How to make a programmable LM2596S Buck Booster???

[UPI]

We need to create 7 programmable 4V-31V supplies for a test jig. I have purchased some of the LM2596 Buck Converters to see if we can make these work since they are dirt cheap.
http://www.ebay.com/itm/300611676528?ssPageName=STRK:MEWAX:IT&_trksid=p3984.m1423.l2649

Reading some other postings lead us to add a 1K resistor to the Feedback pin (pin 4) of the LM2596S, with the built-in voltage adjust pot set to max output of 35V, and then to feed 0-5V into this 1K Feedback pin resistor. What we found was:

* We can control the voltage from 3.6V to 35V
* A 10mv change on the Feedback pin changes the output by approx. 100mv
* The output adjustable range was achieved with control voltages between 1.5V thru 4.7V

The attached pic is basically the same design as the one we are manipulating.



Do you guys know of a better way to do this?

[Bored@Work]

Better way: If you didn't already, calculate the resistor values in the feedback, don't just take 1k and the existing divider. It is about injecting some additional current into the feedback node.

 http://electronicdesign.com/passives/easy-way-roll-your-own-programmable-power-supply

explains this, but you better derive the equations on your own, since the ones in the article are wrong.

[UPI]

That is exactly what we needed!

With the following conditions:
We are currently using a variable 0-5V supply but will probably use an MCP4725 DAC
R1 (from FB to GND) is a 330 ohm
R2 (from FB to output) is a 7.5k resistor (in place of the original 10k pot)
VDACL = .5V (not sure why I decided on this but it is fine)
VTH = 1.23V (internal voltage reference of LM2576S)
VOH = 35V

Then:
R3 (between FB and the DAC output) should be 973 ohms

Finally:
Using a 976 ohm resistor on the DAC out, we get down to .4V on the Buck Converter out with a DAC out of 5V and we get up to 34.1V with a DAC out of .6V. We had to load the output with a ~500 ohms to get it to come down all the way to the .4V solidly and quickly.

Thanks for your help!!!

[mrflibble]

Hey, that looks exactly like what I am using those cheapo ebay lm2596's boards for.

I was about to point out that you may want to put a pullup resistor on the ON/OFF and then pull it low with an mcu (or whatever you have in there) after you made sure you have a valid DAC voltage. But then I noticed you were using an NV DAC so that's probably not needed.

As for the article ... yeah, looks like there are a few errors there.

Should be something like this: Vout = [R2 / (R1||R2||R3) ] * Vref - (R2/R3)*Vdac

Where R1||R2||R3 is the equivalent resistance of R1,R2,R3 in parallel.

It's exactly the same sort of thing as a summing node on an opamp.

Incidentally ...

It is about injecting some additional current into the feedback node.

If I understand the datasheet correctly, for the ADJ part ideally (assuming an ideal error amp) there should be no current flowing into the feedback node. As in only the amp bias current...

[Bored@Work]

Incidentally ...

It is about injecting some additional current into the feedback node.

If I understand the datasheet correctly, for the ADJ part ideally (assuming an ideal error amp) there should be no current flowing into the feedback node. As in only the amp bias current...

I think you mean the feedback pin. While I mean the network node where the feedback pin and the three resistors are connected together. Kirchhoff says the sum of all currents in a node is zero. So be injecting some additional current you shift the other currents. With that the voltage at the feedback pin changes, so the regulator needs to change its output to restore the voltage at the feedback pin.

By the way, the circuit gets a bit easier if you use a current DAC instead of a voltage output DAC. Then you can omit the additional resistor.

[mrflibble]

I think you mean the feedback pin. While I mean the network node where the feedback pin and the three resistors are connected together. Kirchhoff says the sum of all currents in a node is zero. So be injecting some additional current you shift the other currents. With that the voltage at the feedback pin changes, so the regulator needs to change its output to restore the voltage at the feedback pin.

No current flowing into the feedback pin of the LM2596 is indeed what I mean. I suspected you meant your last post in the network analysis sense, but found it a little ambiguous. You even had me doubting if I had read the datasheet correctly in the past, so I rechecked. ;-) Because for the fixed voltage parts there are internal resistors, but for the ADJ part the FB pin goes straight to the error amplifier.

By the way, the circuit gets a bit easier if you use a current DAC instead of a voltage output DAC. Then you can omit the additional resistor.

Good point. Aren't current dac's generally a bit more expensive though? Right now I just use the dac in whatever mcu I happen to be using, so additional resistor it is...

edit: fixed quote.

[fcb]

As you don't mention the current draw, efficiency, accuracy or noise required it's really difficult to comment.

You might find that LM317T/K's are perfectly good for your application and very simple to control (DAC into OPAMP into LM317 adjust pin).  Limiting the maximum current with a 317 is trivial also.

[UPI]

I was doing some more playing around with these eBay LM2596S Buck Converters and it appears that they are actually re-badged LM2576 regulators rather than LM2596. They are labeled as 2596 but their switching frequency is 52kHz rather than 150kHz. I adjusted them to 8V, loaded them with 8 ohms and scoped the output pin 2 for the test.

We replaced one with a Mouser sourced real "On Semi" LM2596S and it does operate at 150kHz.

Using an analog scope, we see ~150mV vs ~80mV on the board outputs with the only difference being the regulator IC.

We will switch to real 2596 regulators on more critical applications.

[mrflibble]

Counterfeit parts on Chinese ebay goodies? Never! I also have a few of those, so would be interesting to see if ALL of them are counterfeit or just some batches. Another thing I noticed is that the inductors on these converters get damn hot. So no doubt they skimped on that as well.

[fmaimon]

I've bought these and they seem to be real ones, as the work at the right frequency.

[UPI]

I bought these because they had the extra inductor and cap on the output and claimed to be low ripple. What a sucker.

http://www.ebay.com/itm/300611676528?ssPageName=STRK:MEWAX:IT&_trksid=p3984.m1423.l2649

[smbaker]

Last night I tried this technique with some of my LM2576 regulators, and had mixed results. I used the same resistor values that UPI used near the top of the thread (330 ohm, 7.5 K ohm, and 976 ohm). Using an MCP4728 DAC, I was able to adjust the power supply from about 3.6 V to about 35 V.

However, I destroyed three LM2576 regulators while doing this. In each case, the output driver for the LM2576 appears to end up 'stuck on' regardless of feedback voltage. There's no 53 Khz wave so far as I can tell. Thinking that it was maybe a voltage surge on the feedback pin that killed the regulators, I installed a 2V zener (smallest one I had) on the feedback pin. Still ruined the LM2576.

In the last experiment, the LM2576 was  damaged not when applying the DAC voltage, but when removing it. I had the DAC outputting ~ 4.8 VDC, the regulator outputting ~ 3.6 VDC, and I disconnected the wire from the DAC. Upon disconnecting the DAC, the voltage at the feedback pin should have dropped from 1.23V to 0.15V, and everything would have been well within bounds. I just don't see anything going on there that should be killing regulators.

Thoughts?

Thanks,
Scott

[Kalvin]

Here's the pdf for the article:
http://e2e.ti.com/cfs-file/__key/communityserver-discussions-components-files/858/5241.Programmable_5F00_Power_5F00_Supply.pdf

[smbaker]

My question wasn't about the overall technique as much as it was about the application of the technique to this particular regulator. An abrupt change in feedback voltage, and *poof* it's instantly dead. It doesn't even release any magic smoke.

I put the scope in infinite persistence mode to try to look for any transient voltages on the feedback pin, but did not note any.

Scott

[prasimix]

My question wasn't about the overall technique as much as it was about the application of the technique to this particular regulator. An abrupt change in feedback voltage, and *poof* it's instantly dead. It doesn't even release any magic smoke.

I put the scope in infinite persistence mode to try to look for any transient voltages on the feedback pin, but did not note any.

Scott

Hm, I'm afraid that is not something specific to LM2596 but at least for another two TI controllers: LM5118 and LM5088. One of the desperate attempt was using opto-coupler with LDR presented here. That was the only one that not kill LM5118 instantly. The situation with LM5088 is better but still a mystery to me. It could even works with typical pnp based tracker circuit what you can find in few LTC's application notes.

[smbaker]

I figured out the issue I was having. Looking to save $0.10 and ten seconds soldering time and knowing that I was supplying regulated DC to my prototype from my benchtop PSU, I decided to omit the input capacitor for the LM2576. Apparently that was a bad idea. I should have read the datasheet:

To maintain stability, the regulator input pin must be bypassed with at least a 100 uF electrolytic capacitor.