for heat sinking,  Even if they can work OK without sinking, they will have better heat stability with heat sinking.  There are little metal heat sinks that can just be attached to the back of flat packages that are not grounded and can help a lot also.   Heat or rather temp change is a big enemy of stability is any PS.

[dietert1]

Small heat sinks don't really take more heat away than the device pins, unless they are in thermal contact with the board. Some of those small heat sinks are meant to be soldered into the board. One can also put larger copper pads next to the power transistor to help spreading the heat into the board. Sometimes those pads are kept free in the solder mask in order to cover them with solder, once more reducing heat resistance.
Another method is mounting power transistors flat on the board and soldering them onto a cooling pad like SMD parts.

Regards, Dieter

[Therm Mr.]

This is a pic of the inside of a 361X I scrounged off of the net. The voltage regulators seem to have oversized heat sinks soldered in. The Darlingtons are the TO-3's mounted to the rear cast alum heat sink. When looking close they all seem to have mica insulation, but it is hard to tell. The TO-3's are obsolete and I am using a TO-247 package. It would be difficult to try replicate the large alum casting so I am experimenting with using two fan cooled large heat sinks with one TIP142 Transistor mounted on each one. They mount back to back and share a cooling fan.

enut11 posted a very simple but nice thermistor controlled on off fan switch but I worry about temp swing using an on/off type. I think I found a relatively simple variable fan circuit that is thermistor based and may help to keep a stable temp. I don't know till I try. Or I may just go with full time fan on and be done with it. I wanted to build something difficult and this definitely fits the bill so far.

I appreciate any and all advice, thanks to all.

Therm

[Wallace Gasiewicz]

You can probably mount each of the TO 247 using the screw holes of the TO 3 transistors. Most of the back of the TO 247 should be in contact with the heat sink using the insulating mica and compound. You can probably use the existing screw and nut you have since the 247 does not require insulating the screw.

[enut11]

enut11 posted a very simple but nice thermistor controlled on off fan switch but I worry about temp swing using an on/off type. I think I found a relatively simple variable fan circuit that is thermistor based and may help to keep a stable temp. I don't know till I try. Or I may just go with full time fan on and be done with it. I wanted to build something difficult and this definitely fits the bill so far.

Therm

Hi Therm
That 2 transistor fan controller from Reply #76 is a variable fan speed control. Ie, the hotter the heat-sink the faster the fan will spin. As the temperature rises the thermistor resistance drops which allows more base drive current for the transistors. The pot is just for adjusting the threshhold temperature. This could be around 35-40C, but your choice.
enut11

[Therm Mr.]

Hi Therm
That 2 transistor fan controller from Reply #76 is a variable fan speed control. Ie, the hotter the heat-sink the faster the fan will spin. As the temperature rises the thermistor resistance drops which allows more base drive current for the transistors. The pot is just for adjusting the threshhold temperature. This could be around 35-40C, but your choice.
enut11

I did not catch that. I will give it a try it is a very nice simple circuit.

Thanks

Therm

[Therm Mr.]

This came in today and gave me a good chuckle. I had no idea how small it was. But rolycat was correct about being it being pretty accurate. I hooked it up to my ham radio 12v power supply and it read 14.02 v. I hooked my Fluke 87 to the same supply and it said 14.01 V. Pretty close for a China cheapie.

I have not started mock up yet because the heatsink for the Darlingtons did not come with the cam clips needed for attaching them. So I had to order some and now I'm just waiting again. I did get to attach the fan and test that it works so I guess that is progress. I plan to monitor the case temp with my infrared thermometer to see if it will be able to reject enough heat. We shall see.

[xavier60]

With those panel meters, I have never been able to make use of the 4 retaining tabs because they never want to spring in for me. I have been breaking them off and relying on a neat cutout in the front panel.
I'm curious about how the display refresh compares to the advertised specs.

[pqass]

Hi Therm,

Did any documentation come with the meter?
It would be helpful if you can provide a closeup of the backside.

[Therm Mr.]

Hi Therm,

Did any documentation come with the meter?
It would be helpful if you can provide a closeup of the backside.

Absolutely zero documentation. The only reason I knew how to hook it up was the markings on the board and rolycats post. I can't figure what that push button in the center does? It is tiny, but since I have it I may as well experiment with it. Pic attached. Best I could do.

Thanks

Therm

[pqass]

The two issues I identified with these cheap meters in my reply#36 were their burden voltage, and (on some meters) that you couldn't dial-down your PS below a minimum voltage (4.5V) without starving the meter for power.  Fortunately, the burden voltage from the meters built-in (low-side) shunt can be removed by attaching the wiper and lower fixed contact of (10T) pot R37 to point B (vs. A) in the attached diagram (from rolycats link).  And by providing the meter with its own FULLY ISOLATED supply, it will happily read 0V with R37 fully CCW.   

Having two shunts in the circuit does introduce another voltage drop (probably by a few 100mAs) between unregulated to regulated sides, however, I think the simplicity of the meter installation makes up for that; ie. NOT having to configure the meter to use your existing shunt (R2), choosing a full-scale Vref, calculating dividers, and re-calibrating.  And the bonus of eliminating R11, R12, R18, and R17.  UNFORTUNATELY, by using this meter (with built-in shunt) you won't be able to set/view your current setting by pushing momentary SW1A.  Instead, you'll have to short the outputs while adjusting R19.

These cheap Chinese meters are not based on the custom logic ICL7107 ADC but instead on a cheap ($1) STM8 MCU with 10bit ADC.  It probably uses oversampling to get 4 extra bits to achieve 10,000 counts (at 2 Hz refresh).

[Therm Mr.]

The two issues I identified with these cheap meters in my reply#36 were their burden voltage, and (on some meters) that you couldn't dial-down your PS below a minimum voltage (4.5V) without starving the meter for power.  Fortunately, the burden voltage from the meters built-in (low-side) shunt can be removed by attaching the wiper and lower fixed contact of (10T) pot R37 to point B (vs. A) in the attached diagram (from rolycats link).  And by providing the meter with its own FULLY ISOLATED supply, it will happily read 0V with R37 fully CCW.   

Having two shunts in the circuit does introduce another voltage drop (probably by a few 100mAs) between unregulated to regulated sides, however, I think the simplicity of the meter installation makes up for that; ie. NOT having to configure the meter to use your existing shunt (R2), choosing a full-scale Vref, calculating dividers, and re-calibrating.  And the bonus of eliminating R11, R12, R18, and R17.  UNFORTUNATELY, by using this meter (with built-in shunt) you won't be able to set/view your current setting by pushing momentary SW1A.  Instead, you'll have to short the outputs while adjusting R19.

These cheap Chinese meters are not based on the custom logic ICL7107 ADC but instead on a cheap ($1) STM8 MCU with 10bit ADC.  It probably uses oversampling to get 4 extra bits to achieve 10,000 counts (at 2 Hz refresh).

After seeing the size I am not likely to use this meter for my project. It will be ok to experiment with a bit but it's just too small for my liking. After hearing your assessment of it and now knowing the CC switch wont work. Well that kind of seals the deal for me or the no deal anyway. It is an interesting little meter though. I bought it just to have something to use till I made a final decision. But did so with an open mind in case I really liked it.

I tested it again against the Fluke and they are both the same. I just needed to be sure the little test clips were seated properly. Here are Pics of dual power supply wiring style, both Amps and Volts. The load is the Heatsink fan. I am amazed at the accuracy of this cheap little meter. The amps fluctuated but both meters stayed together.

Thanks

Therm