TL431-related (circa) problem - TNT Indiscreto power supply

Dear guys,

I've just built a nice phono preamp, designed by Giorgio Pozzoli for TNT-Audio. Here's the English page, with description and schematics: inDiscreet Phono Pre-amplifier [English]

I did the first listening test and I was being surprised by its good sound, but all of a sudden a prominent hum appeared. First thing I did was to check the power supply, which must provide a +27V and a -9V rail.

indiscreet3.jpg


(note that U1 and U2 are TL431)

I read a good -9.1V, and a horrible +33V. I checked the reference voltage across the R12 resistor, which should have been 2,5V (reference to the TL431), and it was 0 V. On the other rail, there was the correct value of 2,5V, and in fact that rail, as already stated, was doing right.

I thought the TL431 in the faulty rail was probably defective, replaced it and followed the power-up process with a multimeter. I noticed that, despite the fact that it should have taken about 40 minutes to reach the nominal voltage, in less than circa 10 minutes I was already at 27V. The reference voltage across the resistor ramped proportionally, reaching about 2,5V at 27V rail. At past 27V rail, the reference dropped from 2,5-2,6V to zero (!). A few seconds later we were at 28.5V and going, still the reference across R12 dead to zero, and I quit the test.

I tried a couple TL431's with the same result.

I'd really appreciate some help in identifying a culprit, here...

Thank you in advance, have a nice Sunday!

Giacom
 

Mark Johnson

Member
Paid Member
2011-05-27 3:27 pm
Silicon Valley
Wow! the (R7, C8) timeconstant is 470 seconds; that's 7.8 minutes.

I guess I would temporarily disconnect the rest of the preamp from the power supply and install another brand new, known good TL431. If this one burns out then you know the fault is somewhere on the power supply board. If this one does not burn out then you begin to suspect that the fault is not on the power supply board.
 
Wow! the (R7, C8) timeconstant is 470 seconds; that's 7.8 minutes.

I guess I would temporarily disconnect the rest of the preamp from the power supply and install another brand new, known good TL431. If this one burns out then you know the fault is somewhere on the power supply board. If this one does not burn out then you begin to suspect that the fault is not on the power supply board.

Thank you, I guess it was poorly described in the project page.

Well, that's a test I will surely do, but is there anything "past" that point that could fry the TL431? I didn't think so, therefore I didn't try disconnecting the preamp in the first place.

indiscreet1.jpg


Now that you pointed that out, I will try anyway.

R9 = 470 ohm ,did you use 47 ohm ?
Mona

I checked that already, it's the correct 470 ohm value. (hard to make a mistake there, since the resistors are PRP with clearly printed values)

Thank you both,

Giacomo
 
You are absolutely right about the pinning? If yes, did you load the regulator? Without load the TL431 should dissipate 400 mW or more = hot. What is your voltage after the fet?

Hi, yes the pinning is easy, since the PDF included within the project page has the PCB layouts and the components drawn "from above". Very easy to just copy and follow; anyway, since I had this issue, I already checked the pinout of both the TL431 and the IRF630, and they're all correct.

Also, I initially didn't know that the value had to be so precise (that should be the point of a regulated supply, now that you mention it... :D ), so when I tested it without load, and read about 32V, I thought it was due to my slightly high domestic AC, and went on connecting the boards as if everything was good. So, all my latest results are with the phono boards attached, under actual load conditions. Now that you make me think of it, I can therefore already answer that the phono section is OK (since I was measuring high right from the beginning, before connecting them), plus it also sounds very good until the hum emerges. It makes me guess that everything is just fine on that side of the preamp.

Should I measure the voltage past the IRF630 with or without load? And, with or without a (broken) TL431?

Thanks,

Giacomo
 
Last edited:
A correct working 431 should have 2.50 volts across REF and ANODE. If you have zero volts you have some problem. You should also study the datasheet for the right capacitance value. How hot did the 431 get?

Yep, I tried to see what was happening with the reference voltage already: put (several) new TL431's and powered up, with a voltmeter hooked up across the 1k resistor. The voltage would ramp slowly to 2.5V, and then suddenly drop to 0V. Meanwhile, the rail voltage would ramp slowly, without stopping at 27V; 27.x is the value when the reference voltage drops to zero, while it keeps ramping to 32, 33 or so.

I touched the TL431 immediately after the reference drop, and it doesn't feel hot.

Giacomo
 

seanvn

Member
2010-09-30 9:07 am
Voltage regulator chips will absolutely oscillate if you don't meet the output capacitance requirements given in the datasheet. In the case of the TL431 you can connect it to a circuit that has no supply line decoupling at all or you must connect across the TL431 at least a 100nF ceramic capacitor and really you should use a 1uF multlayer ceramic capacitor. Voltage regulator chips will definitely go nuts if you don't do what the data sheet tells you to do.
 
When the ref voltage is zero, how much is your output voltage? Voltage across R12, R11 and R9?

When the ref (voltage across R12) drops to zero, we are at about 27,4 V rail output. Didn't measure the voltage across R11 and R9, will do tonight!

Voltage regulator chips will absolutely oscillate if you don't meet the output capacitance requirements given in the datasheet. In the case of the TL431 you can connect it to a circuit that has no supply line decoupling at all or you must connect across the TL431 at least a 100nF ceramic capacitor and really you should use a 1uF multlayer ceramic capacitor. Voltage regulator chips will definitely go nuts if you don't do what the data sheet tells you to do.

Thank you, didn't know this. Help me get it: the big 4700 uF filter cap, placed between the +27V and the GND on the phono board, may be the cause? So you say I should put a 1uF cap across, say, R11, to stabilize things?

Giacomo
 

Hearinspace

Member
Paid Member
2008-06-03 5:18 am
I don't know a whole lot but I did spend a bit of time with shunt regulators a few years ago. I found that the 431 would bite it if the supply came up too slowly as it was trying in vain to hold a voltage that couldn't be there. Does it seem right to the more knowledgeable here that the 431 is set for final reg V by a resistive divider but the multiplier is set to bring up the rail slowly?
My other question is about working the lower 431 into no load - understandable as it's the ground rail - but usually you want the shunt reg to work into an impedance, no?
 
I pulled off all those three resistors nearby the TL431, and measured them. They are all correct.

Now, I also detached the phono boards and let the circuit run WITHOUT the TL431 in that rail, just to measure voltages around. I find some oddities. Here are the voltages, compared to the ones specified in the schematics:

- first off, domestic AC is 238V (a tad high)

- Secondary voltage(s): 17,9 + 17,9 V (instead of 15 V)

- Voltage on top of the upper diode, just past the first 4 capacitors: 48,8 V (instead of 44 V)

Then I let it charge for about 30 minutes, to see how it did in the output section:

- Voltage at the source of IRF630 (I'm actually using a 640, but whatever): 38,4 V (instead of 34 V)

- Vref (across R12): 3,3 V (instead of 2,5 - don't know if pulling the TL431 had an impact here)

- rail output: 36,8 V

...What do you guys think of these numbers? Anything helpful to diagnose?

Giacomo
 
All that seems to be within the limits.It's the action of te TL431 to pull down the 38V so that voltage on R12 stays 2,5V.
With 38V to 27V you get a drop of 11V on 470 ohm that's more than 22mA and 27V makes the TL431 dissipate allmost 600mW (if there is no external load).
To withstand the heat :) the IC has to be mouted very close to the PCB with a acceptable copper area.But I don't think that this the reason of a sudden dead,probably more slowly.
Try again with 1k in place off the 470 ohm,see what happens.
Mona
 
All that seems to be within the limits.It's the action of te TL431 to pull down the 38V so that voltage on R12 stays 2,5V.
With 38V to 27V you get a drop of 11V on 470 ohm that's more than 22mA and 27V makes the TL431 dissipate allmost 600mW (if there is no external load).
To withstand the heat :) the IC has to be mouted very close to the PCB with a acceptable copper area.But I don't think that this the reason of a sudden dead,probably more slowly.
Try again with 1k in place off the 470 ohm,see what happens.
Mona

Thank you for the assistance and clarifications ;)

The TL431 I have (I had :rolleyes:) are classic rounded-plastic three pin components, with no surface to place a heatsink; should I look for another enclosure?

I anticipated you with the 1k test: after discussing the matter with some colleagues at work, today, I tried that mod already. Swapped the 470 ohm with a 1k ohm. Result: nothing different happened. Same behaviour.

What's next? :D

Giacomo

P.S.: Please note that the 1k test was done prior to detaching the phono boards, therefore with load (and any issues it may carry; however, I measured it now that it's detached, and its DC resistance (+27 pin to ground) is infinite). Seeing that nothing changed, I put the 470 ohm resistor back in its place, so what I said above with voltages etc. is again with the 470 R.
 
Last edited:
Uhm, I was re-reading the TL431 datasheet, and noticed the V(KA) max (cathode to anode voltage), which is 37V.

If I read more than 38V (actually, it ended up at 39 when I checked a hour later) at the source of the above MOSFET, couldn't that be the cause why I'm smoking the regulator?

Giacomo