Please help me understand\fix this discrete linear power supply

[budm]

Re: Please help me understand\fix this discrete linear power supply

Experiment time.
So if the "-.77V for the OP." then that 1n4148 Vf will be marginal with not much head room for error, which may explain why the problem is not always there 100%.
May be AGENT can implement my suggestion to see what will happen.

[kaboom]

Re: Please help me understand\fix this discrete linear power supply

Yes, that would work if the extra drop out voltage can be spared.
Larger silicon diodes have a lower forward drop for a given current than smaller silicon diodes do.

You're just adding another junction's worth of drop the "fault" current will flow thru- you won't fix it.



The above won't always work.




Strictly speaking, the above does not totally bypass the current, some still flows thru the resistive divider. But the lower Vf of the schottkey keeps the potential across the 78xx lower, hopefully by enough to keep from interfering with the 78xx's internal VRef and biasing.

[budm]

Re: Please help me understand\fix this discrete linear power supply

No that is not what I have suggested, the original Clamping Diode D430? is still in place, not removed, see post 82.
Since the added diodes are in series with internal resistors and that RD network and is in parallel with the external D430 (that clamps the Voltage to 0.6V) for the added diodes to conduct then you must have 0V on the resistors and the Diode gnd which cannot be. And if there is not external diode, the added diode at the GND pin will not conduct with the negative back feed applied.
Low Vf diode will be a way to go but mine is just the experiment to prove my theory.

[rievax_60]

Re: Please help me understand\fix this discrete linear power supply

Yes, that is right. I just tried it. The allowed reverse current increased from 8ma to 24ma. It is only worth a quick try only if a schottky diode is not available. That is all I'm saying.
I did mention the use a schotty diode back in post 78.

[budm]

Re: Please help me understand\fix this discrete linear power supply

What did you try?

[rievax_60]

Re: Please help me understand\fix this discrete linear power supply

What did you try?

I tried a single 1N4007 across the output of a L7812CV. It failed to start when 24ma of reverse current was supplied into the output compared to only 8ma when no diode was fitted.
Oh!, I was replying to Kaboom.

[budm]

Re: Please help me understand\fix this discrete linear power supply

Can you please try my suggested circuit? And what is the reverse Voltage level turns out to be, >0.7V?
I am glad you can do the experiment.

[rievax_60]

Re: Please help me understand\fix this discrete linear power supply

Can you please try my suggested circuit? And what is the reverse Voltage level turns out to be, >0.7V?

I have no doubt that your idea will solve the original problem so long as the adding of the extra 0.7v drop in series with the regulator's output won't cause regulation drop out. There is only 2v to spare.

[budm]

Re: Please help me understand\fix this discrete linear power supply

I have no doubt that your idea will solve the original problem so long as the adding of the extra 0.7v drop in series with the regulator's output won't cause regulation drop out. There is only 2v to spare.

Yes I know, it is operating in marginal range, may be that is the SINCLAIR way.
I just want to prove my theory that the backfeed was the problem but I did not understand what is really going on inside the IC until you provide me with that great handbook) that's all, and now we know what the correct way to prevent the problem.

[rievax_60]

Re: Please help me understand\fix this discrete linear power supply

Yes I know, it is operating in marginal range, may be that is the SINCLAIR way.
I just want to prove my theory, that's all.

I tried your arrangement shown in post 103, I didn't bother with the jack up diode between the common and ground.
Using 1N4007 doides, 1 amp of reverse current caused -0.45v at the regulator IC's output pin. The regulator was able to start up ok.

[budm]

Re: Please help me understand\fix this discrete linear power supply

Thanks, the jackup diode at the GND PIN will block the reverse current due to that added diode to the GND PIN. I also think it will also work without that original clamping diode in place if the jack up diode is in place which blocks the reverse current.

your knowlege of Sinclair is lacking,
they make a habit of selling stuff with terrible over-complex power circuits, then constantly revising them.

the Sinclair spectrum computer had 7 major revisions requiring pcb changes, and several intermediate changes with parts bridged on the pcb!

all to make THIS actually work.

,
https://www.badcaps.net/forum/showpo...6&postcount=82
My knowledge in SINCLAIR may be lacking but not in theory.

[stj]

Re: Please help me understand\fix this discrete linear power supply

What about Sinclair? Is this UNIT MADE BY SINCLAIR, same as Sinclair computer, excuse my ignorant! I do not see SINCLAIR name any where in the service manual, except the hand written SINCLAIR on the first page, otherwise I see the name MAURITRON.

Mauritron is a company that sells poorly photocopied manuals.

the unit is made by Thurlby Thandar,
they are a test instrument division of Sinclair.

look at the address at the bottom of the page - i think it's the original place!
http://www.ttid.co.uk/

Glebe Road, Huntingdon, Cambridgeshire PE29 7DR

[kaboom]

Re: Please help me understand\fix this discrete linear power supply

I tried your arrangement shown in post 103, I didn't bother with the jack up diode between the common and ground.
Using 1N4007 doides, 1 amp of reverse current caused -0.45v at the regulator IC's output pin. The regulator was able to start up ok.

That's after the junction heated up- Vf falls with increasing temperature.

Try that with 10mA and and cold 4007- you'll see that Vf is higher. Hopefully not above ~650mV.

[rievax_60]

Re: Please help me understand\fix this discrete linear power supply

I have put everything away now. I remember taking a voltage reading across the clamp diode of about 0.83v at 1 amp. At 10ma, I would expect the voltage to be even lower.

[Agent24]

Re: Please help me understand\fix this discrete linear power supply

i found this in my collection.

start around page44

That is a very nice book. Should come in handy for future too... Thanks

"Figure 4-2 is an example of a typical connection for obtaining a positive and negative power supply. The diodes
between output and GND are for preventing latchdown at startup and are absolutely necessary in the case of loads
shown by solid lines. Without the diodes, current flows in the separation regions between elements as described in
Figure 4-2 is an example of a typical connection for obtaining a positive and negative power supply. The diodes
between output and GND are for preventing latchdown at startup and are absolutely necessary in the case of loads
shown by solid lines. Without the diodes, current flows in the separation regions between elements as described in
chapter 2 and the output voltage does not rise (refer to Figure 4-3)."

Thanks for that one too, that does indeed look like the problem here!

+7 is slow to come up, not because of some "mysteriously slow" transistors, but because of a bypass cap (C306, 2.2u) from +7 to the wiper of P300. It causes the regulator to back down on any sudden positive-going noise on the +7 rail.

But it also slows down the rate-of-rise when the thing's powered up with all caps discharged. When power is first applied, pin 3 on U300 is effectively tied right to the output, until C306 charges. After that, the pot and resistive divider take over to provide an accurate sample of +7.

The -7 output has a similar soft-start, but notice one thing! C307 is only 1u. And the resistors are different values, also giving a different time constant.

Increase C307 to 2.2/50, or even 4.7/50 to slow down the rate-of-rise for -7, while keeping that of +7 the same.

The 'slow' transistor idea was only a theory. I had the idea because the +5v rail already had a bad TIP32 transistor and it seems that transistor had gone partially open or had decreased gain, leading to the base resistor overheating and the transistor completely dying when I gave it a dummy load.
Since the +7v rail also used a TIP32 with the same manufacturer and datecode, I thought that perhaps it was also going bad.

I see your explanation of the opamp and the soft-start capacitors makes more sense. I don't know if changing those capacitors would be wise though, since the opamps are not the same for both rails, maybe the values need to be different for the thing to work. I think stj is right about Sinclair, and this thing is already silly enough. I'm not going to poke at it more than I have to...

If D403 is a 1N4148, Then replacing it with a larger silicon diode might be enough. Power diodes with low voltage ratings are supposed to have a lower forward drop.

It was a 1N4148. Maybe the original D403 and 78L05 were close enough that it worked to start with and over the years things drifted and it stopped.
I replaced D403 with a 1N5819 schottky as per the databook you uploaded (I think I trust that over Sinclair!) and it seems to have solved the problem

Thanks everyone for the help! And those books will be a good read. Now I can hopefully use this thing and have some fun. I see someone wrote some PC interface software, which I will be trying out: http://sourceforge.net/projects/dsaquire/

Experiment time.
So if the "-.77V for the OP." then that 1n4148 Vf will be marginal with not much head room for error, which may explain why the problem is not always there 100%.
May be AGENT can implement my suggestion to see what will happen.

Thanks for the double-diode idea but I think I'll leave it alone for now. Replacing D403 with a schottky looks like the 'correct' way to do it, and is easier.

[fzabkar]

Re: Please help me understand\fix this discrete linear power supply

I can't shake the feeling that the Schottky rectifier is compensating for a problem rather than solving it.

IMHO a reasonable design current for a 78L05 would be 50mA. This allows for a 2x margin. Clearly 150mA is excessive for this device. Perhaps the Vf of a 1N4148 diode is less than the latchdown voltage for the 78L05 at 50mA but higher at 150mA. This would explain why the problem has only shown up now rather than at the design stage.

[Agent24]

Re: Please help me understand\fix this discrete linear power supply

I can't shake the feeling that the Schottky rectifier is compensating for a problem rather than solving it.

IMHO a reasonable design current for a 78L05 would be 50mA. This allows for a 2x margin. Clearly 150mA is excessive for this device. Perhaps the Vf of a 1N4148 diode is less than the latchdown voltage for the 78L05 at 50mA but higher at 150mA. This would explain why the problem has only shown up now rather than at the design stage.

A good point, but I don't know what the problem could be, as the thing seems to work fine otherwise, even when I used the 78L05\1N4148 (provided I switched it on twice)

The main ADC\DAC chip does get a little warm during operation, and I would expect such a chip to require some current. Also, the datasheet for the 10-bit UVC3130 (The one here is the 8-bit 3120, aside from that they're pretty much the same) gives it a maximum power consumption of 150mA...

Add the other parts on that rail and maybe 140mA is not unreasonable. In fact if anything it shows the design choice of a 78L05 was probably silly to begin with.