Voltage reference in Jung regulators

Sigurd brought up the question or claim that LM431 is pretty OK. Some people wrinkle their noses but if you compare the data between LM431 and LM329, the 431 is better on most parameters except for tempco.

It seems also that Texas has the 431 more specified than National.

Notice that the dymanic impeance is much better and the noise is lower, and also the price.

I have used the 431 much in work and am very pleased with it just because it's very troublefree, has decent precision and cost little.

What do you say folks?

The discussion was started here

I missed this thread as I have ben busy with several audiorelated projects :D

as far my research is concerned I have found that ONSEMI's TL431 BI/BV is the best audio voltage reference around (when used as a normal Zener diode).

Note, it must be from OnSEMI and it must be the TL or BV version of the TL431.




Datasheet can be found at

OnSemi's TL431
Have you checked ADR42x from Analog Devices, not so bad either.

The incredible Invisus Super Regulator uses it.

Yes, I have checked the ADR42X (X = 0, 1, 3, 5), and they all
are not by far as good as TL431BI/BV from OnSemi.

Output impedance is almost a factor 10 better fo rthe TL431BI/BV.

Voltage noise is almost twice as good on the TL431.

See below:



if it's a "Reference" then the noise density should also be spelled out from 100 mHz to 10 Hz --

you have to measure itself -- TI has a useful setup on the TL431 PDF -- the entire assembly should be placed in a shielded container --

two cheap references can be sa good as one expensive one -- as "noise", if it is indeed "noise" is random -- will cancel.
we had the option to fit either a Zener voltage reference or, at a separate location fit an LM4040 as the reference.

These were feeding a discrete voltage regulator.

I argued for a TL431 option by adding a third pad. It would then allow all three options at the same location and save board space by eliminating the need for a separate Zener location that could not fit the two pin To92 pad layout. The freed space would now allow one resistor of the pair needed for an adjustable tl431 reference. We could even move the pot from the output side to the tl431 (reference) side and have a truely adjustable reference and no extra space requirement.

I did not have the knowledge to argue the advantages of the tl431. I could only argue the logic of the physical pad side of the proposed change. I gave up when three posters all thought I was being penny pinching for no technical advantage.
peranders said:
I think the price is often related to the stability rather than the noise. It's hard to make something stable but it's easy to remove the nosie.

noise is random -- so when we specify noise or the voltage of a reference we have to integrate voltage over time and over a specific bandwidty -- and this is what a capacitor or RC circuit does -- the integration. stability is like a "random walk" where the mean value will change over time -- over a long enough period of time it will revert to a mean value -- internal thermal relations, susceptibilty to heating or cooling due to ambient conditions etc.

noise isn't always an enemy.
I'm gearing up to modify an existing Jung regulator to the most current version, with preregulator, and turn it into a standalone high performance bench supply for projects. This thread on voltage references caught my attention; I'll likely be incorporating the TL431 into my updates.

I'm also wondering about incorporating the TL431 into the Leach amplifier's input stages' zener-regulated 40 volt supplies, but it's only rated to 37 volts max. I suppose I could try running the supplies within the part's specifications, but translating the TL431's performance to a much wider range of voltages might be useful if there's a relatively simple way to do it without compromising that performance. A mini-pcb containing all the parts to drop in place of the zener(s) would work neatly.

At any rate, I've ordered 10 of the OnSemi parts for experimentation.

Some further research turns up an Analog Devices AD588 voltage reference, 10 volts @ .01% tolerance(!) without trimming. Expensive too; the AD688 is half the price for a slight decrease in tolerance. All of my test bench meters need calibration sources, and this would exceed their specs, easily.