The Power Supply Modding Guide (or TL494 -> Anything)

[Th3_uN1Qu3]

This topic is a WIP, and for now it will mostly be a placeholder for things to come. Here i will try to cover the basics of power supply modding, ranging from the simplest to the extreme. I will also present the TL494 in detail.

First we will start with Tools of the Trade. This is a bare minimum of what you need to survive as a PSU modder. You will need:

• At least one multimeter with frequency and duty cycle measurement. Two are better.
• Another two cheap meters, but make sure one of them can measure AC current. So you can measure input voltage, input current, output voltage, output current at the same time and determine efficiency. Working only with line powered PSUs? Then you can substitute a Kill-A-Watt or similar for two of the meters. Still, having a bunch of meters handy is always good. I have six.
• Oscilloscope, 10MHz at the very least. A x10 probe with compensation is also mandatory.
• Two soldering irons. One low power ~25W for general work, the other 60W or more, chisel tip. For removing big things like transformers and inductors.
• Desoldering braid and desoldering pump.
• Enameled copper wire for making your own transformers and inductors.
• The Mega SMPS Design Cheat Sheet
• Screwdrivers, screws, boxes, things like that.

I will not bore you with basic theory and i assume you have some knowledge of that. Instead, we will explore hands-on how to modify a power supply and what to look for, and whether what you want is achievable with your particular supply.

A lot of us want to turn old PC PSUs into proper all-round bench supplies. I will show you how to achieve that, even if the PSU uses a custom controller. I have several "universal" TL494 daughterboard PCB designs which i will make available, which you can use to run *any* PWM-based power supply.

Questions? Feel free to ask.

[everell]

Re: The Power Supply Modding Guide (or TL494 -> Anything)

Another tool I find quite handy is a Dremel drill and a number 60 drill bit. Seems like I am always needing to drill another hole so that I can add something. The number 60 drill bit is the smallest I have found at the hardware store, and is small enough to add a hole without demolishing several circuit traces at once.

[newtoCaps101]

Re: The Power Supply Modding Guide (or TL494 -> Anything)

Thank You uN1Qu3! I will keep an eye on this thread, I have plenty of power supply's laying around. Some broken, some dont work, Others I used up for parts

[TheLaw]

Re: The Power Supply Modding Guide (or TL494 -> Anything)

Unique you're a boss. Thanks for sharing all of this.

Your theme-song: http://www.youtube.com/watch?v=nW4Acfqzpz4

[Th3_uN1Qu3]

Re: The Power Supply Modding Guide (or TL494 -> Anything)

I forgot to mention that at least one of said multimeter(s) will need to be able to measure inductance as well, if you want to wind your own transformers or inductors.

Stay tuned, coming soon... I have LOTS of stuff to do including a couple exams coming up that i failed in the summer so imma have to take them again, and my router box just crapped out (was a Celeron 566 in a Slot1 board, the CPU slot was wiggly, and when the box was moved for cleaning under that desk, it failed altogether) so currently my only source of internet is leeching wifi from my neighbors. Yay.

[Th3_uN1Qu3]

Re: The Power Supply Modding Guide (or TL494 -> Anything)

Fixed my router box a few days ago btw... Found its original PII-300 which seated snugly and fired right up. I'll have to buy it a slightly better chip but for the time being it does fine.

So, i give you the first TL494 daughterboard layout! It is the standard halfbridge mode with collector output and 5v reference voltage. Assumes that driver transistors and pull-up resistors for the TL494 are on the power supply PCB. Brings out both error amps and the deadtime control input for external protection support. There is space for 2nd order compensation on both EAs (if 3rd order is necessary, then the other RC group can simply be installed on the main power supply PCB, since the voltage divider resistors are going to be there anyway), and error amp 2 is configured with a variable reference, to be used for current limiting. You can see pics of the actual board here. Use ExpressPCB to open and print the layout for toner transfer.

Next up, the accompanying LM339 board, supporting overvoltage, undervoltage (which also acts at short circuit), and overtemperature protection! Complete with LED indicators for each.

[Th3_uN1Qu3]

Re: The Power Supply Modding Guide (or TL494 -> Anything)

Is it Christmas already? Looky what i built... a protection board! Undervoltage/short circuit, overvoltage, and two channels of overtemperature protection.

A few issues were found and fixed in the prototype, hence rev 1.1. Most notably, the LEDs have been spaced properly. I found out after i did the board that 5mm leds wouldn't fit because they were too close together... so i had to use 3mm for mine. 5mm ones will fit with the new layout.

Connect the "out" of this circuit to pin 4 (DTC) of the TL494. If the comparator detects any problem, the TL494 is shut down. Note that the circuit isn't latched: overvoltage and overtemp protections will allow the 494 to restart once the fault condition clears. This may, or may not be desirable for your use.

Use this voltage divider calculator to find out the resistor values for your needs. Output voltage must be 2.5v (as this board uses a TL431 2.5v reference), and select resistors so that you have 1-2mA of current. You do remember Ohm's Law, don't you?

For the nitpickers:

1. You'll have to mount one electrolytic cap on the bottom of the board. You need a cap over the top resistor of the undervoltage protection to create a time constant long enough for the main caps of the supply to go past the undervoltage protection threshold (that's for yourself to figure out), otherwise the protection circuit won't allow the power supply to start up at all.
2. Yes, the screw holes are too small and too close to the corners. You fix it.
3. I only installed one channel of overtemperature protection on my board, don't point that one out. If you don't need one of the protections just take out the LED. And yes, the LEDs are a working part of the circuit. If you don't want them use 1N4148 diodes instead and increase the 5k1 resistors to 10k.

[everell]

Re: The Power Supply Modding Guide (or TL494 -> Anything)

Both pc boards seem to me to be looking at them from the perspective of the top or component side of the pc board. This would mean that to etch the boards, the mirror image would have to be used. Is that correct?

Would you give a simple explanation of the difference between a half bridge psu and a full bridge psu.

(I am still sitting in the dummies section)

[Th3_uN1Qu3]

Re: The Power Supply Modding Guide (or TL494 -> Anything)

Both pc boards seem to me to be looking at them from the perspective of the top or component side of the pc board. This would mean that to etch the boards, the mirror image would have to be used. Is that correct?

That's why i gave the .pcb file. You just open it in ExpressPCB, go to File -> Print, select Bottom Layer, and it'll do the rest of the work for you.

Would you give a simple explanation of the difference between a half bridge psu and a full bridge psu.

In the half bridge, the voltage applied to the primary is Vin/2. In the full bridge, the voltage applied to the primary is Vin. Which means primary current is now half of what it would be in the half bridge. So, given the same transistor current ratings the full bridge can offer twice the power. The transformer does need double the primary turns because you have double the primary voltage - but with half the wire thickness, you end up with the same size transformer as for the half bridge.

[Th3_uN1Qu3]

Re: The Power Supply Modding Guide (or TL494 -> Anything)

Coming with a rectification of what i said earlier: The time constant needs to be on the Vcc input of the comparator, not on any of the dividers. The reason is that since the power supply is off, there's no voltage to charge that cap in the first place. And so it'll stay off.

The resistor you choose needs to supply 10-15mA. The comparator and TL494 will run on a lot less, but i decided to have a safety margin, so they're more resistant to interference.

[Th3_uN1Qu3]

Re: The Power Supply Modding Guide (or TL494 -> Anything)

Now it's time for another talk, this time we'll cover unregulated supplies.

Why unregulated? Because they're simple. There's no feedback loop to compensate, no worries about interference within the power supply itself (since the control IC is basically turned into a dumb pulse generator), and no big and expensive output inductor! There is a lot of stuff that runs on unregulated supplies with mains transformers, wouldn't it be nice if we could shrink that transformer to, let's say, 1/10th the size? The lack of a sensitive feedback loop also makes them trivial to build - if you've wired it right, it's guaranteed to work.

Audio amplifiers in particular have almost always used unregulated supplies. So let's talk about audio SMPS.

One of the most important differences is that since there is no feedback loop to eat the mains frequency ripple, the primary capacitors have to be much larger than in a regulated one. For a regular half bridge with the primary caps in series for 230v, or in doubler configuration for 120v, i recommend 2x 1000uF at the very minimum. Otherwise you'll end up with a lovely hum in your speakers.

Again, since there is no feedback loop, the secondary capacitors will also have to be big. In fact, they will have to be as big as they would be if you were using a mains transformer. Basically, the only thing that shrinks is the transformer - but i feel this is entirely worth it.

Now that i mentioned the transformer, it too needs special precautions. Since the power supply will run at maximum duty cycle all the time, the transformer will have to be designed with nominal line in mind, instead of minimum line for a regulated supply. That means there will be more turns in the primary. This is to some degree offset by the fact that the current will be lower (if mains voltage drops output voltage drops too, so current does not increase), so you can use thinner wire.

Also, since the PWM IC runs at maximum duty cycle, there is no need for overvoltage protection. For an audio amp supply, primary overcurrent protection (also called overpower protection) and thermal protection are all that is needed.

There is a way to get line regulation and small primary capacitors in an unregulated SMPS: By adding an active PFC stage in front of it. Depending on the ripple rating of the cap, it may be made 3 times smaller. Since the active PFC stage keeps the voltage on the primary cap constant, you now also have immunity to changes in line.

That's all for today.

[Th3_uN1Qu3]

Re: The Power Supply Modding Guide (or TL494 -> Anything)

I did a lot of talk with no pics, so now i provide pics of an actual working unregulated supply built by yours truly. It has the following specs:

• 600W continuous power output
• +/-35v and +/-20v output voltages, unregulated
• Undervoltage (serving as short circuit) and overtemperature protection
• Full range thermal fan control, fan is OFF until temperature reaches 40C.

A few clarifications are in order. First of all, don't ask me about the obvious bad brands on the caps. The small Fuhjyyus you see on the board aren't even used anymore. As for the others, an audio amplifier supply stresses them a lot less than a computer supply. And most are in non-critical places which put almost zero load on them. Exception to that are the 1000uF Jakec caps filtering the +/-20v supply, but even that is going to 7815/7915 regs so there's no problem.

And yes, i know those Chemicons aren't genuine. But i don't care who made them as long as they do their job.

That input filter is way too close to the stripboard with the protection circuit. I know, but that was the only place i could put it, the layout of this power supply case is horrible. It has no chance to touch that board, and it doesn't disturb the comparator with interference, so i'm alright with that. In fact, it originally was soldered to the power connector, so i couldn't have even fit the modded board in place at all!

There should have been overpower protection. I was too lazy. In a real situation i wasn't even able to trip the undervoltage protection with shorts - the huge output capacitance simply blew the wires to bits, while the maximum current drawn from the mains (my meter has a peak hold function) never exceeded 0.33A. The only way to short this would be if someone dropped a screwdriver on the outputs, and then if that's the case the protection will step in and do its job.

By the way, there's a pair of 10 000u 50v caps in the supply. There are another three pairs, one for each amp (two bridges and one stereo module, all using TDA7294 ICs), so that's a total of 8 pairs of caps on the output. Divided by two (since they are in series across the +/- supply), that gives 40 000 uF with 70 volts across it. That can do some damage, believe me... If something shorts across that, it's more likely to blow to bits than to trip the protection.

The 40v output is with zero load (other than the fan, which hadn't even turned on at that time). As soon as even a small load is put on it, it'll drop to the designed 35v. The idle consumption of the amplifier ICs alone is enough to do that.

Awaiting your questions.

[b700029]

Re: The Power Supply Modding Guide (or TL494 -> Anything)

By the way, there's a pair of 10 000u 50v caps in the supply. There are another three pairs, one for each amp (two bridges and one stereo module, all using TDA7294 ICs), so that's a total of 8 pairs of caps on the output. Divided by two (since they are in series across the +/- supply), that gives 40 000 uF with 70 volts across it. That can do some damage, believe me... If something shorts across that, it's more likely to blow to bits than to trip the protection.

40,000uF @70V = 98 Joules.

[Th3_uN1Qu3]

Re: The Power Supply Modding Guide (or TL494 -> Anything)

After a fairly long pause i'm back in business. This time we have a Powerlink "400W" unit which will go thru a radical transformation which will (hopefully) have as end result a split adjustable regulated supply. This PSU was recapped by me, but yes, that's still offbrand caps like G-Luxon and YC. Anyway, i just needed it to start up, not power a computer for any length of time.

This PSU was originally controlled by a SG6105D. The 6105 is a TL494 + two TL431 + a preconfigured quad comparator (LM339). Since the voltage limits are configured internally, there is no way of drastically changing the voltage. So what i did was pull out the 6105 and install a daughterboard with a TL494 instead.

But first things first. I took a couple measurements of the original configuration. I found the oscillating frequency of the 6105 to be 76kHz, which made the working frequency of the supply 38kHz (for half bridge, switching frequency is fosc/2). Since i am going to be using the original primary parts, i am going to keep the same switching frequency. Then i measured the peak voltage at the transformer secondary, this will be necessary later for calculating the compensation. Of course, the secondary voltage is going to change because i want higher voltage outputs, but the first step was to get the TL494 working with the existing circuitry.

I found peak secondary voltage to be 27v at the 12v output and 11v at the 5v output.

After i removed the SG6105D, the first thing i needed to do was install a TL431 to regulate the 5vsb. Since the SG has two onboard 431s, the 5vsb doesn't use a separate one and i had to supply my own. Then i wired up the 494.

Also, an interesting piece of info: This supply had a 14v auxilary rail on the 5vsb transformer. In 494 based PSUs this is used for feeding the 494 and driver transistors. In this PSU, it wasn't used for anything! But it was complete with rectifier and output capacitor. So, there was no problem adapting a 494 to this design. I also found out that the SG6105 has built-in pull-up resistors for the outputs, so i had to wire those in as well before the 494 would start the supply. They are the 220 ohm resistors you see on the back, wired between the 494's collectors and the aforementioned 14v auxiliary.

I'm going to be using a 5v not a 2.5v reference, so i only took feedback from the 12v rail. Please note: I have no intention of using this PSU to power a computer ever again, and it is no longer suited to do so. I just wanted to make sure it will run with the 494 before getting into the big mods. Obviously, i had to redo the compensation. I also added a 10k pot in series with a 10k resistor for some degree of control over the former 12v rail's output voltage. I can vary it from 11v to 16v without any issues. I'm not going beyond that yet, because the 12v output cap is 16v.

The TL494 PCB that i made is missing a resistor btw. You should cut the trace going from Vref (pin 14) to the inverting input of the first error amp (pin 2) and install it on the back. It should be the same value as the upper resistor in the voltage divider that sets the output voltage. I finally figured out why my compensation wasn't working, and that was the issue.

Attached pics, schematics and charts. In the last pic i'm charging a small SLA. Next up will be the big mods - replacing the diodes and caps and rewinding the transformer and inductor.

[SuperDuty]

Re: The Power Supply Modding Guide (or TL494 -> Anything)

I just did a very basic mod : adding SATA connectors and a +4 ATX connector to a Fotron power supply. Removed the wiring from a crap power Heden power supply, and soldered in the right predrilled holes on the Fotron PCB. Nothing fancy, I don't have the tools and I do not have the knowledge to do something more complicated.
I did this on two PSUs. I have plenty of power supplies, many crap ones, and the good ones are missing some connectors for modern computers.

[Evil Lurker]

Re: The Power Supply Modding Guide (or TL494 -> Anything)

Questions... its been a while since I tinkered with one of these IC's but I did get it to oscillate in open-loop mode using the on board voltage reference.

What is the correct configuration for driving a single mosfet? If I understand correctly, the TL494 has two outputs that can be configured in totem pole configuration or parallel by either connecting one of the pins that drive the internal flip-flop or whatever its called either high or low. Could you please elaborate a bit on this?

Also, with the two onboard comperators and feedback pin, how are they generally configured?

[Th3_uN1Qu3]

Re: The Power Supply Modding Guide (or TL494 -> Anything)

What is the correct configuration for driving a single mosfet? If I understand correctly, the TL494 has two outputs that can be configured in totem pole configuration or parallel by either connecting one of the pins that drive the internal flip-flop or whatever its

For single ended output you wire the OUTPUT CTRL pin (pin 13) to GND instead of vref, and you can then parallel the outputs, collector or emitter, whatever you need. In that case the output frequency is the oscillator frequency, unlike the push-pull output case where output frequency is 1/2 fosc.

[badcoils]

Re: The Power Supply Modding Guide (or TL494 -> Anything)

I am also on the way of modding an 450W ATX Power supply into a +-40V dual power supply suitable for audio power amplifiers. The "normal" feedback is OK, and it is supplying the required voltage into a dummy load. However, the last thing I have yet to perform successfully, is the feedback loop compensation. I have tried the type III compensation but it is not working very well; the over-power protection circuit trips very often when the audio amplifier (which is powered by the power supply) is producing some bass. I added the loop compensation components by adding the R and C components onto the error amplifier in the controller IC (WT7520), mainly to save cost. Perhaps I was not doing it correctly?

But I have a question: How do you actually performed your loop compensation? Did you use a separate op-amp to do the compensation, and then connect its output directly to the PWM modulation voltage pin?

[Th3_uN1Qu3]

Re: The Power Supply Modding Guide (or TL494 -> Anything)

I added the loop compensation components by adding the R and C components onto the error amplifier in the controller IC (WT7520), mainly to save cost. Perhaps I was not doing it correctly?

But I have a question: How do you actually performed your loop compensation? Did you use a separate op-amp to do the compensation, and then connect its output directly to the PWM modulation voltage pin?

That's what it's supposed to be done. Of course you can use an external opamp - you would do that when the one inside the IC doesn't fit your needs, for example if it is too slow (low bandwidth). But i suspect this is not the case here.

Feedback loop compensation has nothing to do with the overpower protection. That's a different animal and is tweaked from a different place. It could be either on pin 4 or pin 12, but my guess is pin 12, which has a threshold of 0.62v. Follow the trace and you're likely to find a resistive divider fed by a diode coming from the driver transformer. Play around with that divider and see where you get.

[badcoils]

Re: The Power Supply Modding Guide (or TL494 -> Anything)

Yes, you are right, it is the pin 4 that controls the overpower protection. However, it serves two purposes in the SMPS: when the voltage crosses 1.25V it shuts down the supply(overpower protection), when the voltage is lower than about 0.6V it disables the undervoltage protection. I found some voltage divider (linked to the base drive transformer) and also a capacitor near that pin, presumably to provide the delay during startup before enabling the UVP.

However, what I feel is that, the error amp responds to the transient drop in the output voltage too aggressively by commanding the switching transistor with high (very high) duty cycles and causes the over-power protection to trip too soon. I was able to reduce the occurrences of the undesirable tripping by adding some compensation circuitry, but still not very satisfying.

Tweaking the voltage divider at the pin 4 is not a good idea, the voltage there is already near the lower threshold of the Disable-UVP function. Lowering the voltage will only void the short circuit protection!

Another thing, the error amp in the WT7520 has its inverting input biased to 2.5V, as mentioned in the datasheet. Dealing with this kind of error amp, what is the best approach to take? It doesn't seem feasible to apply the 2pole-2zero compensation circuit, or I missed something?