The TeddyReg is an ultra low noise regulator designed to be used in pre-amplifiers, CD players, DACs etc. Its advantages are very low noise, and excellent filtering of the mains and rectification noise. Since it has such a high filtering capability it allows connecting many regulators to the same power rail, avoiding the need of special or multiple transformers. In addition, it does not require large transformers and smoothing capacitors.

On the downside, its output impedance is higher than most linear regulators (like the ALWSR or monolithic regulators like the LM317). This is not a problem in most circuits consuming low current. But it is not suitable for circuits requiring high variable currents like class A-B power amplifiers for this purpose another regulator the PowerReg can be used.

Another limitation is its slow startup time. The basic circuit takes several seconds to achieve the final output voltage, which may be problematic in CD players where the clock, the DAC, and other circuits need to synchronize when it is turned on. This limitation can be resolved however using the optional accelerator circuit described below.

The TeddyReg can be implemented both as a positive and negative voltage regulator.

Principle of Operation

The TeddyReg is built as a combination of a linear regulator followed by a very powerful gyrator. The linear regulator operates as pre-regulator and provides a first level low frequency noise filtering and line regulation. Line regulation is the capability to maintain a constant output voltage level on the output channel of a power supply despite changes to the input voltage. The gyrator is a low pass filter that controls a transistor in a way that the transistor passes only low frequencies. Our goal is to make it an ultra low pass filter, that is, pass DC (zero Hz) only and filter anything else.

The first stage is a standard implementation of the LM317 (U1, C1, C2, R1, R2, and C3). Better implementations such as Tracking Pre-regulators (TPR), or ALWSR can be used, but I found no benefit in doing that (unless you already have them).

The gyrator consists of a 2nd order low pass filter. The first stage (R3, C4) uses a 100K + 22-33uF Tantalum capacitor which filters noise at the audible spectrum. R4 works with R3 as a voltage divider to create some dropout on the gyrator. This dropout is necessary for a proper operation of the transistors. The second stage (R5, C5) uses a 100K + 0.1uF Ceramic capacitor to filter high frequencies up to tens and even hundreds of MHz. Don't ask me why, but filtering noise at these frequencies which are clearly above the audible spectrum has a very important effect on the sound.

R6 is used to prevent oscillations. R7 is optional and is used to bias Q1 (slightly reducing the output impedance)

Most gyrators are using a bipolar transistor at the output. The output impedance of a gyrator is approximately the sum of the filter resistors divided by the transistor hfe (current gain). As an example, if we use a D44H11 with a current gain of 100 and two 100K resistors (R3, R5), the output impedance will be 2000 Ohm, which clearly too high. For this reason most gyrators use a much smaller resistor with a much larger capacitor. The disadvantage of such a solution is that large electrolytic capacitors have limited filtering capabilities in high frequencies. In many cases a Darlington transistor which has a much higher hfe is used. The Darlington combination presents however other problems. First problem is that its hfe increases with the current that passes between the collector and the Emitter, so actually with low currents the hfe is not large enough. Second problem is that the Vbe(voltage between the base and the emitter) of a bipolar transistor increases when the current increases, this by itself increases the output impedance. In a Darlington configuration there are two Vbe, therefore twice the impedance.

The solution used here is a combination of a JFET and bipolar power transistor. Unlike bipolar transistors, in a JFET there is no current passing between the gate and the source (the base and the emitter), which means no current through R5, and R6, and therefore large resistors can be used without affecting the output impedance. In addition the Vgs (Vbe) of the JFET is negative approximately canceling the Vbe of the power transistor.

Building the TeddyReg

The total voltage dropout should ideally be distributed equally between the linear regulator and the gyrator, with at least 2.5V on the LM317 (or 2V for the 1086).

The output voltage of the Linear Regulator (Vlr)is (R2/R1 + 1) x 1.25
The output Voltage of the gyrator is approximately Vlr x R4 / (R3+R4)

R1 = 250R

R2 = see formula above

R3, R13 = 100K

R4, R14 = see formula above

R5 = 100-150K

R6 = 200-2000R

R7 = 2-3K optional

C1 = 0.1uF X7R Ceramic (optional)

C2 = 10-20uF Tantalum

C3 = 10-20uF Tantalum

C4 = 22-33uF Tantalum

C5 = 0.1-0.3uF X7R Ceramic

C6 = 10-20uF see component choice below

C7 = optional see component choice below

Q1 = 2SK117GR

Q2 = D44H11

Q3 = ZTX651 or almost any mid power NPN bipolar

U1 = LM317/LT1086

R13, R14, and Q3 are the accelerator circuit and are optional.

As an example, to obtain 24V from a 33V rail, it is recommended to have 4.5V on the LM317 and 4.5V on the D44H11. To obtain 28.5V at the output of the LM317 R2 should be approximately 5450 Ohm. 
In order to reduce it to 24V by the gyrator, R4 should be approximately 526K.
The actual output voltage may vary in +/- 1V due to variations in the JFET Vgs

For the negative-rail version all capacitor polarities need to be reversed, U1 is LM337, Q1 is J74 or other p-channel (pay attention to pin layout), Q2 is D45H11, and Q3 is ZTX751 or any mid-power PNP transistor.

The Accelerator

The basic circuit has a relatively low startup time of several seconds. This is due to the time it takes to charge a 22-33uF capacitor through a 100K resistor. The accelerator circuit consists of a transistor and two resistors (Q3 R13, R14) which are equal in value to the voltage divider in the first low pass filter. The idea is that the transistor conducts as long as the capacitor is not charged. During this time it bypasses the 100K resistor and charges the capacitor. The current through the transistor is hfe times the current through the 100K resistor, which means it charges, depending on the transistor being used, about 200 times faster.

Component Choice

The choice of component is very important, especially in the gyrator part.

The Tantalum capacitor should be of high quality, I recommend using brands like Kemet, AVX, Panasonic, Vishay, etc. Low quality Tantalum capacitors tend to leak, causing performance degradation or even changes in output voltage. Mil-spec axial Tantalum capacitors perform very well in this position.

The Ceramic capacitor should be X7R or NPO/COg. NPO/COg are better but more difficult to source, especially in these values, SMT versions are more common. In any case X7R are good enough.

The capacitor at the output of the gyrator has an effect on the tonal balance and can be used for fine tuning according to your preference and the rest of your system. 10-20uF Tantalum will give a warm sound, a 1-10uF film is a good all rounder. A 0.01-0.1 will give very transparent and accurate sound. Combinations work as well 10uF Tantalum bypassed with 0.01-0.1uF film gives the best of both worlds.

The D44H11 is the best choice I've found for the output transistor. It has good power handling and works well at high frequencies. For low current and low dropout applications smaller transistors such as the ZTX651 (or ZTX751 for the negative version) can be used. For the JFET, transistors it is important to check that it can withstand the required voltage. Toshiba 2SK117 and 2SK170 are very good. The 2SK117 withstand higher voltage. The 2sk170 has lower noise and has a p-channel (j74) version for negative regulators. All versions (Y, GR, BL) can be used.

Credits and Links

This circuit was designed with the help of members of the PinkFishMedia DIY forum. Many thanks to those who contributed, tested, and provided feedback. More info can be found in the following PFM threads:





This circuit is published with the intention that DIY-ers can build this circuit for themselves or their friends for non-profit use. There is no patent or intellectual property but I expect that it will not be used to create commercial products competing with my own commercial products without my authorization.