low noise voltage regulators

http://sjostromaudio...view&id=57&Itemid=27

Thanks for the siva...
also looking at tentlabs and burson audio regulators.

The TL431 is a cheaper option .. still need to dig into this stuff.

I think the power supply is a huge problem in my DAC..desperately seeking an upgrade.

Sachi, will it be possible to use Nimh cells or VRLA as power source? For regulation you can go in for simpler series regulators built using discrete devices. 5V supply can be buffered after the 3 terminal regulator. Ref voltage source should be selected for low noise.

I am not into DIY for some years (decades so my ideas too will be a generation old

Regards
Anil

http://www.wenzel.com/documents/finesse.html

You don't even need to replace the regulators, this circuit will shunt almost all the noise voltage to ground. Basically you have to open up the output pin of the regulator chip and add a small resistor in series with the output. The value of the resistor should be such as to only create a few millivolts of voltage drop, given the current flowing through it. A 100 ohm resistor is a good starting point for figuring out the current. Once you have decided on a value, insert the value and construct and attach the Finesse.

Pulse response of the 78xx type regulators is pretty poor also, this can be significantly improved by adding a large output capacitor, but then you'll need a diode across the terminals to prevent the chips from blowing. If the pulse response is an issue (for low current devices it normally is not) then you need to think about an outboard supply.

I recommend the Finesse, it requires little construction, and is independent of the quality of the regulator preceding it. The devices are easy to obtain, and the noise is dropped to a few nanovolts. Intrinsic device noise plays no detrimental role, as the device is in shunt mode and not series mode.

thanks for the clapton...that's probably the simplest circuit that can be easily soldered point to point on a small bare PCB.

gonna have to rummage in that parts bin of mine..

Welcome

Use a reasonably fast and low noise transistor if you're building the first circuit. The recommended type is a good starting point. Stay away from older transistors as they will hurt rather than help the cause - though the actual device selection should not impact the performance too much, only reduce the effectiveness of the circuit.

If you're hooking up P2P or using perfboard, use thick traces for all return paths and try to keep the Finesse as close to the load as physically possible. Low inductance and short traces will help the device shunt the most noise.

The series resistor may already be built into the circuit, so there may be no need to cut the current path.

Another thing you may try is a bit of aluminum cooking foil glued to the tops of the DAC and output opamps, and using tape connect the foils to ground. It may help reduce a bit of EMI interference.

Maximum improvement is usually obtained by replacing or bypassing the output stage totally, as in entry-level players they can be built with cheap and cheerful (or nasty) components, I never cease to be surprised at the compromises present in commercial equipment.

Have fun. This thing can become quite addictive, so be warned!

I have been following this thread, and have been reluctant to intrude, for the sake of being obtuse.

However, I feel compelled to share "MHO" - "My Humble Opinion"

The circuit at

http://www.wenzel.com/documents/finesse.html

is a novel and interesting "Noise Cancellation Circuit" but A SHITTY Voltage Regulator.

Infact it makes Nonsense of the Voltage Regulator before it, practically negating and spoiling Every aspect of the Voltage regulator before it, for the sake of lowering the noise. Is That what you intend to do ? Then the suggested circuit is ideal...

If your DAC or whatever else, uses a Voltage regulator. I suspect, it needs Voltage Regulation.

Some basics, Any Voltage regulator is required to provide Load Voltage regulation, ie the Output Voltage must remain Constant, when different load currents are drawn.

The humble 7805 is a 5 Volt regulator, that will not let the output voltage change by more than 1/100 th of a Volt ( 10 milli Volts ) with the Output current varying from 5 milliAmps to 1500 milliAmps ! (a change of 1495 milli Amps)

Compare this with the admitted Load Voltage variations of the suggested circuit... 300 millivolt CHANGE FOR A 20 milli Amp change in current drawn.

Further, I suspect that the current drawn from the 780X regulators in yr DAC will be more than the very small Current variations assumed in the circuit description of the 'finesse.'

Obviously, like any designer, an attempt is made to highlight the improvement in the output Noise, and under-play the Hugely negative impact on the Load regulation.

The output Voltage ( and hence the Output Current) in a DAC can swing from 0 Volts to atleast 2 Volts, causing MUCH larger current demand swings on the Voltage regulator, than what has been assumed on the Finesse site. As indicated on that site, the circuit is probably more suited for instrumentation amplifiers, that need Ultra Low Noise. ( not audio ).

Though not mentioned on the site, the Finesse may be best suited for Shunt regulation rather than Series Regulation (such as 3 terminal regulators).

Please consider the overall impact, before you venture on the project. DIY often suffers from "I built it, and it works, so it MUST sound GREAT!" syndrome.

Incidentally, the 7805 has an output Noise of 40 microVolts.

eric_clapton, please do not take this as a personal assault. I am just providing my views on the 'Finesse' circuit, and its (in)-suitability in audio.

No malice intended.

**ahem** it IS a shunt regulator. It's not meant to be used in series, or did you not read the whole description? Or are you thinking that shunt and series can be interchanged? On the page you can see it is only connected in shunt mode, so I don;t know where you got the idea that

Though not mentioned on the site, the Finesse may be best suited for Shunt regulation rather than Series Regulation (such as 3 terminal regulators)

Anybody can see that the applications have a voltage regulator input and a load output.

Most voltage regulators use brute force to prevent the load voltage changing. They are normally used in series with the load, and use feedback from the output to correct any variation wrt a reference. For the 7805 et al, that reference is ground. This is also useful to get higher/odd output voltages out of a 78xx regulator, but we won't get into that here.

The finesse is NOT a voltage regulator, is not meant to be, nor claimed to be. It's a noise shunt. It is meant to be used *in conjunction with* a voltage regulator. Please read my post and the page correctly and completely. I'm not insulted, but a little surprised that you would misconstrue a page and such a simple circuit.

I would really like to know how

it makes Nonsense of the Voltage Regulator before it, practically negating and spoiling Every aspect of the Voltage regulator before it, for the sake of lowering the noise.

There must be some explanation for that, and I'd be glad for you to be the first to point it out in the world of power supplies.

The circuit works by shunting noise current to ground, which is very small compared to the overall output current - measured in microamperes. It doesn't affect the fixed regulator in any way. See that the base of the transistor is AC-coupled to the source side of the resistor, and will therefore only shunt AC. AC on a DC power supply=noise. The Finesse does *not* interact with the load, and it's not supposed to.

I also have no idea where you got those figures from

300 millivolt CHANGE FOR A 20 milli Amp change in current drawn

. I can't see it anywhere on the page. Oh wait, that graph at the bottom? That's not voltage change, that's noise voltage [nv divided by square root of frequency]. 20nV of noise voltage at the output with 300 nV
at the input. Nano, not milli. n=nano, m=milli, p=pico, etc. Horizontal scale is frequency. The upper plot is voltage input, and the lower plot the output. The vertical scale is noise voltage.

This is a pretty common approach now (Wenzel associates published it in the 80s IIRC) in instrumentation supplies. As you probably know, instrumentation is very suceptible to interference as the measurments are in nanos and picos, not something our 'golden ears' pick up.

However I'd be interested in a rebuttal if you're interested, even though you've identified a noise shunt as a voltage regulator. Or was it an honest mistake of not having read through properly?

SHOOT ! e_c... you have taken off Just the way I hoped you would not ...

1. OFCOURSE the Finesse is a shunt circuit

I was refering to the 780X regulator being the Series ckt.

I had said:

the Finesse may be best suited for Shunt regulation rather than Series Regulation (such as 3 terminal regulators).

What I was saying was that the Finesse would probably be better used in Conjunction with a shunt Voltage regulator, rather than a series Voltage regulator.

e_c said :

The finesse is NOT a voltage regulator, is not meant to be, nor claimed to be. It's a noise shunt. It is meant to be used *in conjunction with* a voltage regulator.

I TOTALLY agree. I had said :

A_N had written: is a novel and interesting "Noise Cancellation Circuit" but A SHITTY Voltage Regulator.

Incidentally, Savyasaachi in this post has requested

I am looking for a low noise voltage regulator to replace the archaic 7805, 7912, 7812 that I am using in my DAC.

So he DOES want a Voltage REGULATOR. A circuit that reduces noise but also spoils voltage regulation is not what he is looking for.. IMHO ofcourse


2.

e_c said : I would really like to know how it makes Nonsense of the Voltage Regulator before it, practically negating and spoiling Every aspect of the Voltage regulator before it, for the sake of lowering the noise.

I had already explained :

A_N had written: Some basics, Any Voltage regulator is required to provide Load Voltage regulation, ie the Output Voltage must remain Constant, when different load currents are drawn. The humble 7805 is a 5 Volt regulator, that will not let the output voltage change by more than 1/100 th of a Volt ( 10 milli Volts ) with the Output current varying from 5 milliAmps to 1500 milliAmps ! (a change of 1495 milli Amps) Compare this with the admitted Load Voltage variations of the suggested circuit... 300 millivolt CHANGE FOR A 20 milli Amp change in current drawn.

The series resistor in the Finesse does the following:

1. Trashes the Load Regulation of the 780X regulator.

2. Increases the output impedance of the 780x regulator

e_c said : I also have no idea where you got those figures from 300 millivolt CHANGE FOR A 20 milli Amp change in current drawn . I can't see it anywhere on the page.

Para 4 on the site http://www.wenzel.com/documents/finesse.html :

Consider the low power version shown in fig. 1 which might be suitable for cleaning up the supply to a low current device. A 15 ohm resistor is inserted in series with the regulator's output giving a 150 millivolt drop when the load draws 10 mA

I speculated that 20 mA would be more typical ( even CONSERVATIVELY) for variations in the 2 channels (stereo) of the analog section .... so I have scaled by a factor of 2, the exact figures used on the website.

e_c said : This is a pretty common approach now (Wenzel associates published it in the 80s IIRC) in instrumentation supplies. As you probably know, instrumentation is very suceptible to interference as the measurments are in nanos and picos, not something our 'golden ears' pick up.

YUP ! That IS what I re-iterated too !

A_N had written: As indicated on that site, the circuit is probably more suited for instrumentation amplifiers, that need Ultra Low Noise. ( not audio ).

e_c said : ... an honest mistake of not having read through properly?

Maybe that was so in yr case ?

No, didn't take off - I figured you were suggesting that the finesse be used on its own, which kind of worried me. Maybe it wasn't so clear to me that you saw it being used together with another regulator.

The one point I suppose you have is the increased output impedance due to the introduction of a series resistor, but to the regulator the resistor is a part of the load. To the load, the supply impedance is increased only marginally. We are not talking about power amp level currents, so we'll assume relatively constant load current. Unless you're running headphones off the same supply, this can be assumed to be true. A 2 volt swing (say the full output voltage of an analog stage) in output across a 10kohm load (typical amp input impedance) will require a mighty 200 microamperes of current, so we'll assume that there is no issue of an added resistor. However, we'll solve that equation too.

Let's see - a single LM4562 has a current draw of about 6mA at full output and 30 volt supply (+/-15V) into a typical 10K load (I've actually measured this), and a couple of them yielding a full balanced output config (not common in CD players, but we'll hold it for now) is about 12mA. Let's say 20mA, as you suggest, on each supply rail, resulting in your 300mV drop with a 15 ohm resistor.

It's as simple as dropping the resistor to 1 ohm if you don't like lowered voltage (that brings your drop down to 20mV, pretty insignificant but enough to pick up the microvolts of noise), and that will drop the output impedance of the combination (btw a LM7805 has an output impedance of 0.5 ohm at 100KHz, so not as benign as you think). And if you don't want to drop any voltage at all, you can always insert a single 1N4148 diode in series with the ground of the regulator, that will push up the voltages by about 0.7 volts. Or a resistor (the current through this pin is 10mA, so you can calculate different resistor values to get the required output voltage) A nice trick for some extra free voltage.

Regulation is basically for two purposes - one is constant voltage for operating circuits, and the other for removing spurious products from the supply. An insertion of a small (say <1% of the voltage dropped across it) resistor will not affect the former, and if it is inserted with the express purpose of noise removal, may be more beneficial than harmful. A single 7805 has noise voltage of 40uV, which is about 15,000 times that of an LM4562 opamp. That's much more detrimental than a drop of 0.02 volts.

A good modern opamp will have a PSRR of over 95dB, which should be able to withstand a small variation in supply impedance - which, by its nature, is at low frequencies. The problem is that PSRR drops off rapidly at increasing frequencies, a problem that a good noise shunt will take care of. Any series regulator will have a problem with self-noise, and the 78xx are the worst. A typical 7815, for example has a noise voltage of 90uV - which is 33,000 times the noise of a good opamp.

As far as the TS is concerned, he was looking at a problem and I offered a solution. I don't consider it as an exact answer, but as a suggestion of how he could get a solution by approaching it from a different angle. I could have suggested another commercial product to be dropped into the place of the three-terminal regs, but there is a possibility of achieving similar noise reduction without spending too much money, so in my book this is a very elegant way to reach low noise figures.

As for implementation, most acclaimed drop-in 'audio' supplies including super-regulated have some form of noise shunt, so it's not just instrumentation or RF that can benefit. They don't call it Finesse so they don't get sued, but the principle of any active noise shunt is basically the same.

This is a good discussion - don;t get me wrong. I like people who can get me to think. Maybe I didn;t see where you were heading earlier, but now that I see your point of view more clearly I hope you understand why I still think it's a good recommendation.

Edit: amp_nut: I realised you posted the correction about 200 microamperes as well, thank you for pointing that out. My mistake.

Hello e_c.

Thanks for sharing your viewxs and insights.

A Very small point, 2 Volts accross a 10K resistor is 200 micro amps.

Typically, the low voltage consumption will be dominated by the quiecent currents in the circuit, and as you rightly said, the current into the input impedance of the pre amp or power amp will be small. The standing current of the buffer op amp will be Significantly higher.

The DAC that Savyasaachi uses, most probably has a Valve buffer..

I dont think you can drop that series resistor arbitarily low, ( eg 1 ohm where the series current thru it is 10 mA is probably too low ) and still have the Transistor pick up the PSU noise, invert it in phase and cancel it. I suspect that the values indicated on the wenzel site are about as low as you can go ( for the IxR drop ) and still get good noise cancellation.

I believe that the (undesirable) output impedance hike by maybe even a factor of 10 or more is an undeirable effect of this circuit. The super reg recommended by Siva stays clear of any series sampling resistor, and infact I have not come accross Any super regulator with this. Ofcourse, I hastily add, that I my decade old knowledge on the topic is rusty... ( though not fautly )

However, I found the wenzel circuit VERY interesting and thought provocing, as a noise cancellation ckt. I would love to ( plan to ) design a shunt regulator, with the wenzel circuit built into it... Maybe several already exist on the net....


Yes, modern Op amps have Superb PSRR. The 95 dB ( approx 50,000 times ) that you mention will not only reject the Power supply ripple ( as you point out for the effects of the series resistor ) but will also attenuate the noise by a similar amount. .... so if the ripple wont hurt on the Power supply, neither will the noise, by the same argument.

Thanks for sharing, on the forum. I CERTAINLY learned about the wenzel circuit, which I was not aware of.

If you know of other interesting links that use the wenzel, in conjuction with a shunt reg, do share.

Cheers

If you know of other interesting links that use the wenzel, in conjuction with a shunt reg, do share.

Afraid not - since it's patented I doubt anyone will publish a design using it publicly. I'm not sure how many people have used it privately though

20 mA of current though a 1 ohm resistor is 20mV, I suspect it'll be enough for a 2N4401. You can drop it even lower, but then you need a better error amp (a circuit on that page uses a .05 ohm resistor, but with an opamp).

As for series super regs, they do exist and are easily available either as PCBs, kits or fully built units. Usually they are based on discrete devices, and may use an error amp fed from the output (one of the Jung regulators fed the opamp rails from the post pass voltage, IIRC) or the input DC. They can be used standalone or with a three-terminal. The link sivat provided is similar, but this IS the DIY forum isn;t it?

A link to get you started: http://www.tangentsoft.net/elec/opamp-linreg.html

Have fun.

Hi e_c,

Thanks for the link / pointer.

I am aware of this link, a GReat intro on Super Regs, but it is for :

1. Series Regulators ( not Shunt )

2. Does not use wenzel type series resistance... as part of a Super Reg design.

Infact I have not been able to find Any Super reg design with a series resistor/ wenzel or a Noise shunt ckt....

That is what I am keen to learn / know of, since its new to me... ( Noise Shunt used in any Super regulator design, which you referred to in your post )

Sivat's link does Not use a noise shunt ....

I have also read up on Audio Note's regulator... another interesting approach.

'Super regs' by design will not use a noise shunt, as they are designed to be quiet from the get-go. Noise shunt is meant for noisy supplies, the Wenzel is an aftermarket design, meant to clean up noisy supplies.

If I'm designing a new regulator ground-up, I wouldn't use a series resistor/Wenzel. If I'm looking for a method to quieten up an existing supply, I would use the Wenzel or similar. Horses for courses.

Like I said earlier, most super regs (including the ones in the tangentsoft link) have a noise shunt - you can't see it because it's buried in the feedback path, the opamp is doing that job. Look closely. It doesn't have to be a series resistor to shunt noise. When comparing the output voltage to a reference (assuming the reference is quiet), noise will be part of the error signal and therefore eliminated. The principle of operation is exactly the same. The Wenzel does it post-reg, and the super regs build it in.

I don't see this discussion heading anywhere anymore, btw. If you're asking me whether I know actual implementations of a noise shunt with a series resistors in published designs or commercial products, I don't. That's because it is not meant as a build-in design, and I thought that was clear from both its description and my posts. I'll take off now, if that's all right by you.

Have a nice day.