LM350T failure - counterfeit component or fundamental construction problem? Opinions requested.

[dinsen]

This is my first post, but I've been lurking for a while.

I'm restoring a Soundcraft Spirit Studio console for my daughter and a core part of the project is building a new power supply as the original was missing.

I chose an LM350T for the positive 17V rail and a LT1033 for the negative rail. The LM350T, however, has failed twice for me, and in a rather unpleasant way as it short circuits which sends roughly 33V onto the 17V supply line I want to understand the failure to ensure it doesn't happen again.

I chose to build the resistor/capacitor/diode part on a stripboard, and this is one potential source of problems as I suspect a short circuit of the regulator could damage it. I like strip boards, but I realize they may not be too great for power supplies as the distances between the strips are easy to short. I suspect perhaps peice of metal let inside the box from drilling could have caused a short. The LM350T is reportedly a tough component but looking at the internal schematics in the data sheet shorting the ADJ pin to either IN or OUT is likely to damage it. I cannot detect any short circuits nor see any metal in the box, so there's no smoking gun - it's just a hypothesis.

Being a trained engineer, I of course devised another theory: The power supply has about 1.5m of cable to connect it to to the console. Also the console has lots of 47uF decoupling capacitors inside, plus about one metre of additional wiring inside. My circuit has the usual protection diodes from OUT ->|- IN and ADJ ->|- OUT. I fitted no protection diode from +17V -|<- GND, as I didn't think that should be necessary. Theoretically however, inducing a current in the cable could perhaps cause problems... I've done some simulations in LTSPICE, but nothing that reveals any real problems (perhaps a little if I simulate the self inducatance of cables 1500 km long!), so I doubt if that could be the cause of my problems.

So I clearly needed a third theory, and I was thinking that perhaps the LM350T's I bought were bad. I just checked the numbers and they both seem to have the same batch numbers as some tested by someone on youtube (video below). I did not buy them from eBay, but from a decent supplier here in Denmark, so I'm kind of surprised if this is the issue.

We would appreciate opinions and experience from all you experienced builders - should we give up our stripboard design, include protection diodes, or find some original LM350T's (or even better some LT1085's, but they're expensive and harder to find)

Best wishes from Copenhagen,
Anders

 

[mad.ax]

Welcome aboard!

What is the size/power consumption of your console?

roughly 33V onto the 17V supply line

So you have 16V differential between the input and the output of your regulator. Given a Pd max of 25W, about 1,5A is the limit. Could it be that you're exceding it?

Axel

 

[dinsen]

Welcome aboard!

What is the size/power consumption of your console?

Thank you!

The console is 24x8x2, channels are purely with TL072 opamps, NE5532 in the groups and master, and LED vu meters with current sources driving them so I estimate the consumption to be 1-2A. Original CPS150 PSU is specified to 1.25A but fitted with LM338K's which are current limited at 5A. I decided to design for 3A.

 

[My3gger]

Welcome!
Can you measure temperature of LM350 with a temperature probe? As Mad.Ax wrote you have very high differential, make sure heatsink is big enough and reg. properly thermally coupled to it .
Veroboard is better for this kind of thing, i used to do LV and HV supplies on them.

 

[mad.ax]

Original PSU is (in)famous for being 'weak', even for 16 channels!
LM338K have a Pd max of 50W and near 6A max with a differential of 16V. And LM338K are TO3 packages, hence they are much better at dissipating heat...

Axel

 

[ruffrecords]

Definitely fit the protection diodes.

Cheers

Ian

 

[gyraf]

..seem to have the same batch numbers as some tested by someone on youtube

This very much a telling sign - many counterfeiters have one standard batch number per counterfeited parts. And if it's laser marked, not screen printed, it's even more suspect too..

Or use 78S15, cheap, plentiful and easy 15V/2A regulation

Just fit reverse-protection diodes for when powering down, and you should be safe

I've seen fake parts from several Danish suppliers - it seems to be everywhere now. EL-Supply on Bornholm usually have a rigorously controlled inventory, and are one of the few I find trustworthy..

/Jakob E.

 

[dinsen]

So you have 16V differential between the input and the output of your regulator. Given a Pd max of 25W, about 1,5A is the limit. Could it be that you're exceding it?

Good point, I didn't think of that! The box is essitially one big 190 mm x 260 mm heatsink. I didn't get a chance to measure the temperature of the LM350T on casing. However, when it worked, the temperature was essentially the same all over the outside of the heatsink, top to bottom, so I doubt if it was much more than the 27 deg C I got. Also, I would have expected the LM350T to thermally shut down instead of breaking down?

Protection diodes are a no-brainer and I definitely will be fitting the one on the output in addition to the two shown in the datasheet. I'll probably also scrap my strip board design, just to be sure.

 

[dinsen]

This very much a telling sign - many counterfeiters have one standard batch number per counterfeited parts. And if it's laser marked, not screen printed, it's even more suspect too..

Or use 78S15, cheap, plentiful and easy 15V/2A regulation

Just fit reverse-protection diodes for when powering down, and you should be safe

I've seen fake parts from several Danish suppliers - it seems to be everywhere now. EL-Supply on Bornholm usually have a rigorously controlled inventory, and are one of the few I find trustworthy..

/Jakob E.

The batch numbers are hard to read, so I suspect they are laser marked

Thanks for the tip about 78S15 and EL-supply, Jakob... 78S15's don't seem that easy to fin (RS doesn't have them).

 

[chilidawg]

Why not using the proven method of increasing the current output of LM317 using a pass transistor? Though the 16V differential between input and output is what I would deal with first. Not sure what kind of power transformer you are using, and if its secondary output has other AC voltage option that will get you 20V DC or so after rectification, hence lowering the thermal factor by 5x

 

[abbey road d enfer]

I chose an LM350T for the positive 17V rail and a LT1033 for the negative rail. The LM350T, however, has failed twice for me, and in a rather unpleasant way as it short circuits which sends roughly 33V onto the 17V supply line I want to understand the failure to ensure it doesn't happen again.

First of all, why starting with such a high unregulated voltage? It is typical to start with about 24-25V for 17V rails.
I suggest you do the "spit test" for temperature. Put a drop of spit on the regulator's case. If it boils it's too much.
Thermal protection has many ways to fail...

 

[gyraf]

Thanks for the tip about 78S15 and EL-supply, Jakob... 78S15's don't seem that easy to fin (RS doesn't have them).

I get them here:

https://secure.reichelt.com/dk/en/voltage-regulator-fixed-15-v-2-a-to-220--a-78s15-p23486.html

 

[dinsen]

First of all, why starting with such a high unregulated voltage? It is typical to start with about 24-25V for 17V rails.
I suggest you do the "spit test" for temperature. Put a drop of spit on the regulator's case. If it boils it's too much.
Thermal protection has many ways to fail...

I was inspired by the original PSU which is driven by a 2x24V transformer and when I was able to lay my hand on a cheap, new 2x22V toroid transformer I went by that route. In hindsight it might not have been a good idea, but I'll stick to it now. This high voltage also allowed me to add a phantom power supply without an additional transformer, driving the bridge rectifier through a pair of capacitors. Finally, I found a nice box with big integrated heat sink (cheap too).

I'm sure the regulator did not get particularly warm as I measured less than 30 deg C on the outside of the heatsink directly where the mounting screw was going through, even after powering the whole console for an hour (where luckly the regulator didn't fail).

I estimate the voltage will be 31V at 1-2A current draw after the filter caps so the voltage drop over the regulator will be about 15V. The below diagram is taken from the LM350T data sheet and indicates it should be perfectly able to supply the less than 2A being drawn by the console.



I think I'll order a new real LM350T from EL-Supply (thanks for the recommendation, Jakob, and by the way they even have the 78S15's!). I'll add a protection diode on the output, but otherwise keep my design, i.e. transformer and strip board construction as is.

But first, I'll "sleep on it"

I'm very grateful for the suggestions you generously shared!

Best,
Anders

 

[soapfoot]

I'm sure the regulator did not get particularly warm as I measured less than 30 deg C on the outside of the heatsink directly where the mounting screw was going through, even after powering the whole console for an hour (where luckly the regulator didn't fail).

One word of caution here--

If there happened to be sub-optimal coupling between the regulator and the heat sink, that would have two consequences:

It'd obviously make the part far more prone to overheating, but it'd also make the temperature of the heat sink an unreliable indicator of the temperature of the actual part.

There are a few things here that cause some suspicion (mainly the lot number of the parts coinciding with known-counterfeits), but actually checking the temp of the regulator itself is a pretty easy (and free!) step before doing anything more invasive (or expensive).

 

[Rob Flinn]

Thank you!

The console is 24x8x2, channels are purely with TL072 opamps, NE5532 in the groups and master, and LED vu meters with current sources driving them so I estimate the consumption to be 1-2A. Original CPS150 PSU is specified to 1.25A but fitted with LM338K's which are current limited at 5A. I decided to design for 3A.

The current rating of these devices depends on how much voltage you drop from in to out on the regulator. The greater it is the less the current rating. Refer to the data sheet which normally has a graph for this parameter.

 

[radardoug]

You say you are a trained engineer. Having up to 20 volts across the regulator tells me you are not.
Also, you are guessing about your consoles power draw. Use a couple of lab supplies and FIND OUT what it draws.

 

[gyraf]

Radar, this gets too close to a personal attack - we don't need that around here. Please try minimizing such outbursts.

 

[Nick Nack]

Why not just add a resistor in front of the regulator, to drop a few volts?

 

[dinsen]

I have a M.Sc.(eng), so I am indeed a trained engineer. I'm not an electronics engineer as I've been working since my graduation in 1995 with software systems. This journey was set out to build a linear power supply copying a SoundCraft design, and in that respect it has so far failed. I'm now learning, by experience, how to build stable and powerful linear power supplies, and I'm going to finish the project, of course, also thanks to the input I get here. I appreciate the friendly comments I get.

Here are my learning points so far:

1. I copied a bad design with a transformer with 24V secondary windings and therefore a very high voltage drop over the regulators. Repeating other people's bad designs is a bad idea. I should have researched more.
2. I reduced the voltage in my design to 22V as I found a good toroid transformer with that specification. Unfortunately it is actually a 24V transformer even though it is labeled as 22V @ 230V, so I've learnt that what's stamped on transformers can be wrong.
3. I bought a fake LM350 that failed in a disastrous way. I knew about fake components so I bought it from a credible source. I've learned a lesson there, which you have confirmed, Jakob. Thanks.
4. I chose an LT1033 as the negative regulator, and that was a wise choice. It perfoms quite well. I got it directly from Analog.
5. I have now fitted a presumably good LM350 in the positive supply. Appreciating that about 15V voltage drop over the regulator is on the high side, according to the TYPICAL curves in the data sheet it should be able to supply about 3A without big issues (see the chart above). It is not, as I will show below.

I did a test with the new presumably good LM350 yesterday and measured output voltage, temperature, and ripple.

At 16 Ohm load, the output voltage measured at the end of the rather long cable is about 16.2V on both negative and positive rails. This indicates a voltage drop over the cable as it measures 16.4V unloaded. I expected that. Temperatures measured with an IR thermometer at the back of the heat sink where the regulators are bolted stabilize at less than 30C.

At 8 Ohm load the negative rail still measures about 16.2V, but the positive rail drops to 9.5V. The temperature of the negative regulator is now about 30C whereas the positive regulator is now working very hard and increases to 35C. The LM350 obviously can't cope with the large voltage drop at 3A although the typical specification would have indicated that.

Also, somethign else starts to happen then. The scope screen shot below shows the ripple of both. Clearly the LM350 is pushed beyond it's limits and apparantly oscillating while the negative rail seems to be copying that same oscillation. This is measured at the end of the long cable with no capacitance at the end, only resistors and decoupling capacitors near the regulators.

Note that the ranges of the two channels are different, and that there's a 1.4 kHz oscillation occurring. CH1 is the positive rail, CH2 is the negative. (The channels are shown as DC, but my scope doesn't have the AC mod, so I put a capacitor between the supply rail and the probe.)



I have not made any measuremens with the console connected, of course as I need to solve the issue with the LM350. I see the following two options:

1. Replace the transformer with a 2x18V @ 2A. (It will be just enough to drive my ac coupled phantom power supply.)
2. Replace the LM350 with an LT1084, which is guranteed to be able to spply 2A with 15V drop.

Or both, of course.

 

[abbey road d enfer]

I reduced the voltage in my design to 22V as I found a good toroid transformer with that specification. Unfortunately it is actually a 24V transformer even though it is labeled as 22V @ 230V, so I've learnt that what's stamped on transformers can be wrong.

No. The rating of a power transformer is with its maximum load, so under reduced load (less current demand) it delivers more than nominal. And it's very likely that your grid delivers more than 230V. Where I am I get consistent 234V.
And your big mistake is to neglect the crest factor. A sinewave has a crest factor of sqrt 2 (1.414), so, without load, teh rectified voltage is 1.414 times the rms value minus a fraction ov volts due to teh rectifiers. It is very common to use a 18V rms winding for producing 18Vdc regulated.

options:

1. Replace the transformer with a 2x18V @ 2A.

Note that, as the rectified voltage is about 1.4 times that of the rms voltage, current drawn from the xfmr is 1.4 times that which is delivered to the DC load. For a 2A DC rail, you need a 3Aac winding.

1. (It will be just enough to drive my ac coupled phantom power supply.)

You will have to use a tripler instead of a doubler. A doubler would deliver about 50V unregulated, which leaves almost nothing for teh regulator. A tripler would easily deliver 70V, which leaves plenty of room for a CRC filter and a regulator.