wenzel associates finesse shunt

[kp93300]

Hi, more experienced diyer,

Please see the attached images.

The circuit is from Wenzel associates .
Finesse Voltage Regulator Noise!

I build the shunt hoping to clean up a 24 V dc supply for my pre.
The 2n 4401 becomes too hot to touch after 5 minutes. However, there is no problem if i powered with 12 V dc.
I have checked the spec of 2n 4401 and is within operating range.
What are the values to change to prevent overheating of the 2n 4401 ?

thanks

kp93300

[DF96]

Does the transistor need a heatsink? The power dissipation given in the datasheet will probably assume a maximum case temperature. Or is it oscillating?

Better decoupling/smoothing might be a better solution - much simpler!

[agdr]

If it were me I would switch to a BD139-16, then heat sink it like DF96 said.

http://www.fairchildsemi.com/ds/BD/BD139.pd

[agdr]

...and you need to change the 4300R resistor to 8450R if you are going to feed it with 24V rather than 15V and want to preserve the 18.7mA idle current through the shunt it seems to be designed for. With the 4300R bias resistor you get 40mA through the shunt at 24V, which equates to 14.5V across the transistor and Pd = 580mW, right at the heat-sinked max for that 2n4401.

How much load current do you intend to pull at the 24V? The values seem to be set up for an expected load of around 100-200mA.

[kp93300]

I build this to try and improve the performance of this amp from JUMA.
I intend to replace the cap multiplier before the j310 .
I have no idea what is the power requirement , probably less than 100ma.
thanks for the useful suggestions , i will try the suggestions here.
thanks

kp93300

[kp93300]

I inserted the shunt as per schematic above into a dac power supply and the improvement is obvious.

I am building a second shunt with an input of 24V dc replacing 2n 4401 with the bd 39 and 4300R with 8450 R.
Are there any other changes that is necessary ?

I do not have a scope to adjust these values.

thanks again.

kp93300

[agdr]

Something to experiment with would be doubling the series resistor to 0.1 ohm, then adjusting the op amp gain according to Wenzel's formula:

The values are not critical except that the gain of the amplifier should be very near the ratio of the transistor emitter resistor to the series shunt resistor.

Which means the 300k non-inverting op amp feedback resistor changes to 150k 1% to lower the gain from 300 to 150 (15R/0.1R=150).

As it sits the 18.7mA through the shunt subsequently causes a (18.7mA)(0.05R) = 1mV drop across the series resistor at idle. Making the change above would double that to 2mV and potentially provide more noise cancellation ability.

He also talks about the possibility of making a gain resistor variable to allow precise nulling. I don't know if that will work as well as Wenzel states, but the resistor to try that with would be the 1K between the op amp inverting input and the 100uF cap. Maybe try something like a 200 ohm 20-turn trimmer in series with 900 ohms of fixed resistance to allow +/- 100 ohms of variance. Center the trimmer before you hook it up, of course. That would vary the gain between 137 and 168. Or for more precision a 100R trimmer in series with a 953R fixed, gain 142-157.

[agdr]

I forgot to say - the 8450R resistor is for the 2n4401 version. The transistor power dissipation drops to around 400mW and you would be OK with a heatsink. For the BD139-16 you would need to use a 13k bias resistor. The BD139-16 should be OK with no heatsink at just 400mW, although I would still probably use a clip-on (TO-126).

You may want to experiment with replacing the 100uF caps on the output of the op amp and from the 1000R to ground with 220uF to give lower impedance at low frequencies. As Wenzel says-

low frequency performance is impacted by the size of the coupling capacitors

The LM4562 at $3 USD (same as LME49720, or LME49710 for single) is a newer part that beats the $1 LM833 in most specs but quiescent current, including noise and PSRR. I doubt if any of the improvements would even be noticable, but for $2 more it might be worth a try. Wenzel:

noise floor is limited by the performance of the LM833 and resistor noise

[kp93300]

thanks agdr,

i will do the changes as suggested by you . I have populated 1/2 of parts on the breadboard this morning.
To increase the idle current through the series resistor, i am contemplating placing a 1K R from + to ground at the output of the shunt.
Any comments ?

Can you estimate the current through the j310 ?

My electronic knowledge is very basic.

much appreciated.
regards
kp93300 .

[agdr]

I breadboarded one today too.

I think you are OK without any extra load on the shunt. It appears that the circuit AC couples the op amp to the transistor, so only the noise or other signal riding on the input DC gets shunted, regardless of the current draw of your load. The more load current though, the more drop across the shunt resistor of course, and hence some loss of voltage level regulation.

That is an LME49740 in the picture, the quad version of the LME49710, using 1/4 of it. The math almost matched reality - I had to go up to a 9.1k bias resistor instead of the 8.45k to get 18.8mA shunt current through the 220 ohm resistor. I haven't tried it with a BD139 yet, I'll do that this weekend. That 2n4401 definitely needs a heat sink at the 400mW - starts to slowly drift without.

If you are learning electronics you might have some fun figuring out the DC bias point of the op amp. At DC the caps are all opens, of course, so no DC can flow that way (in reality there is a tiny leakage current, but ignore that now). Then apply the two laws of op amps: (1) the input voltages are forced to be the same by the feedback, and (2) the inputs draw no current, meaning all currents at each input node sum to zero. The inputs really do draw a small bias current and have offset voltage, but ignore those here.

The 10K voltage divider applies half the incoming voltage to the non inverting input - 12V. Any AC riding on the DC, like 2mV of noise, also gets cut in half. By (1) the non-inverting input must also be at 12V. With the caps open the only path to ground for current is through the 300K resistor (I've built it up here with the 300K and the 0.05R shunt). Since there is 19mA through the transistor emitter resistor, the 300k resistor returns to (0.019)(15) = 0.28V. Assume here the emitter current is much larger than the current through the 300k. So the current in the 300K resistor is (12V - 0.28V)/300k = 39uA.

Since no current can go into or out of the inverting op amp input by (2), all that current must come through the 47K+1K resistors to the op amp output. That gives a drop across those resistors of (39uA)(48k) = 1.88V. Now add that to the 12V at the inverting input and you get the op amp output DC bias point, (12V + 1.88) = 13.88v.

On your J310 - from the datasheet is looks like the maximum drain current of that jfet is 24mA - 50mA. so your circuit would pull no more than that. You are probably OK with the shunt as it is.

You may want to take a look at this review article of the Wenzel shunt:

Reducing Power Supply Ripple and Noise

They were not impressed. The conclusion: "the attenuation at the null point changes over 30 dB for a 3% change. Since the same change could easily occur with normal temperature variations, we conclude this circuit is too unstable to be very useful."