Author Topic: LM431 (TL431) 0.1-10 Hz Vp-p noise.  (Read 2418 times)

0 Members and 1 Guest are viewing this topic.

Offline MasterTTopic starter

  • Frequent Contributor
  • **
  • !
  • Posts: 851
  • Country: ca
LM431 (TL431) 0.1-10 Hz Vp-p noise.
« on: June 11, 2023, 12:51:07 am »
 I've been verifying noise level of the voltage references, and get puzzled by one part number from OnSemi. Here is the list:
AZ431 -1.2V/2.5V  Diodes Incorporated
TL431 - TI
TL1431 - ST
TL431 - ST
TLVH1431A - TI
TL431BCLPRMG
https://www.mouser.ca/ProductDetail/onsemi/TL431BCLPRMG?qs=xZq1yRCsb1eu65fxX1gBtQ%3D%3D

Others V.Ref compared with:
LM4040 - 2.5/3.0/4.096 - TI
REF6141 - TI

 All references shows more or less close data in the 0.1-10 Hz range, 5-10 uVp-p. Except TL431BCLPRMG, that have 3-5 times lower noise. Reading DS I notice  Figure 13. Spectral Noise Density, that is missing in DS from other makers. 

Does anyone else notice that not all 431 are the same?
« Last Edit: June 11, 2023, 01:11:16 am by MasterT »
 
The following users thanked this post: magic, Pulsepowerguy

Offline Pulsepowerguy

  • Regular Contributor
  • *
  • Posts: 52
  • Country: us
Re: LM431 (TL431) 0.1-10 Hz Vp-p noise.
« Reply #1 on: June 11, 2023, 01:07:06 am »
Unfortunately that Onsemi part is EOL. There are a few TI parts with noise data, and they appear to be way noisier than this one, however, their bias currents are 1mA whereas the Onsemi part's bias current is 10mA.
This is the best one I found:
https://www.ti.com/lit/ds/symlink/tlvh431b-ep.pdf?ts=1686444947389&ref_url=https%253A%252F%252Fwww.ti.com%252Fpower-management%252Fvoltage-reference%252Fshunt-voltage-reference%252Fproducts.html

« Last Edit: June 11, 2023, 01:10:11 am by Pulsepowerguy »
 

Offline magic

  • Super Contributor
  • ***
  • Posts: 7454
  • Country: pl
Re: LM431 (TL431) 0.1-10 Hz Vp-p noise.
« Reply #2 on: June 11, 2023, 08:36:25 am »
TI is the original maker of TL431 and they have a noise density plot in their datasheet: 220nV/rtHz at 10Hz.
Indeed, onsemi's plot looks much better. It seems to apply to all packages and accuracy grades.

Not sure what the remark about 10mA bias current means; all onsemi's DC specs are similar to TI's. Minimum cathode current for operation is 1mA (edit: noise spec is given at 10mA, as are several other specs, but I doubt that cathode current makes a difference for noise).

The fact that onsemi's broadband noise is twice lower than TI's suggest that they may be using more bias on some internal transistors and possibly different topology altogether. What I do known is that many generic TL431 from China, such as the fairly common "WS TL431" in TO92 package use completely different circuit topologies; dunno about their noise performance.


There is a zillion chips functionally similar to TL431 with confusingly similar names, but  meaningful differences: other reference voltage (1.24V is common and surely requires a very different internal circuit), lower cathode maximum rating, lower minimum current for regulation (these are likely to have sky high broadband noise, physics has no mercy). The link above is an example of such part, it is not a 100% compatible substitute.
« Last Edit: June 11, 2023, 08:55:59 am by magic »
 

Offline MasterTTopic starter

  • Frequent Contributor
  • **
  • !
  • Posts: 851
  • Country: ca
Re: LM431 (TL431) 0.1-10 Hz Vp-p noise.
« Reply #3 on: June 11, 2023, 02:18:32 pm »
TL431's topologies from TI and OnSemi is exactly the same , both depicted in DS.  Seems like some improvement was made in technological processes by OnSemi, but it's not documented.

 I see a big problem with Voltage References IC documentation & standardization  in general, each manufacturer specified IC in different way - many omits spectrum charts.

 Having V.Ref with 10 uVp-p noise is only good up to 18-bits, what about 24 ? 
TL431 was developed in the 12-16 bits era, and it works well, but times changed. I was tested couple more IC REF5025 and MAX6164 - both show no differences with TL431 from 0.1-10 Hz noise perspective. 

There is some parts from AD in the 1-2 uVp-p range, that  insanely overpriced  and still at 20-bits level

 Is it a gap in the education, that nobody even understand that there is no V.ref. for 24-bits ADC on a market at all? 

 
 

Offline magic

  • Super Contributor
  • ***
  • Posts: 7454
  • Country: pl
Re: LM431 (TL431) 0.1-10 Hz Vp-p noise.
« Reply #4 on: June 11, 2023, 03:02:04 pm »
Bandgap references are inherently noisy, as they amplify the noise of two transistors usually ~10x per each 1.2V of output voltage. And each halving this factor squares the area occupied by reference transistors, so it gets impractical fast. LT made some expensive and power hungry bandgap references with markedly lower noise, maybe these were the AD parts that you found.

Flicker noise can be improved by fab process, but wideband noise (which is also lower for onsemi) is often bottlenecked by collector current and can only be reduced by cranking up the bias. Not 100% sure what limits TL431 performance, although I feel like it's too noisy for base spreading resistance (the other limit on wideband noise) to be the dominant factor.

Resistor values in TI datasheet are known to be bogus, i.e. not representative of actual TL431 produced by TI. Resistor values in onsemi datasheet are the same as TI, so ::)
 
The following users thanked this post: thm_w

Offline dobsonr741

  • Frequent Contributor
  • **
  • Posts: 706
  • Country: us
Re: LM431 (TL431) 0.1-10 Hz Vp-p noise.
« Reply #5 on: June 11, 2023, 06:43:57 pm »
I do not see any of the data sheets showing the 0.1-10Hz range for "Equivalent Input Noise Voltage vs Frequency". The lowest is 10Hz. Both on Ti and OnSemi. Can you point to your refs?

The REF6141 easily beats both 431s, and a heavier weight, like LTC6655 is even better. Agin, no-one shows below 10Hz on my data sheets.
 

Offline MasterTTopic starter

  • Frequent Contributor
  • **
  • !
  • Posts: 851
  • Country: ca
Re: LM431 (TL431) 0.1-10 Hz Vp-p noise.
« Reply #6 on: June 11, 2023, 07:16:35 pm »
I'm not quite understand question. If you are referring to first message:
" All references shows more or less close data in the 0.1-10 Hz range, 5-10 uVp-p."
than I was talking about experimental data measured with my new toy ad7172.

 Basically, this is exactly a problem I 'd like to bring attention, spectrum charts are not in the same freq. band or missing in the DS.
I find adr3225 shows start 0.1Hz, LM4040 - 1Hz, REF50xx & MAX616X - no data.

REF6141 - in my test setup shows good results compare to 431 from ST and TI (noisy version) but worse than OnSemi's. I used huge 22uF NR cap in both cases, with REF6141 and TL431B. My understanding, that REF6141 is kind of from different leage, specifically for fast SAR ADC - not that I'm researching SD ADC

 
 

Offline dobsonr741

  • Frequent Contributor
  • **
  • Posts: 706
  • Country: us
Re: LM431 (TL431) 0.1-10 Hz Vp-p noise.
« Reply #7 on: June 11, 2023, 10:31:34 pm »
I thought you are only looking at data sheet data and not your own measurements.

The 1/f noise has always been back magic for me. If you are chasing the lowest possible noise reference here is one, now removed article on xdevs: https://web.archive.org/web/20220528043957if_/https://xdevs.com/doc/_Metrology/ciofi2007.pdf
 
The following users thanked this post: MegaVolt

Offline iMo

  • Super Contributor
  • ***
  • Posts: 5570
  • Country: va
Re: LM431 (TL431) 0.1-10 Hz Vp-p noise.
« Reply #8 on: June 13, 2023, 03:32:07 pm »
..
.. Having V.Ref with 10 uVp-p noise is only good up to 18-bits, what about 24 ? 
TL431 was developed in the 12-16 bits era, and it works well, but times changed. I was tested couple more IC REF5025 and MAX6164 - both show no differences with TL431 from 0.1-10 Hz noise perspective. ..
The 431 ref was never intended for ADCs, like 12-16bit.
It is a cheap general purpose ref with 50+ppm/C TC (and long term stability??).
It has been widely used in all kind of power sources where having voltage with 3 digits precision is all you need.
The 12-16bit ADC/DAC era references are the buried zener based like 1021/399 and all the bandgap ones like REF5025 and MAX6164, etc.
I doubt there is a reference for 24bit available today, except perhaps the LTZ1000/ADR1000..

A "reference for 24bit ADC" means the Vref fluctuation (thermal, 1/f, other noise, long term stability, hysteresis, etc) is (well) below 1 LSB (0.15uV in your case, imho).
Readers discretion is advised..
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 15157
  • Country: de
Re: LM431 (TL431) 0.1-10 Hz Vp-p noise.
« Reply #9 on: June 13, 2023, 03:48:54 pm »
AFAIK the TL431 only comes in a plastic case and for this reason along is prone to drift from humidity. It is not made for high performance, more like low cost for things like SMPS or similar.
It may be OK for lower resolution ADCs (e.g. 8-10 bit inside a µC).

The max6164 is a low power version and for this reason a bit more noisy. Low noise naturally comes with a higher suppler current.

For the noise it also depends on the application: in a DMM one cares about the drift and very low frequency noise like 1mHz - 1 Hz.
For an audio application it is more the low end of the audio range (the upper part is easy to filter) and rarely anything below 1 Hz.

For the very low frequency part the type of noise may also make a difference there can be more random walk like noise and discrete step pop-corn type noise. This can make a difference. The spectrum alone may not tell the difference. This is why some also show time domain noise.  Especially the 1/f part of the noise can also vary between units quite a bit and this part is hardly tested in production as the test naturally takes a lot of time.
 

Offline MasterTTopic starter

  • Frequent Contributor
  • **
  • !
  • Posts: 851
  • Country: ca
Re: LM431 (TL431) 0.1-10 Hz Vp-p noise.
« Reply #10 on: June 13, 2023, 04:33:58 pm »
A "reference for 24bit ADC" means the Vref fluctuation (thermal, 1/f, other noise, long term stability, hysteresis, etc) is (well) below 1 LSB (0.15uV in your case, imho).
Not quite, noise 0.15uVp-p in sub 10 Hz range, this is why I'm saying "NO single Voltage Reference on a market for 24-bits ADC".
 I'm considering one parameter only, it's written in the topic's name, so  /aging /temp. stability /long-term / humidity - not to discuss, please.

 I start this research during evaluation   SD ADC from AD, and explain in another thread that there is No solution to get noise below 0.15 nVp-p
except using simple RC network filtering . You may continue to critisize this , but w/o offering alternative to RC filter, words would not have any sense to me.

 TL431B is evaluated for one parameter only, and it's works exceptionally/ perfectly well - my SD ADC running almost for a week now

 Comments to image:  3 TL431B in the test, one is connected to adc V.Ref input, two others - to two diff adc inputs.
Filtering is swithed off for all 3 at the moment of capture image, but I also tested with 3 identical (1500 uF + 10k) filters .
Wiggle lines (about 4 uV) similarity on both channels proves that I have a problem not with the filters but with variation on the input current at Ref. Input of the ADC. Strange things, DS says +-100 nA but measurements imply +- 3-5 nA absolute values, with 4 uV / 10k = 0.4 nA bumps
 

Offline iMo

  • Super Contributor
  • ***
  • Posts: 5570
  • Country: va
Re: LM431 (TL431) 0.1-10 Hz Vp-p noise.
« Reply #11 on: June 13, 2023, 05:00:01 pm »
With such a cheapo reference like the 431 you may put, say, 100 of them in parallel and the noise drops 10x, afaik..  :D
Readers discretion is advised..
 

Offline MasterTTopic starter

  • Frequent Contributor
  • **
  • !
  • Posts: 851
  • Country: ca
Re: LM431 (TL431) 0.1-10 Hz Vp-p noise.
« Reply #12 on: June 13, 2023, 08:46:02 pm »
With such a cheapo reference like the 431 you may put, say, 100 of them in parallel and the noise drops 10x, afaik..  :D

You mean to build a "Godzilla" reference with 1 A current?  ::)

 I'm weighting costs of small auxiliary SD ADC to monitor those current bumps on reference inputs lines. To create "dual core" ADC, where main unit is sampling 2 channels w/o been disturbed by switching MUX to those lines, as I thought earlier. Same concept as ads1263 - can't get one for a year, out of stock.
 

Offline dobsonr741

  • Frequent Contributor
  • **
  • Posts: 706
  • Country: us
Re: LM431 (TL431) 0.1-10 Hz Vp-p noise.
« Reply #13 on: June 13, 2023, 08:59:55 pm »
I had good experience with the Zonri ADS1263 modules of CN origin. After replacing the sw reg on them.
 

Offline iMo

  • Super Contributor
  • ***
  • Posts: 5570
  • Country: va
Re: LM431 (TL431) 0.1-10 Hz Vp-p noise.
« Reply #14 on: June 14, 2023, 08:34:38 am »
With such a cheapo reference like the 431 you may put, say, 100 of them in parallel and the noise drops 10x, afaik..  :D

You mean to build a "Godzilla" reference with 1 A current?  ::)

There are the LM385-2.5 references, running well from 20uA current. That would be 2mA with 100pieces.. :)
Readers discretion is advised..
 

Offline magic

  • Super Contributor
  • ***
  • Posts: 7454
  • Country: pl
Re: LM431 (TL431) 0.1-10 Hz Vp-p noise.
« Reply #15 on: June 14, 2023, 09:03:25 am »
That's 1100nV/rtHz at 10Hz, so 25 of them will match TI TL431 and about 300 will match onsemi.

I don't know why onsemi is more efficient than TI, but as a general rule, noise is determined by the combined current of all references, not by how much you split that current into multiple chips ;)
 

Offline iMo

  • Super Contributor
  • ***
  • Posts: 5570
  • Country: va
Re: LM431 (TL431) 0.1-10 Hz Vp-p noise.
« Reply #16 on: June 14, 2023, 09:11:45 am »
Sure, each reference with its own cathode resistor and combined via a summing resistors. As it is done with 399 when more pieces in parallel..
Thus 100 references, and 200 resistors (100*125k cathode ones and ie. 100*10k summing ones with 5V power)  :phew:
The total current will be ~2mA.

PS: those 1150nV/rtHz is for 100uA, thus with 20uA it will be even more. Anyhow, simply scale it up to 1000 pieces, still only 20mA total current  :D :D
« Last Edit: June 14, 2023, 09:45:12 am by iMo »
Readers discretion is advised..
 

Offline magic

  • Super Contributor
  • ***
  • Posts: 7454
  • Country: pl
Re: LM431 (TL431) 0.1-10 Hz Vp-p noise.
« Reply #17 on: June 14, 2023, 12:56:18 pm »
PS: those 1150nV/rtHz is for 100uA, thus with 20uA it will be even more.
You can try, but I doubt it.

The first 10μA (LM385) or 400μA (TL431) flows through the bandgap reference, the rest is bypassed by a power transistor in parallel with it.
Throwing more current at the chip increases power transistor bias and may reduce impedance and improve dynamic response, but I wouldn't expect much difference in noise.
They all ought to be OK at their minimum operating bias.
 

Offline jmelson

  • Super Contributor
  • ***
  • Posts: 2852
  • Country: us
Re: LM431 (TL431) 0.1-10 Hz Vp-p noise.
« Reply #18 on: June 14, 2023, 07:30:15 pm »
A really good low-noise reference is the LM4140.  Available with a few output voltages, noise is less when run from less than 4.5V, and you have to put about a 1 mA load on the output.  There might be even better ones now, but we found this one to work really WELL in a noise-sensitive application.
Jon
 

Offline MasterTTopic starter

  • Frequent Contributor
  • **
  • !
  • Posts: 851
  • Country: ca
Re: LM431 (TL431) 0.1-10 Hz Vp-p noise.
« Reply #19 on: June 14, 2023, 07:39:37 pm »
VN Output noise voltage(7) 0.1 Hz to 10 Hz 2.2 μVP
Charts 11 and 12 show about 200nV/sqrt(Hz), compare to 48nV TL431B
 

Offline GigaJoe

  • Frequent Contributor
  • **
  • Posts: 524
  • Country: ca
Re: LM431 (TL431) 0.1-10 Hz Vp-p noise.
« Reply #20 on: August 29, 2023, 05:28:01 am »
assuming both bandgap type ,
lm4140 - consumed 0.2 mA ;  it series ,  I assume fraction to bandgap feeding
tl431 -  can be 0.4 mA to 0.6mA - i assume higher current for bandgap.
AS431 (diodes)  - 1mA

lm4140 noise as 2.2uV p-p defined for lowest 1.024 Vout and it less then native bandgap voltage output
tl431 - 2.5V out and 2X bandgap as well 2X internal noise of bandgap

may i guess - level of current consumption may indirectly define noise output ? 

btw , comparing lm4140 noise with output 1.024 ,and something else with output 5V , would be incorrect ,  as per 4140  5V out actual Vp-p   = 2.2 X (5 / 1.024)

 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf