PSU Ripple Rejection Ratio: I need to solve this!

[Philip_BlueFX]

Hello, my name is Philip and this is my first post in this epic forum.
Been following the Vlog a lot but I gotta get in touch at last.

Let me explain the situation.
I am designing this tube, mic pre-amplifier (for the first time) based on existing schematics (would like to skip the details on that for now).
I have decided to get rid of the input-output Lundall Transformers for two reasons.
   a) They are way expensive and I am trying to cut the cost down.
   b) I want to experiment! I may try a switching option between Trafos and Electronically balanced in-outs in the future and see what works where.

I should mention that this is my degree project as an electronics engineer so it's more of a learning matter than a product design.
Let's discuss the issues.

My Mains is 220V through a 220:15+15 Transformer so I get the 15V rectified via a Voltage Multiplier up to ~70V.
This stage is built for the 48V Phantom power and the TIP121 Darlington takes over the regulation.
Now, since I am going for the Electronically Balanced in-outs, I chose a TL072 opamp for the purpose. Actually two of them, and the need for a dual-rail supply emerges! After many failed tries, I ended up with this circuitry that actually works fine, ideally!
When I power the thing up through my bench supply with V0phantom=60V, it actually works perfectly.
BUT, when I power up using the transformers, meaning that I get to regulate a fluctating AC voltage from the mains, I have huge issues with VCC,VDD outputs.

-Yellow waveform represents my V0phantom at about 65V
-Blue waveform represents VDD. VCC is almost identical.

So, my VCC and VDD supplies present a very unsatisfying ripple effect that makes it's way through the output of the balanced circuits.
I believe this to be a huge problem and I would very much like to get rid of it!

I tried using the BD683 transistor as a "Capacitance Multiplier" but it doesnt seem to work at all, also I don't even know if this is a correct approach.
I have also played a little with the smoothing capacitors, adding more or bigger ones but with no greater results.
Attached are the PSU circuitry, a couple of waveforms and the in-out circuit.

Would love some help on the matter.
Thank you all!

[Ian.M]

You've got 15V AC (RMS), which you should be rectifying directly to feed regulators for the +/-10V Vcc and Vdd supplies, rather than dropping the nom. 60V output of the voltage multiplier through a Zener!   This would also eliminate the TC7660 charge pump voltage converter you are using for the negative rail, which is undesirable to keep as its likely to be a source of considerable switching noise at its 10KHz operating frequency.  It may well be possible to full-wave rectify the unreg. feeds for those supplies to double the ripple frequency, making it considerably easier to filter.

Also, a capacitance multiplier works best on the input to a linear regulator, as it works by keeping its output voltage down to close to the ripple trough voltage under load.  If the regulator has already eliminated most of the ripple a capacitance multiplier on the output will have a negligible effect on the residual ripple.   A properly designed capacitance multiplier (not a Darlington one) has a little under 1V drop from the ripple trough at full load, so you'll need to check there is sufficient headroom for your regulator to work with when the mains supply to the transformer is at its lower permitted limit (which should be 216.2V in EU countries).  You *may* need to use LDO regulators, and Schottky rectifiers for  a bit more headroom if you wish to retain the capacitance multiplier.

[Philip_BlueFX]

I would prefer to rectify directly to the +/-10V regulators but since I need this +15V AC for the +48Phantom, I don't think I can do much more than this.
I mean, I don't have a spare clean +15AC supply so I need to work with the ones I got.

I mentioned my Trafo being a 220:15+15 but the spare 15AC supply is used for the 15V heaters of the tubes and it's a pretty noisy area to attempt feeding my balanced in-out circuitry with.

Don't you agree?

[Ian.M]

So, effectively you believe you've only got one 15V AC RMS secondary available, though I doubt any noise from the filament supply would be significant after rectification and filtering.

If you keep them separate you can only get half-wave rectified feeds for the positive and negative regulators, but its still possible, although you'll need larger capacitors and higher rated diodes than the full-wave circuit requires

If its acceptable to ground the filament supply at one end, you can use the classic full bridge with center tap arrangement for the positive and negative rails, and drive the phantom power voltage multiplier off the 'hot' end of one secondary as you are doing currently, omitting the first diode and capacitor as it can 'stack' on top of the unreg. positive rail.

[Philip_BlueFX]

I really appreciate your help.

If you pardon my stubbornness, I would like to explore any ideas of how to fix my issue on my current circuitry.
I have seen this work on a 'perfectly regulated' supply, so I assume that my transformer supply fluctuations become amplified through the TL783 and produce this effect. I believe this is what we call Ripple Rejection Ratio of the regulator.
Now in the output, the ripple is about 150mVpeak so it's too much.
Are there any ideas on how I can lessen this fluctuation?
TL783's Rejection is actually higher in low output voltage (~70dB @10V). Should I look for a replacement?

To be really honest, my Zener solution to drop the voltage down from ~65V to ~20V seems pretty shady. It seems to work (though I feel the zener gets too hot) but is it really enough? I can feel that this is the problem but I don't have any real ideas on how to drop 40V and maintain enough current on both outputs.

I would appreciate any inputs on this matter.
If I don't find a possible solution, I am going for the half wave rectifier.

Thanks for your time.

[Andy Watson]

I would echo the above comments about not using the +60V to supply all the power.
An observation: The blue wave does not follow the ripple on the capacitors - in fact, it alternates in polarity. To me this looks like it is mimicking the charging current that is flowing between the transformer and the smoothing capacitors. Check this path and remember that even copper wires have resistance. In particular, check where the reference (ground pin 3 ?) of the TC7660 is situated with respect to the transformer currents.

[mikerj]

To be really honest, my Zener solution to drop the voltage down from ~65V to ~20V seems pretty shady. It seems to work (though I feel the zener gets too hot) but is it really enough? I can feel that this is the problem but I don't have any real ideas on how to drop 40V and maintain enough current on both outputs.

This is why people are suggesting that deriving your Vcc and Vdd rails from the 65v supply is a bad idea. You have a lot of voltage to drop which, in a linear regulator, means plenty of heat.

Have you checked the dropout voltage of the TL783?  It's huge, starts at 5v at very low current and gets close to 20v under worst case conditions (high junction temp, high current).  Have you performed a voltage "budget" on this design as it's possible you have no voltage overhead for the TL783?  How much current is drawn from Vcc in total? (i.e. op-amps and the nasty flying capacitor noise generator)

Does your scope trace show the V0Phantom rail under load?

[Philip_BlueFX]

I would echo the above comments about not using the +60V to supply all the power.
An observation: The blue wave does not follow the ripple on the capacitors - in fact, it alternates in polarity. To me this looks like it is mimicking the charging current that is flowing between the transformer and the smoothing capacitors. Check this path and remember that even copper wires have resistance. In particular, check where the reference (ground pin 3 ?) of the TC7660 is situated with respect to the transformer currents.

The blue wave does not indeed follow the cap ripple but it alternates in polarity, a behavior I fail to understand.
The ground is indeed connected to one (let's say the negative) 'pole' of the transformer. However, all grounds of the circuit are joined together to the mains' ground or common pin, so I believe it is indeed a zero voltage point reference.
Moreover, the alternating signal represented by the blue wave can also be found at the TL783 output, even without the TC7660 connected, with a resistor load or even without a load at all. It actually is found there without even supplying the damn thing with a voltage but having mains connected.


@mikerj
TL783 is not under a lot of heat (can touch indefinitely) since it's input is ~20V and it's output ~14V. Most voltage is dropped across the zener which heats up a lot.
I use a TL783 for the +HT 245V (hooked on a different transformer though) which is mount on a beefy heat sink (will definitely upgrade on the final product!) and it gets freaking hot (But can withstand it or it would go on a thermal breakdown as it used to do before I mount on this sink). So, this TL783 is living an easy life
My current consumption is about 35mA with a dummy phantom load, the opamps driven but without signal input.

Andy Watson's comment about my ground reference troubled me. Went on scoping around my different ground points and what (-probe on a ground, +probe on an other) you think I found? The same signal at about 50Hz and ~400mVp-p which is getting bigger the further I go between the two points. Mind you, this PSU is all on breadboard.


If this all is actually a ground issue and not a schematic issue, then I believe it can be fixed if I build it on board and maintain a single ground point as I have done with the rest of my design (which is pretty noiseless yet).
I like the idea of the half wave rectifier if it gets rid of that sketchy diode voltage drop I came up with, so I may try that and get 'on board' (got it? ) right after.

P.S. I WOULD REALLY LOVE an idea to drop this voltage (instead of just a zener) down and not go with a whole new rectifier though.
I don't know why I'm getting obsessed with this but I feel it can be done somehow.

Will update if I get a breakthrough. Thank you all for the help. Let's now get back to work

[Andy Watson]

I would echo the above comments about not using the +60V to supply all the power.
An observation: The blue wave does not follow the ripple on the capacitors - in fact, it alternates in polarity. To me this looks like it is mimicking the charging current that is flowing between the transformer and the smoothing capacitors. Check this path and remember that even copper wires have resistance. In particular, check where the reference (ground pin 3 ?) of the TC7660 is situated with respect to the transformer currents.

The blue wave does not indeed follow the cap ripple but it alternates in polarity, a behavior I fail to understand.

Exactly, in the blue waveform you are not looking at capacitor ripple. You are seeing an artefact/interference from something else.
I believe it is due to the charging of the capacitors. You are assuming that your circuit is ideal and conforms to the schematic. Grossly simplified in the first diagram. Your reference is the ideal ground at the base of the capacitor.

However, in reality, wires and connections have resistance, albeit small. So the real circuit is more like the second diagram. Where is your reference ? Is it "seeing" the voltage across the potential divider formed by Rwire and Rtransformer ? This potential divider will be conveying the capacitor charging current - which will be many times the average supply current.

The ground is indeed connected to one (let's say the negative) 'pole' of the transformer. However, all grounds of the circuit are joined together to the mains' ground or common pin, so I believe it is indeed a zero voltage point reference.

So there may be ground loops to add into the mix?

Moreover, the alternating signal represented by the blue wave can also be found at the TL783 output, even without the TC7660 connected, with a resistor load or even without a load at all. It actually is found there without even supplying the damn thing with a voltage but having mains connected.

This is confirmation of what I was saying - i.e. it is not a ripple problem, but a ground/reference problem.

The reason I believe the blue wave corresponds to the capacitor charging current is that the negative pulse has one "hump" - due to the two flying capacitors being charged-up. The positive pulse has two humps, one is due to the flying capacitors transfering their charge and the othe hump is the re-charging of the first capacitor in the chain.

P.S. I WOULD REALLY LOVE an idea to drop this voltage (instead of just a zener) down and not go with a whole new rectifier though.
I don't know why I'm getting obsessed with this but I feel it can be done somehow.

The simplest option to remove the zener would be to take the low voltage power from the first capacitor in the voltage multiplier. You are not gaining by going through the multiplier and then via the zener (in fact you are losing).

Mind you, this PSU is all on breadboard.

Breadboard!  Hmm

[Philip_BlueFX]

Ladies and gentlemen, I do believe we solved it!

I thank you all for the guidance. It was indeed a ground loop scenario.
I fought a fierce battle but the key to victory was very simple.
Let me take you through it briefly.

The phantom PSU circuit (our topic's patient) was all on breadboard, while the op-amp in-out circuit was on the same breadboard as well.
The whole issue was that these two sections (PSU and OpAmp in-outs) where sharing a common ground wire back to the main preamp circuit which is actually the ground zero of the mains. At first glance (and many many glances after), this did not matter at all. It should be fine.

The solution came to me when Andy Watson (thank you sir) mentioned that the signal I am seeing, resembles the charging and discharging of the filter caps, and it actually WAS!

So, finally, I did isolate the two circuit grounds by running an other wire for the opamp ground and there we go.
This simple solution took me over 3 days of experimenting and measuring to come up with and It's what I should have done from the beginning.
I have also made a few tweaks in my design. Circuit is still on bread but I hope I can have it on perfboard by the end of the week.

I want to thank you all for the help. I will update when this is final and I hope no new problems will emerge.

P.S. Attached are the 'currently final' design of the PSU and a few pictures of my real world.