Purpose of precision high-current series regulator's resistors and capacitor?

?

Hi, Don:-

There is no "the" TL431 datasheet, as there are multiple versions from TI a nd at least 8 manufacturers (and probably a few more in China).

But presumably this is the schematic:

The 4.7K is to keep the output 2N222 from amplifying the leakage in the other 2N222.

The other parts are for stability.The 30 ohm for keeping the Darlington from going all Colpitts on us and the cap to keep the TL431 happy.

I'm sure someone will point out if this explanation is lacking or incorrect in some way. ;-)

I'm not sure the 4.7K and 30 ohm are actually necessary.

Best regards, Spehro Pefhany

at least 8 manufacturers (and probably a few more in China).

It may be worthy of a simulation while we wait for the critics.

Thank you,

--
Don Kuenz KB7RPU 
There was a young lady named Bright Whose speed was far faster than light; 
She set out one day In a relative way And returned on the previous night.

speff wrote:

The url is

No, they are not. In fact, the first 2N2222 (emitter follower) is not necessary. The pass transistor can supply only a limited current, perhaps

50 mA at 80C, so a darlington is not needed.

The TL431 uses a bandgap reference and is extremely noisy, around

450nV/root Hz at 10 Hz, with a pronounced peak just above 10 KHz.

However, Fig 29 can be modified to bring the noise down to 3.3nV/root Hz at

1 Hz, and 0.6nV/root Hz at 10 Hz. Here is a 5V regulator in LTspice:

Version 4 SHEET 1 1608 1700 WIRE -464 48 -480 48 WIRE -352 48 -464 48 WIRE 96 48 -352 48 WIRE -480 64 -480 48 WIRE -352 64 -352 48 WIRE 96 112 96 48 WIRE -480 160 -480 144 WIRE -352 160 -352 144 WIRE -320 160 -352 160 WIRE -288 160 -320 160 WIRE -176 160 -208 160 WIRE -96 160 -176 160 WIRE -16 160 -96 160 WIRE 32 160 -16 160 WIRE -176 176 -176 160 WIRE -96 176 -96 160 WIRE -16 176 -16 160 WIRE -176 256 -176 240 WIRE -96 256 -96 240 WIRE -16 256 -16 240 WIRE -352 288 -352 160 WIRE -320 288 -352 288 WIRE -240 288 -256 288 WIRE -352 320 -352 288 WIRE -304 352 -320 352 WIRE -240 352 -240 288 WIRE -240 352 -304 352 WIRE -176 352 -240 352 WIRE -112 352 -176 352 WIRE 96 352 96 208 WIRE 96 352 -32 352 WIRE 144 352 96 352 WIRE 160 352 144 352 WIRE 96 368 96 352 WIRE -176 384 -176 352 WIRE -352 400 -352 384 WIRE 96 464 96 448 WIRE -176 480 -176 464 FLAG -352 400 0 FLAG 96 464 0 FLAG -16 160 Q1B FLAG 144 352 Vout FLAG -304 352 U1B FLAG -464 48 VCC FLAG -176 480 0 FLAG -480 160 0 FLAG -320 160 U1C FLAG -176 256 0 FLAG -16 256 0 FLAG -96 256 0 SYMBOL npn 32 112 R0 SYMATTR InstName Q1 SYMATTR Value 2N3904 SYMBOL res 80 352 R0 SYMATTR InstName R1 SYMATTR Value 470 SYMBOL cap -192 176 R0 SYMATTR InstName C1 SYMATTR Value 1000uf SYMATTR SpiceLine Rser=12m Lser=10n Rpar=3e6 Cpar=3e-9 SYMBOL voltage -480 48 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V1 SYMATTR Value 10 SYMBOL res -336 160 R180 WINDOW 0 36 76 Left 2 WINDOW 3 36 40 Left 2 SYMATTR InstName R2 SYMATTR Value 1k SYMBOL res -192 368 R0 SYMATTR InstName R3 SYMATTR Value 10k SYMBOL res -16 336 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R4 SYMATTR Value 10k SYMBOL res -192 144 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R5 SYMATTR Value 10k SYMBOL TL431AS -352 352 M0 SYMATTR InstName X1 SYMBOL cap -256 272 R90 WINDOW 0 0 32 VBottom 2 WINDOW 3 32 32 VTop 2 SYMATTR InstName C2 SYMATTR Value 1000uf SYMATTR SpiceLine Rser=12m Lser=10n Rpar=3e6 Cpar=3e-9 SYMBOL cap -112 176 R0 SYMATTR InstName C3 SYMATTR Value 1000uf SYMATTR SpiceLine Rser=12m Lser=10n Rpar=3e6 Cpar=3e-9 SYMBOL cap -32 176 R0 SYMATTR InstName C4 SYMATTR Value 1000uf SYMATTR SpiceLine Rser=12m Lser=10n Rpar=3e6 Cpar=3e-9 TEXT -488 -64 Left 2 ;'TL431 Noise Spectrum TEXT -88 -32 Left 2 !.lib TL431AS.lib TEXT -488 -32 Left 2 !.noise V(vout) V1 oct 1000 1 1e6

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

{ Npanes: 1 { traces: 1 {524290,0,"V(onoise)"} X: ('M',0,1,0,1e+006) Y[0]: ('n',1,5e-010,5e-010,6.5e-009) Y[1]: ('_',0,1e+308,0,-1e+308)

Log: 1 0 0 GridStyle: 1 } } [AC Analysis] { Npanes: 1 { traces: 1 {524290,0,"V(vout)"} X: ('M',1,0.001,0,1e+007) Y[0]: (' ',0,1.99526231496888e-006,6,0.00398107170553497) Y[1]: (' ',0,80,20,280) Volts: (' ',0,0,5,2.500764,6e-006,2.500842) Amps: ('m',0,0,4,0.01000308,2e-008,0.01000334) Log: 1 2 0 GridStyle: 1 PltMag: 1 } } [Transient Analysis] { Npanes: 1 { traces: 1 {589827,0,"V(vout)"} X: (' ',1,0,0.2,2) Y[0]: (' ',4,4.9938,0.0001,4.9952) Y[1]: ('m',4,1e+308,3e-007,-1e+308) Volts: (' ',0,0,4,4.9938,0.0001,4.9952) Log: 0 0 0 GridStyle: 1 PltMag: 1 PltPhi: 1 0 } }

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

• TL431AS.lib

• Models have been edited for LTspice syntax, but are otherwise per originals

• Model developed by analogspiceman

************************************************** .subckt TL431AS A K R ; Anode Kathode Reference D1 R K Dc R1 A R 1.3e6 tc1=3m G1 A 5 R 4 1 C1 5 A 1n Rpar=600 D2 7 R Dc D3 7 4 Dn R2 7 4 950k R3 6 5 600k D4 A 6 Di C2 K 6 40p Rser=5k Rpar=1e6 G2 K 7 6 A 95m D5 A 7 2V5 D6 7 A Dk C3 7 A 50p D7 A K Dc .model Dc d Ron=10 Vfwd=0.65 Vrev=36 Epsilon=50m .model Dn d Is=1p Kf=0p2 Cjo=0p3 .model Di d Ron=10m epsilon=1m .model Dk d Ron=5k7 Vfwd=0.8 Epsilon=0.5 .model 2V5 d Ron=27m Vfwd=0.6 Epsilon=10m Vrev={Vr} revEpsilon=10m .param Vr=2.5+dt*(95u-dt*2u6) dt=temp-10 .ends TL431AS

Forgot the ASY and some other files. That's too many to handle, so here's the whole thing zipped:

This uses Helmut's model, and produces 2 nV/root Hz at 1 cycle.

Hm. I think you might not really get that performance if you build it. Base spreading resistance might be a problem and you might need at least some special transistors to fix that.

Yep, much better parts in Table 8.1 in AoE3.

Also, the circuit model needs to be stable with an output capacitor.

>

--
 Thanks, 
    - Win

Adding 50 ohms in series with the Q1 base increases the noise to 2.2 nV/Hz^

1/2. Base spreading resistance is not a problem.

The 2N3904 is a surprisingly good transistor. It also has an entry in both LTspice IV and XVII

With Spice Models? What's wrong with 0.6 nV/Hz^1/2 at 10 Hz?

Adding a 1000uF from Vout to gnd drops the noise from 0.6 nV/Hz^1/2 to 0.02 nV/Hz^1/2 past 10 KHz.

The capacitors have 12 milliohms series resistance and 10 nH series inductance.

The transient response shows 209uV overshoot with a 0.1V step at the input. It damps out in 2 cycles. The output voltage settles 80uV higher.

Wasn't the 2N222 obsolete in 1970? :)

My 1974 Archer _Transistor Substituti RB pinches off the base of the final transistor and makes the Darlington configuration faster and more linear. RA is called a base stopper and destroys the Q of what could otherwise be a resonant circuit due to combination of the output load on the regulator, interelectrode capacitances in the transistors, and the output gain. Emitter followers are quite susceptible to oscillation. C1 adds local feedback to the TL431 to reduce its gain long before the phase shift of the output transistors can cause a positive feedback situation.

Basically, those three components all are there to give stability. Depending on the type of load, they may not be needed. The TL431 by itself is a pretty slow device, but the writers of the app note probably wanted to protect against a worst-case scenario.

BTW, there are lots of TL431 variations from different manufacturers. Just because it's marked TL431 doesn't mean it is anything like one from TI. Including the pinout :)

Thank you,

--
Don Kuenz KB7RPU 
There was a young lady named Bright Whose speed was far faster than light; 
She set out one day In a relative way And returned on the previous night.

Try adding the missing '2', 2N2222.

SPICE models cannot nay-say bench measurements. As our table shows, we measured 1.35 nV/rt-Hz for 2N3904 at 10mA; worse at lower currents.** That's with Rs=0, r_bb'=110 ohms. With many manufacturers of jelly-bean transistors, you will get noisier ones (we used Fairchild and On Semi). Our table has 30 better choices, and the manufacturer and process are more dependable. Paul and I spent many many months buying transistors, making laborious measurements, analyzing and presenting the data. So make good use of it. Don't use SPICE models to deny reality.

** Much worse at 10Hz.

--
 Thanks, 
    - Win

Of course. This is perfectly obvious.

However, you are measuring the transistors alone. The LTspice simulation uses the transistor in a feedback network with heavy filtering.

The TL431 model developed by Helmut specifically includes flicker and broadband noise. This is low-pass filtered before it reaches the transistor. The transistor output is fed back to the TL431. The transistor model is supplied by NXP, which you do not use. There is no way to compare the simulation results with your measurements.

Adding a 1000uF cap from Vout to gnd significantly reduces the noise. You do not add capacitors to filter the output of your measurements.

The simulation is essential to tell you if a circuit is worth spending time on the bench. It is invaluable to help troubleshooting when something goes wrong on the bench. It allows you to evaluate different component parameters and see which ones are critical to obtaining the desired results. It allows you to optimize the circuit and components for best performance. You can examine the open-loop response in a feedback circuit and optimize the loop gain, and gain and phase margin. This can be difficult or impossible to do on the bench.

Bench work only tells you the results of one particular set of components. It can be difficult and time-consuming to change components. You may have to fight noise, poor connections, unwanted oscillations, test equipment limitations, bad components, and a host of other problems that interfere with your evaluation. You may lack the equipment needed to make critical measurements that you can easily do in simulation.

Don't deny the value of simulation.

What is the change?

Or you could use the 80 pV/rt-Hz circuit in AoE 3.

Have you made any measurements yet? It's possible that your SPICE model may be completely wrong, and there will be no need for any special low-noise preamp to take the measurement.

--
 Thanks, 
    - Win

See the zip file

That is rediculous. It is intended for ribbon microphones, with a source resistance of 0.2 Ohms. It is also too noisy.

The noise is too low to measure. All the other measurements are satisfactory.

After 40 years running SPICE, it is more likely the results are highly accurate, and offer performance you could never dream of.

Hundreds of thousands of engineers have been taught SPICE in colleges and universities. How did you miss out?

Steve, did you realize whom you were replying to? Some major fraction of those hundreds of thousands have been taught out of AoE. I mean, a healthy ego is one thing....

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

Many others were taught ot of Terman, Radiation Labs, Radiotron Designer's Handbook, various engineering magazines, etc.

AOE discusses Intusoft ICAP, but never seems to use it anywhere in the thousands of opportunities throughout the text. This could have vastly increased the value of AOE, especially if LTspice were used instead of ICAP.

In this day and age, lack of skills in LTspice is a career-limiting move.

Those who learned from AOE are at a severe disadvantage.

I was looking to give you a way to climb down honourably, but if you want to double down instead, well, you do that.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

Chortle! Best smile of the day.

I truly fail to see what you are driving at. We are talking about careers and livelihood. You have posted a number of LTspice files. Could you do this work without LTspice?