H.V. regulated power supply

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

Deepdark

Well-known member
Joined
May 19, 2013
Messages
1,321
Location
Quebec, Canada
Hi guys

I would like to have your opinion about the linked suply I drawn. It' s more of a practice and Learning point of view than a official design or what ever. When speaking about high voltages, I speak about voltages used in tubes circuit (preamp, compressor, etc.). I know tubes don't require regulated supply and lot of time, a simple RC filter stages is plenty enought, but again, this is simply a kind of homework I would like to test.

So, it's based around a TL783. I figures a Vout of 300v, just for the exemple. Vin is 260v x 1.414 = around 370v. Vout - Vin = 70v. Tl783 requires between 1.25v and 125v so I'm well between.

The spec sheet speak about a ripple red of 76db. I add some filtering in hope it will help reducing that value closer to 100db and up. Is it overkill, not recommanded, or I could strap some more filtering?

Thanks guys
 

Attachments

  • H.V. POWER SUPLY.pdf
    421.7 KB · Views: 51
Deepdark said:
Hi guys

I would like to have your opinion about the linked suply I drawn. It' s more of a practice and Learning point of view than a official design or what ever. When speaking about high voltages, I speak about voltages used in tubes circuit (preamp, compressor, etc.). I know tubes don't require regulated supply and lot of time, a simple RC filter stages is plenty enought, but again, this is simply a kind of homework I would like to test.

So, it's based around a TL783. I figures a Vout of 300v, just for the exemple. Vin is 260v x 1.414 = around 370v. Vout - Vin = 70v. Tl783 requires between 1.25v and 125v so I'm well between. 
Unless your transformer is oversized, the unreg voltage would probably be closer to 340-350Vdc. But you have to take into account +/-10% variations.
The spec sheet speak about a ripple red of 76db. I add some filtering in hope it will help reducing that value closer to 100db and up. Is it overkill, not recommanded, or I could strap some more filtering?
You haven't really added more filtering, you have just oversized the output cap.
In order to give you a better answer, it is necessary to know what current you expect to draw; that will define the unreg ripple. Where does the 100dB target comes from?
Let's say you want to draw 50mA (which would make worst-case dissipation close to the maximum), the input ripple would be about 0.5Vp-to-p. With 76dB reduction, that would be less than 1mV p-to-p at the output. Is it acceptable?
Increasing the output cap is of limited effect on the ripple reduction. This is all in the datasheet.
If you need additional reduction, you may add an RC filter at the output, or better, an active filter (google for capacitance multiplier), at the expense of reduced load regulation.
 
Thanks for the reply and explanations.  Your 5v p -p is taken from I/2fC, where C is 1uf, isn't it? So 50ma/120*1x10e-6 qhich give about 420mv (0.42v). Am I allright?

and 76db loss is equivalent to 0.0001v p-p, right?
 
Deepdark said:
Thanks for the reply and explanations.  Your 5v p -p is taken from I/2fC, where C is 1uf, isn't it? So 50ma/120*1x10e-6 qhich give about 420mv (0.42v). Am I allright?
No. Ripple is defined by the discharge of the capacitor between two half-periods.
Q=CV => I=C.dV/dT -> dV=I.dT/C e.g. for 0.05A and C=1mF (1000uF) dV=(0.05*0.01)/0.001=0.5Vpp
and 76db loss is equivalent to 0.0001v p-p, right?
  Academically 76dB attenuation is a ratio of 1/6300, so the residual would be 0.5/6300=79uV, but I think the attenuation is not the same at all frequencies so the final result may differ significantly.
 
No. Ripple is defined by the discharge of the capacitor between two half-periods.
Q=CV => I=C.dV/dT -> dV=I.dT/C e.g. for 0.05A and C=1mF (1000uF) dV=(0.05*0.01)/0.001=0.5Vpp
Got it. dT is the time for one cycle, so 1/60hz = 0.016sec * current 50ma / capacitor 1000uf. Right?
  Academically 76dB attenuation is a ratio of 1/6300, so the residual would be 0.5/6300=79uV, but I think the attenuation is not the same at all frequencies so the final result may differ significantly.
[/quote]

Can we simply convert db loss in volt or there is some other factor to take into account?
 
Deepdark said:
No. Ripple is defined by the discharge of the capacitor between two half-periods.
Q=CV => I=C.dV/dT -> dV=I.dT/C e.g. for 0.05A and C=1mF (1000uF) dV=(0.05*0.01)/0.001=0.5Vpp
Got it. dT is the time for one cycle, so 1/60hz = 0.016sec * current 50ma / capacitor 1000uf. Right?
Full-wave rectification happens at twice the mains frequency, hence 100Hz=> 10ms for me; 120Hz=>8.3333ms for you.
  Academically 76dB attenuation is a ratio of 1/6300, so the residual would be 0.5/6300=79uV, but I think the attenuation is not the same at all frequencies so the final result may differ significantly.

Can we simply convert db loss in volt or there is some other factor to take into account?
That's what I did; -76dB => 1/6300 => 0.00016
But the attenuation after 1kHz decreases significantly, so the higher harmonics are not as well attenuated as the lower.
 
TL783 data sheet
 
Fig 4: rip reject is not "76dB" except when Vout is 10V. For Vout of 100V it is more like 50dB.

Only an ideal regulator could have constant rip reject at "all" frequencies. For the speed of devices commonly used, and not counting capacitor-action, rip reject is about zero above 1Mhz and falling by mid-audio freqs. See Fig 6.

*** The max in/out differential is 125V. *** OK, your PDF shows 370V in, 300V out, 70V diff. But in ALL conditions? What about when the input pops-up to 370V, but the output has that monster 1,000uFd cap and will NOT come-up fast? I think you will have 300V across the TL for quite some time, more than enough to kill it.

> it is necessary to know what current you expect to draw

Yes, yes, yes!! It is not a "power supply design" until you specify the POWER it is going to make. Voltage is only half the problem. What is the current??

If the current is 1 Ampere, the 1,000uFd input cap may be appropriate. 1uFd per mA is a fair first-pencil number.

But 300V at 1A is like a TWENTY-6V6 power amp, 150 Watts output. And 6V6 does not need "regulated".

If as I suspect you are juicing 5mA-20mA preamps, say 50mA max, then this design is way unbalanced. 50uFd would be a fine first cap. Using a cap 20 times bigger is poor leverage. Two caps with a resistor between will give better 100/120Hz rejection and much better on the higher buzzz harmonics. As you can probably stand 50V drop at 50mA, the resistor is OTOO 1K. 1K into 50uFd 2nd cap gives a 3Hz corner, 40:1 (32dB) at 120Hz and 100:1 (40dB) around 300Hz where the harmonics get annoying.

Since this zero to 50V drop can be *ahead* of your regulator, the final voltage is not affected. Also this resistor will *reduce* the regulator's dissipation (neater heatsink). And it gives an absolute limit on the current which might get dump in a short circuit (shorts happen).

How much rejection can you need? Usually you need a lot of sorta-clean power and a little super-clean power. It is much neater to add an R-C filter in the B+ to *just* the critical stages.

In tubes we don't need hum/buzz 120+dB down. B+ does not go to the first grid, it goes to the 1st stage OUTput, which is 20 or 50 times hotter. And often only a 1mA stage. Reaching again for a 50uFd cap, and figuring we can accept 25V drop, 25K+50uFd is 0.14Hz so any 100/120Hz crap will be reduced 750 times, 57dB right there.
 
The TL783 is extraordinarily easy to kill in this circuit - I know, I have done it several times.  Just short the output for the briefest of moments. I would recommend replacing the diode between in and out with a suitable zener.

I would also ask why you think you need a regulated power supply for tube pre-amps and line level devices. Tubes are extraordinarily tolerant of HT voltages and an RC string can give you just as much ripple reduction at a fraction of the complexity.

Cheers

Ian
 
ruffrecords said:
The TL783 is extraordinarily easy to kill in this circuit - I know, I have done it several times.  Just short the output for the briefest of moments. I would recommend replacing the diode between in and out with a suitable zener.

I would also ask why you think you need a regulated power supply for tube pre-amps and line level devices. Tubes are extraordinarily tolerant of HT voltages and an RC string can give you just as much ripple reduction at a fraction of the complexity.

Cheers

Ian

Thanks for the advice. For the regulated part, it was more of a Learning homework. I know tubes don't need regulated psu, I tend to made RC stages but, I was just curious about making regulated psu for such high voltages, and theories behind it all. 
 
Thanks PRR for such a descriptive comment. It really helps me. In another context, the phantom psu. While looking at design of 5fish audio, mnats, etc. I found that the design is basically always the same. I saw lot of design including input and output 1000uf caps. If we stick with the 1uf / ma, and a phantom power psu is about, 15ma!? then a 22uf shouldn't be a better guess? I linked the one I made. I would like to know if that looks good. It is made to feed only one channel. PT is a 12VA. I basically followed the LM317 paper but stick the in and out 1000uf caps.
 

Attachments

  • 48v psu.pdf
    601.8 KB · Views: 12
What I think is that at least, the output 1000uf is way too big. Maybe I should replace it with something like 10uf, what do you think?

For now, I tested it unloaded as it is on the linked schematic. I adjusted it for 48V and let in run like this for about an hour without any trouble.
 
Deepdark said:
What I think is that at least, the output 1000uf is way too big. Maybe I should replace it with something like 10uf, what do you think?
If 1000uF were expensive, you may want to consider smaller, but here, it would not be significant.
The rule-of-thumb PRR gave is for GP PSU's; phantom power is a specific case where reasonable oversizing offers good return, in view of the VLN constraints.
BTW, the 100pF caps are utterly useless.
LM317 is not recommended in this application, unless you chose the HV version, or better, the 783. Continuous operation  no problem, but failures may develop at turn-on/turn-off.
 
sorry, didn't see the last sentence. Could I simply swap the lm317 with a TL783 and keep the circuit as it is? And lastly, even if my 100pf caps aren't needed, if they are ceramic, is it ok, or I should take them out?

and here is a picture of the one I made out of the schematic and layout I previously linked.

I was looking at the schematic of a PA2 from TL audio and they made out their 48v psu with a LM317, but they don't use any protection diode, thought. Datasheet recall that we do not need any protection diode with output caps below 25uf, but their external 10uf cap. isn't protected. Anyway, I never heard anybody tells about trouble with phantom power in their PA2
 

Attachments

  • IMG_20150715_213302784.jpg
    IMG_20150715_213302784.jpg
    999.1 KB · Views: 12
Deepdark said:
sorry, didn't see the last sentence. Could I simply swap the lm317 with a TL783 and keep the circuit as it is?
TL783 has a worst-case minimum-current spec of 15mA; that's why you see a 82R resistor at the Adj pin on the datasheet. In fact it may work with the actual 220R but may not in some circumstances (elevated mains or high temp). LM317HV would require no modification to teh circuit.
And is 1N4004 ok for protection diode, or I'd be better off with 4002 (100v instead of 400v for the 4004)?
Don't worry about this; there is no penalty in using higher reverse-voltage diodes here.
And lastly, even if my 100pf caps aren't needed, if they are ceramic, is it ok, or I should take them out?
You can leave them, they won't do any harm, just be useless.
 
abbey road d enfer said:
Deepdark said:
sorry, didn't see the last sentence. Could I simply swap the lm317 with a TL783 and keep the circuit as it is?
TL783 has a worst-case minimum-current spec of 15mA; that's why you see a 82R resistor at the Adj pin on the datasheet. In fact it may work with the actual 220R but may not in some circumstances (elevated mains or high temp). LM317HV would require no modification to teh circuit.
And is 1N4004 ok for protection diode, or I'd be better off with 4002 (100v instead of 400v for the 4004)?
Don't worry about this; there is no penalty in using higher reverse-voltage diodes here.
And lastly, even if my 100pf caps aren't needed, if they are ceramic, is it ok, or I should take them out?
You can leave them, they won't do any harm, just be useless.

Thanks Abbey. I guess this kind of design result in clean 48v, isn't it? About how much ripple red. can we expect? 70db, or more?
 
Deepdark said:
abbey road d enfer said:
Deepdark said:
sorry, didn't see the last sentence. Could I simply swap the lm317 with a TL783 and keep the circuit as it is?
TL783 has a worst-case minimum-current spec of 15mA; that's why you see a 82R resistor at the Adj pin on the datasheet. In fact it may work with the actual 220R but may not in some circumstances (elevated mains or high temp). LM317HV would require no modification to teh circuit.
And is 1N4004 ok for protection diode, or I'd be better off with 4002 (100v instead of 400v for the 4004)?
Don't worry about this; there is no penalty in using higher reverse-voltage diodes here.
And lastly, even if my 100pf caps aren't needed, if they are ceramic, is it ok, or I should take them out?
You can leave them, they won't do any harm, just be useless.

Thanks Abbey. I guess this kind of design result in clean 48v, isn't it? About how much ripple red. can we expect? 70db, or more?
70dB doesn't mean much, since you don't know the ripple of the unregulated voltage, because it depends on the DC resistance of the transformer and the current draw.
 
> 70dB doesn't mean much, since you don't know the ripple

+1.

-OR- the acceptable ripple in the load.

I give you an UN-filtered 60V DC. For 48V Phantom you want some rip-reject. For 48V relays you probably don't.

I give you a 60V tractor battery. That is clean enough for the BBC's 1928 mike amps. (However if the battery is being charged with a raw rectifier or an alternator, it may not be.)

*Define* how clean your load is. Both overall and those key 1st-stages. Figure how dirty your source is. *Then* you can work with rip-reduction numbers.
 
Back
Top