Low Noise Discrete Power Supplies for Phono/Line Preamps

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I've been looking at developments in simple discrete circuits for RIAA preamps and line amps for about 10 years. There are a lot of circuits I've designed and built (and documented here and there on this site) and I'm going to do some sort of thread to consolidate them all when I feel up to it. After 10 years, it's time to do some sorting and summing-up and figure out which topologies have the most promise.

At any rate, some lot of the more simple and ultimately promising circuits (esp. for RIAA first stage duty) have poor supply rejection, so they require a quiet, well-behaved power supply in order to shine. I'll be presenting 4 discrete regulator circuits here that look good - 3 shunt regulators, 1 series. Three circuits use an LED voltage reference, one uses a zener diode reference. Consider them as grist for the mill (in advance).

None of the circuits to be presented in this thread use more than 4-5 active devices, so they can all be built discrete-fashion on a piece of perf board and not suffer for it. Several have been already presented on Diyaudio, but were lost in the shuffle. I'll present circuits, simulation results, and where applicable, a link to the original presentation thread. More later - this is just the teaser.
 
Since three of the four circuits to be presented are shunt regulators they'll get discussed first. The lone series regulator in the suite was just ginned up late last week in simulation, and hasn't been built yet. I'll present it here because it has some interesting features that make it akin to the shunts and perhaps a valuable addition to a designer's tool kit.

The shunt regulator circuits I'm showing here are all based on the architecture shown in the attached diagram - a stiff voltage source fed by a stiff current source. A resistor of appropriate value and wattage could be used instead of the current source, but the current source provides extra attenuation/isolation from whatever nastiness is present on the unregulated supply. The current source is set for the maximum current draw anticipated by the circuit you're powering, plus some extra margin to keep the voltage source happy and stable (and conducting) under all load conditions. A good shunt regulator circuit is set such that the voltage source never goes into cutoff under all load conditions, just like a Class A amplifier.
The circuits I'm presenting were designed mostly around requirements for low noise and low output impedance, so that the output voltage does not vary as a result of the audio load imposed on it. Setpoint accuracy and output drift are ok for the circuits presented, but not by any means spectacular. For most simple amplifier circuits, these won't be as important as noise and output impedance. Circuits with poor PSRR such as the Pacific/Simplistic RIAA preamps and the like demand a quiet supply voltage that doesn't jerk around under load, elsewise you get hiss and other antisocial behavior.

The circuits presented here have no IC opamps. To me, opamps (especially the wide bandwidth, premium performance variety) are like thoroughbred horses, pretty and capable of flashy performance, but requiring a lot of pampering to behave properly, much less shine. Those enamored of the thoroughbreds can check out threads on the Jung, Sulzer, and other opamp-based regulator solutions discussed extensively in other threads. The thrust here is that a small pinch of discrete components can provide impressive performance and predictable behavior, with much less of a need for pampering (and a wider range of operating voltage).
 

Attachments

  • Basic_Shunt.gif
    Basic_Shunt.gif
    6.6 KB · Views: 1,849
For enterprises like this, it's best to get the simple stuff out of the way first. Attached is a set of possible options for discrete current sources. I presented six possible solutions, in the spirit of completeness and perhaps being nice to folks that have fewer options for obtaining the necessary components.

To be honest, though, I rarely stray beyond the simplest solution, "A". This consists of a depletion-mode N-Channel mosfet set up as a current source. Resistor R1 sets the output current, and R2 is a stopper resistor to keep Q1 from inappropriately bursting into song - mosfets operating in the linear region are notorious squealers unless they're dampened down.

For lower currents, the Supertex DN2540N5 depletion mode fet works ok. However, it has a minimum IDSS of 150 ma, which makes it marginal if you want to consider one regulator to power both R and L channels. A much beefier workhorse is the IXYS IXTP08N50D2, which is a 500V, TO-220 device with an IDSS of ~ 800 ma. They don't cost a lot of money, but not a lot of people stock them. If you don't want to go to the trouble, you can use the mosfet/bipolar "ring of two" current sources shown in "B" and "C". I used these circuits before I discovered the IXYS device. I used mosfets for the pass elements in these circuits, as they tend to have a higher output impedance than a bipolar transistor. The bipolar solutions shown in "E" and "F" are cascoded to help get around that problem.

Circuit "D" is one I haven't tried yet, though I see no reason why it wouldn't work OK. It uses a constant current source (Q2) to set up a constant voltage across R2 to act as a voltage reference, which along with R1 and the Vgs of Q1, sets the operating current. C1 bypasses R2 to make the voltage across it more quiet in the audio band. If I remember correctly, Salas used a concept similar to this in his "simplistic shunt".

This is a start. I'll go into more detail later, but I'm nursing a cold right now. Talk amongst yourselves, ask questions - back later.
 

Attachments

  • curr_sources.gif
    curr_sources.gif
    26.2 KB · Views: 1,890
Last edited:
Here is a link to the thread containing my first shot at a discrete low-noise regulator. This was made when I knew less about LED characteristics than these days. If you read the previously cited thread on LED/zener noise closely, one realizes that this regulator is a candidate for either an LED reference or a 10-12V zener.Two instances of this regulator are currently working in my living room system, one with green GaP LED reference and one with red GaAsP reference. Both as quiet as a mouse, at least in ear-to-the-speaker terms, which is more than I can say of the TL431-based regulator they replaced. I'll be posting an updated schematic and simulation profile as time permits. The link -

http://www.diyaudio.com/forums/solid-state/116453-regulator-riaa-preamp-line-amp.html
 
Last edited:
Here is the update simulation of the circuit in the thread referenced in the previous post. The simulation uses a voltage source and a resistor to simulate a string of GaAsP deep red LEDs (low impedance (~1.5 ohms at 10 ma bias), low noise). An exposition regarding this circuit will follow when I've had some sleep, including some general comments on the advantages of a good shunt regulator vs. a series pass solution
 

Attachments

  • firstshunt_bip.gif
    firstshunt_bip.gif
    29.6 KB · Views: 1,691
In the previously shown circuit, the voltage source element consists of a current mirror loaded differential amplifier (Q3-Q6) driving an N-Channel mosfet sink/pass element (M1). Q3 and Q4 are high gain, low noise transistors - there are plenty of Euro and Japanese transistors that could serve as well. I suspect that a pair of 2N4403s selected for high gain would work as well, especially s I've already used them.... I have no trouble using the HFE function on my DVM finding 2N4403s with gains of 300 or greater. The function of the high gain in this circuit is to reduce the amount of base current, to reduce offset. The actual gain of the diff pair is determined by the tail current. I use a current source (J4, R4, R5) to set the diff pair tail current and to provide supply immunity, so that any noise on the incoming supply is not passed to the output.

I've also spun a couple of instances of this regulator using a pair of matched 2SJ74s for the diff pair, set for a tail current of 10 ma (5 per side). This works quite well and gives one the option of additional RC filtering for the reference side of the error amplifier without introducing significant bias current error. However, given the scarcity of 2SJ74s, there are probably better uses for them elsewhere.

Tail current is set for the bipolar circuit shown at 2 mA (1 ma per side), a compromise between gain/drive capability and offset/bias current.

The simulation circuit shown uses a voltage source and resistor (V2 and R15) to model the characteristics of five series-connected GaAsP deep red LEDs, each with a voltage drop of 1.6V and an incremental impedance of ~1.5 ohms at 10mA bias current. Using a fairly large number of series LEDs has a dual purpose - first, it provides sufficient bias voltage for the differential pair/current mirror stage as well as sufficient voltage to drive the mosfet pass element. Secondly, it reduces the noise contribution of the voltage reference to the output voltage by reducing the multiplication factor required to attain the desired output voltage. The noise contribution of the Series LEDs is scaled as sqrt (n), where n is the number of LEDs, while the required multiplication factor scales as n. This means you get an advantage in stacking LEDs for a higher reference voltage. You might also use a buried zener reference like the REF01 for low noise, but I already have a bunch of LEDs....

The LED reference is driven by a current source (J5, R11-12) for supply immunity. A shunt capacitor (C2, R16) provides voltage soft-start to reduce/eliminate turn-on thump.

One common feature of the shunt regulator circuits shown is that they thrive when loaded with lot of low impedance capacitors, something that would drive a lot of series-pass circuits crazy. The output capacitance (C1 and R14) is set for the equivalent capacitance/ESR of four parallel-connected Panasonic FM-series caps, 330uF, 50V (22 miliohms ESR per cap). The low ESR keeps the output excursion small for a fast current transient, while the regulation loop plays catch-up.

Output voltage is set by the reference voltage and R9-10. The vales of R9-10 are kept relatively low to reduce interaction with the bias current from Q3 (3-4 ua). The values shown program an output voltage of ~30V.

C3 and R17 are a speed-up network that passes fast output perturbations directly to the base of Q3 for correction.

I1 is a pulsed current source used to test the transient response of the regulator in simulation. The rise and fall times (100ns) are much faster than what would normally be encountered for an audio frequency load in order to determine the HF stability of the regulator and smoke out any sneaky instabilities.

The voltage source section of the regulator is driven from the unregulated supply (40V) by a current source (M2, R1-2, Q1). I would normally use a depletion mode mosfet and a pair of resistors for this purpose, but I don't have a model for the device I'm using that is easily incorporated into the Cadence/ORCAD version of PSpice.

The modified ring-of-two current source shown can be used by those who can't/wont source the depletion mode mosfets needed for a more simple current source. The output current is set by R2 and the Vbe drop of Q1. One thing that is not necessarily recognized is that the Vbe drop of Q1 is a function of its collector current. This should be set to 1-2 ma to get the B-E junction of Q1 past its "Knee". This determines the value of R1. A better solution is to replace R1 by a current source so that the collector current of Q1 is no longer a function of the unregulated supply voltage. This is shown in the collection of current source circuits a few posts ago.

The response to a fast (100 ns edge) current step is shown in the attached picture. The fast AC response to the step is determined entirely by the ESR of the output capacitors, and works out to a fast AC output impedance of 5 milliohms (no surprise, as that's the impedance of the output caps used in the simulation). The DC response to the 50mA step shows a DC output impedance of 1.8 milliohms - not bad for a small pinch of discrete components.
 

Attachments

  • firstshunt_bip_res.gif
    firstshunt_bip_res.gif
    75 KB · Views: 955
Last edited:
Here's the latest interpretation of the zener-based regulator presented in the previous post. Per Sala's comments. I've included a largish capacitor across the voltage-determining zener diode to suppress noise. It's scaled per the data sheet value of the zener impedance of the regulating diode in question. The simulation shows a 1N4750 zener diode, which, for some reason, is the largest included in the Orcad models I have available. For my purposes, I would ultimately use a 1N4751A zener, a 30V, 1W device.

Simulation-wise, this regulator does not fare badly at all, with a 5 miliohm AC impedance (determined by the shunt capacitor ESR) and a 0.8 milliohm DC impedance. I tried this regulator a few years ago and got home hiss in a sensitive circuit, most likely because I didn't bypass the voltage-determining zener with an appropriate amount of capacitance. This circuit looks like it's worth another try or two.
 

Attachments

  • zener_shunt_shh.gif
    zener_shunt_shh.gif
    18.6 KB · Views: 908
The shunt regulators are done with - here's a series pass regulator that just dotes on large, low ESR output capacitors. I haven't built this one yet - before I marry it with a project, I plan to breadboard it and torture it a while. You could do that as well, as it's not all that complex. And golly gee, not an opamp in sight... The only fault I see is that the regulator will probably need about 5V of headroom to function properly. There may be ways around that, if you care.
 

Attachments

  • NEWSERIES_sch.GIF
    NEWSERIES_sch.GIF
    30.1 KB · Views: 338
Here's the simulated response to a fast load step, showng both the transient and DC output impedance. A usual, the transient output impedance is dominated by the ESR of the output capacitors, and is ~5 milliohms. The DC output inpedance looks to be around 0.2 milliohms, a value that would be difficult to measure in a real situation. At any rate, this looks like pretty good performance for a small pinch of components, so worth the trouble to build and test the real circuit.

That's all there is. I may chime in later with some test results, but if you don't ask any questions, there won't be any answers...
 

Attachments

  • Newseries_res.GIF
    Newseries_res.GIF
    42.6 KB · Views: 313
As an aid to comprehension, here's the three shunt regulator circuits shorn of all simulation doo-dads so that you can see what a practical circuit implementation looks like. All three designs are set up for +40V input, +30V output, my usual working situation. Of course, these designs can be adapted for other output voltages and for negative polarity.
 

Attachments

  • Shunt_Family.gif
    Shunt_Family.gif
    57.2 KB · Views: 345
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.