Force/sense shunt

I'm a recent convert of the lowest possible impedance of power supply based on my experience of adding caps to my chipamp. So I figure the signal stages can't be harmed by reducing their supply impedance either, particularly at LF.

I note there are a few aftermarket regulators around - I had a look at Paul Hynes and Belleson in the past few days. They're a bit pricey for my tastes, given the cost of the components they're using can't be over single digit $ so I've had a look at "doing it at home, only cheaper".

First off, a simple TL431 is about the best bang for the buck achievable, as the part here is 0.2rmb. But the dynamic impedance is typically 0.2ohms and I was hoping and aiming to go a bit lower than this - perhaps an order of magnitude lower, to around 10mohms. Lower than this and the resistance of the PCB tracks come into play and its also very hard to maintain such a low impedance beyond the audio band as cap ESRs (for the best ones) are of the order of 10mohms. The TL431 in my sim pegs out around 10kHz, so above that the job becomes something for a cap (or cap bank).

Adding on a helper transistor gets the impedance into the right ballpark, and also provides the ability to do force/sense which helps nullify track and contact resistance. In the schematic the four 10mohm resistors represent the connector resistance as I'm adding the shunt on to an existing circuit. Even with 20mohms of connector resistance, the shunt still does a little better than 10mohm at LF - schematic and line rejection plot attached ( 100dB = 10mohm ). Parts cost is gratifyingly low for the performance - when fed with an LM317LZ configured as a current source the total is under 1rmb (<$0.20)

Here are a couple of shots of the first prototype - the active components from the top are the LM317L, BC807-40 and TL431. The PCB is 2mm pitch plated through hole proto board.

Update : It was a mistake to build these shunts with 0603 thick film resistors. They show considerable excess noise. It has a tell-tale quality, its pink in shape meaning equal power per decade of frequency, falling at 10dB/decade. Easy to spot this on an FFT. I am in the process of modifying them to use thin-film and 1206 thick film resistors (bigger sized ones don't exhibit nearly as much excess noise). Another thing to point out is that TL431s come in many different varieties and they're not all created equal. My naive assumption initially was that they'd be all very closely similar (as are, say NE5532s from different manufacturers). This is emphatically not the case - the CJ431s (Chinese) I have here are about 5X noisier than a sample LM431ACZ I tried which was in my parts collection from my UK days. A helpful DIYA poster has pointed out that Astec Semi makes the quietest ones, OnSemi looks to be almost as good. I am not sure if I can find OnSemi locally but I will have a go at finding them. TI's part is almost 3X noisier on paper than OnSemi's.

Attached Thumbnails