The rail to rail would have to be much less than rail to ground.
Switch on and switch off is an "indeterminate" zone where a lot of things can go wrong. "Race hazard" situations can exist where if one rail rises before another it can cause latch ups etc. Things like that can be very device specific.
I'm not sure that R1 and R4 in the original circuit are a good idea either. They are very high in value. I would certainly add the inverse diodes on each rail and ditch the zeners.
There is no symmetry at power on and off... it's a race to see which rail gets there first.
If there is any asymmetry it cannot be due to a capacitor connected rail to rail. It could be because of differences in capacitors connected from rail to ground or differences in load on either rail.
Expand the time scale by many many times...
At power on imagine the +16 volt rail has risen to only +3 volts after 1 second but that the -16 volt rail is climbing much quicker and reaches its correct value in that time.
It's possible that the negative rail may "reverse bias" for a few microseconds the positive rail (before it gets going) as a large rail to rail cap "charges" into the positive supply. That could cause a latching action or some other undesirable effect. It could cause the opamp circuitry (depending on whats used) to latch too. Yes thats a bit unlikely but stranger things have happened.
Similar things can happen at switch off too.
Yep.
Please include everything from transformer to measurement equipment. Please also do closeups of your soldering and the component placement/orientation.
Thanks,
Sebastian.
Last edited:
I think this a very real possibility and if this is what is happening, then self destruction of the regulators and op-amps is very likely.
As Mooly pointed out above, besides reverse biasing the internal circuiry of the device under this fault condition, you will probably also be exceeding the Vin to Vout spec - another situation which in its own right will induce failure.
The cure is the reverse clamping diode between the output of the devices and ground .
Note also the minimum load requirements to ensure regulation. I normally use 270 Ohms between the output and the adjust pins, and then calculate the adjust to gnd resistor to ensure I get the correct output voltage. This always ensures the reg has the minumum load current.
this is why I put in place on the amp PCB a pair of diodes across the decoupling or on board smoothing caps. If there are fuses on board then you may have to use double diodes either side of the fuses, depending on what polarised components are there.
This is in addition to the back emf diodes across the output devices.
A total of 6 2cent devices taking up very little board space. It always worth adding this simple and cheap protection.
Funny how you remember things...
Many years ago I built a preamp with CMOS logic selection for the input selector and the two separate record out lines using 4017 counters clocked by a 4011 oscillator. Momentary pushbuttons selected the input... anyway... I found one of the three 4017's kept failing. Once, and that maybe could be a one off or down to poor handling but several times !!
And thinking back I bet this was the same problem. I had used just one split supply with 78 and 79 regs for everything. And this one CMOS IC kept failing. In the end I reasoned it must be a voltage spike or some such and I fitted a diode across that particular IC together with a zener to clamp any surges. It never failed again but back then I didn't really understand the full reasoning behind why it failed.
Something that's easily forgotten.... thanks for the reminder.
Mooly, when you gave the example above where the positive rail reached 3V and the negative one is already at -16V, you describe a valid, temporary state where the positive rail is positive and the negative is negative-- nothing wrong with this. Now after some short time, the rail to rail capacitor draws less current from the rails, but still the same amount from both rails, so the positive rail will gain some more voltage while the negative remains constant. The reason for the different absolute voltages on the two rails were a result of differences in the components that are unique to each rail and not because of the rail to rail capacitor.
I'm tring to think of a way to explain this.
The cap draws its current "equally" from both rails (it has to) but that doesn't mean the voltage on one of the regs can not be forced outside the ratings.
Imagine the circuit without the regs, just a normal regulated PSU with rail to rail caps across the output.
In the positve rail you now fit a series 10 k resistor and in the negative rail a series 1K resistor. When the caps are charged, the rails (assuming no current draw) are at the -/+ voltage of the PSU with respect to ground.
Now think what happens at power on with a discharged cap. What is the voltage at the positive rail going to do with reference to the centre ground point ?
The cap draws its current "equally" from both rails (it has to) but that doesn't mean the voltage on one of the regs can not be forced outside the ratings.
Imagine the circuit without the regs, just a normal regulated PSU with rail to rail caps across the output.
In the positve rail you now fit a series 10 k resistor and in the negative rail a series 1K resistor. When the caps are charged, the rails (assuming no current draw) are at the -/+ voltage of the PSU with respect to ground.
Now think what happens at power on with a discharged cap. What is the voltage at the positive rail going to do with reference to the centre ground point ?
Hi all,
I am using the attached design but the lm317 blows out without any reason I can identify - I already blew 6 of them. And yesterday I blew lm337
Input is +38V 0 -38V and output is adjusted to +16V 0 -16V. It is used to power a few opmaps so current is low. The capacitors are 100uf and the zener dioes are 8.2V and I put them to protect against high voltage at the output.
Any idea what's the problem?
Thanks,
Eli
At start up, lets say that the LM337 is slower than the LM317.
R4, D7 and D8 will put the output of the LM337 at almost -80V wrt the input of the LM317. This exceeds the Vin - Vout of the LM317 so it blows.
If I understand correctly what you were describing, then the positive rail would climb slower than the negative rail, and yes, at times it could even be negative with respect to the center tap of the transformer. But you said it, your model has larger resistance in the positive rail than the resistance in the negative one, and that's what caused the asymmetry. The capacitor itself could not do it alone.
OK, so we all agree the problem is likely to occur because the regs exceeded their voltage limits. What's the solution? I did not mention but the reason the input is so high is because it should be used to power a class D power amplifier as well. If I had another transformer I could provide lower input voltage but I only have one transformer.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.