82v 2a-3a power supply using tl783

Thread Starter

auraslip

Joined Oct 17, 2011
12
I'm looking to make a cc-cv power supply. A very simple one at that! Basically I just want an easy way to get 80v-100v. The tl783 is a linear regulator and can handle that. The data sheet even shows how to use it with some power transistors for higher power. The following is something I made in ltspice, and is copied directly from the datasheet.


I'm looking for reliability at 2a-3a of power. Hopefully it won't be a huge package either. Can anyone advise me on my drawing if it looks good? thanks!
 

SgtWookie

Joined Jul 17, 2007
22,210
Q1 and Q2 must be PNP transistors, not NPN.
With 110V in, 80v out, and a 3A load, Q2 will dissipate 2.7 Watts, the regulator ~9.2 Watts, and Q3 about 70 Watts.

Do you have a plan for getting rid of all that heat? You'll need a big heat sink for Q3, and even then it'll be hard to keep it from melting.

Datasheet for the BUH100:
http://www.onsemi.com/pub_link/Collateral/BUH100-D.PDF
THermal resistance junction to case is 1.25°/Watt, so that's 87.5 degrees above the temperature of the heat sink.

If the temperature of the case can be maintained at 25°C, then the maximum power dissipation of the transistor is 100W; but you derate 0.8W/°C above 25°C

When you are using linear regulation, the only way to reduce the voltage is by burning off the excess power as waste heat. If you live in a cold location, that's not all bad.
 

Attachments

Last edited:

Thread Starter

auraslip

Joined Oct 17, 2011
12
Thanks for the thought out reply. I'm coming to terms with this, despite spending a fair bit of time on this project. It was fun, but I'm afraid you're right.



Here is a current limited, voltage adjustable power supply I made in LTspice. It seems to work.

However, I understand that even if I take care of the heat problem, I would still need a massive transformer. And since this is supposed to be a mobile power supply, I'm not sure I an handle that. And large transformers are expensive as hell. I don't know what to do next to buying a commercial SMPS unit from china of unknown quality :(
 

Thread Starter

auraslip

Joined Oct 17, 2011
12
A 1:1 transformer and rectifier.

Actually, here is an interesting idea; what if I just rectified the AC line, regulated it down to 83v, and THEN used a 1:1 isolation transformer and a rectifier on the output? This way I could use a much smaller transformer but still retain the safety benefits of a having a transformer.
 

SgtWookie

Joined Jul 17, 2007
22,210
The transformer really should come first.

There are ways to use TRIACs (a type of thyristor) as a pre-regulator to get the voltage on the filter cap within range of a linear regulator; I don't happen to have a schematic for one at the voltage you're talking about.

But that's something to look at.
 

#12

Joined Nov 30, 2010
18,160
That schematic in post #3 is a disaster waiting to happen. 2N3055 transistors can not be trusted to share the load like that. You need most of an ohm in each emitter circuit to make them behave properly because their Vbe will not match and the best transistor will go into thermal runaway as it heats up and becomes even more conductive.
 

colinb

Joined Jun 15, 2011
351
A 1:1 transformer and rectifier.
This won't give you 110 V dc, but will give you pulsed dc with a peak value of 170 V. You'll also need some big capacitors to filter the pulsed dc voltage if you want clean output.
Actually, here is an interesting idea; what if I just rectified the AC line, regulated it down to 83v, and THEN used a 1:1 isolation transformer and a rectifier on the output? This way I could use a much smaller transformer but still retain the safety benefits of a having a transformer.
OK, so you rectify the ac to dc and regulate it. But then you can't use an isolation transformer on that dc.
 

Thread Starter

auraslip

Joined Oct 17, 2011
12
My last post is pending because I included a link I guess. Anyways, thanks for the help. It should be clear I have no clue what I'm doing :D

I thought this would be as easy as making a 5v regulator, but I guess not.

The TL783 might still be an option? It seems easy. I'm just trying to wrap my head around how large a transformer I would need to supply even just 1a, let alone 2a. The issues with running multiple voltage follower transistors doesn't seem that big, nor does current limiting. I just don't have a clue where to start on the AC side. I understand the efficiency of linear regulators is Vout/Vin, so with 110vdc input I'm looking at ~30% of the power at 2a (82v * 2a * .30 = 50w) worth of heat. Which isn't a huge deal spread out over a few transistors. I'm not looking for efficiency. I just need to get the job done.
 

colinb

Joined Jun 15, 2011
351
I understand the efficiency of linear regulators is Vout/Vin, so with 110vdc input I'm looking at ~30% of the power at 2a (82v * 2a * .30 = 50w) worth of heat. Which isn't a huge deal spread out over a few transistors.
You better run the numbers. That 50 watts of heat will certainly require some massive heat sinks to keep the transistors from overheating. Possibly some fans would be helpful. You'll need to evaluate the power dissipation in the transistors, thermal resistance from junction to ambient (probably including a heat sink), and maximum temperature above ambient allowed for the transistors.
 

#12

Joined Nov 30, 2010
18,160
The proper way to do this is to get a transformer that will give you the right voltage after rectification. The formula for rectification is: Volts peak = (1.414 times Volts RMS) minus volts lost in the rectifiers. The most efficient rectifier is the 4 diode type and that will cost about 1.4 volts off your peak voltage.

Then you need a filter capacitor and the formula is: I = radical 2 Eripple (peak to peak) x frequency x C.

That's: Amps = .707 times the sag between power pulses times the frequency after rectification times the capacitance in farads. You seem to speak American, so the frequency will be 120 Hz.

This will give you a mostly DC voltage (with a little bit of ripple) to feed the transistors. Make sure the lowest DC voltage after diode losses and sag is about 3 volts higher than the output voltage. That will keep the transistors from running out of voltage.

Be sure to calculate the power dissipation on your zener diode. It looks a bit high from here.

edit: colinb poster while I was typing, in case it matters.
 

Thread Starter

auraslip

Joined Oct 17, 2011
12
50 watts is a lot of heat? lol, guess I'm used to dumping 3kw into my hub motor at 20% efficiency at starts :D of course it's ventilated and I monitor it with a temp sensor.

Thanks for the help ya'll. I'll figure out the math tomorrow and try to find a suitable transistor. I just put together my new ebikes battery pack (20s = 78v nominal) and need to do a test ride. Should be rolling at 40mph!
 

colinb

Joined Jun 15, 2011
351
50 watts is a lot of heat? lol, guess I'm used to dumping 3kw into my hub motor at 20% efficiency at starts :D of course it's ventilated and I monitor it with a temp sensor.
It's all relative... but think about an Intel Core 2 Duo desktop processor, which has a thermal design power of 65 W. This requires a fan-cooled heat sink, which I would call “massive” in the context of common transistor circuits... not to mention the CPU has a big, flat surface area for thermal path to the heat sink.
 

colinb

Joined Jun 15, 2011
351
Variable linear power supplies are MASSIVE. Some of the bench supplies we have around here weigh about 40 pounds. For efficiency I think they usually have multiple “ranges” with different transformer ratios from mains power so that if you want 30 V, you can have it, and if you want 1.5 V, you aren't burning 50 watts of heat to do so.

In your case, you could implement a switch-mode converter instead of linear regulation. It would certainly be MUCH smaller and more efficient, and possibly cheaper. The main disadvantage would probably be in electrical noise on the output due to the switching converter. However, with adequate filtering and/or inserting a low-dropout linear regulator after the switching converter to further reduce noise output I think you could get good results. But I've never made something like this so I may be wrong.
 

Thread Starter

auraslip

Joined Oct 17, 2011
12
I thought ya'll might like to see my @ home charger


I got 6 12v50a server power supplies for $5 each shipped. On the top I'm using a meanwell switching supply to limit current and fine tune the output. Here is the build thread.

Others use the variable lab supplies mention to for voltage and current control. This setup is overkill for my 800Wh pack, but I wanted something that'd be easy to scale up for whatever EV project I build.

Anyways, I was quite happy with the performance I was getting at 20s. I wanted wheelies, and only kinda got them :D so I'm going to be running 22s now which should have a final charge voltage of 22s * 4.15v = 91v. Still not sure how I'm going to achieve this. Too busy with school right now to do anything complicated :/
 
Top