generate voltage reference

Thread Starter

k1ng 1337

Joined Sep 11, 2020
542
Hi,

The boost converter in the schematic is using a 1.25V reference generated by the LM317 (via external power) for the LM393. I would like to replace this reference with one capable of operating at 1.25-2.7V. I have tried a number of methods all of which suffer a voltage decline along with C1 which corresponds to a predictable variance in output voltage. There are many improvements to be made on the circuit though for the time being I would like to focus on generating a reference from scratch. I have reviewed methods for Zeners, charged capacitors and bandgap references, some of which are terribly complex. Any input is appreciated.

Regards,
Mark

boost2.png

edit: forgot to include R9
 
Last edited:

Ian0

Joined Aug 7, 2020
5,171
I'd suggest a LM4040 voltage reference. They will work correctly down to 60uA. Use a fixed 1.2V reference, and move the preset to replace R6 and R7.
 

DickCappels

Joined Aug 21, 2008
8,711
I have spent days, when I did not have access to a proper environmental test chamber, trying to improve on the stability of the TL431 (50 ppm/°C). A time-consuming learning experience. It turned that I could get a far superior off-the-shelf solution without much cost. I ended up using an LM4140 (3 ppm/°C) reference with a PWM adjustable amplifier and offset circuit to get a wide variable but very stable voltage.

Definitely use the TL431 with pots if it gives you the performance you need.
 

Audioguru again

Joined Oct 21, 2019
4,696
You already have an LM317 voltage regulator. It can become an adjustable voltage reference by adding a variable resistor from ADJ to GND to your existing circuit.
 

Thread Starter

k1ng 1337

Joined Sep 11, 2020
542
hi K1ng,
This is the TVL431
E
Thanks. Is there a place I can download all the extra components you guys use? Or do I have to add each like the TLC555?

I'd suggest a LM4040 voltage reference. They will work correctly down to 60uA. Use a fixed 1.2V reference, and move the preset to replace R6 and R7.
Wrong part? " The advanced design of the LM4040-N eliminates the need for an external stabilizing capacitor while ensuring stability with any capacitive load, thus making the LM4040-N easy to use. Further reducing design effort is the availability of several fixed reverse breakdown voltages: 2.048 V, 2.5 V, 3 V, 4.096 V, 5 V, 8.192 V, and 10 V. The minimum operating current increases from 60 μA for the 2.5-V LM4040-N to 100 μA for the 10-V LM4040-N. "

Also, I have replaced R6 and R7 as suggested and noted an output improvement. Thanks for that tip. I knew there was something fishy with the comparator among other disasters. I posted what was my most stable prototype and decided to go from there. What would a be good value for pull-up resistor R9?

I have spent days, when I did not have access to a proper environmental test chamber, trying to improve on the stability of the TL431 (50 ppm/°C). A time-consuming learning experience. It turned that I could get a far superior off-the-shelf solution without much cost. I ended up using an LM4140 (3 ppm/°C) reference with a PWM adjustable amplifier and offset circuit to get a wide variable but very stable voltage.

Definitely use the TL431 with pots if it gives you the performance you need.
Interesting method. How did the amplifier stage affect the precision of the LM4140?

The TL431 has a 2.5V ref. There are other parts that have a lower voltage ref. TL1004.
TL1004 didn't turn up a result?

You already have an LM317 voltage regulator. It can become an adjustable voltage reference by adding a variable resistor from ADJ to GND to your existing circuit.
I've done that. The reviews for using an LM317 as a voltage reference are not good though..

And finally to everyone: As usual, the consensus is to buy a part instead of building it. For scholastic purposes I would like to try generating a voltage reference from basic components. It is clear any design I come up will be lacking in many regards. Part of my interest in this project is identifying shortcomings between circuits and how to make up for them by addressing each component individually.
 

AnalogKid

Joined Aug 1, 2013
10,067
A small signal diode, like the ultra-common 1N914 or 1N4148, can be used as a simple 0.6 V voltage reference. However, it has a very well documented temperature coefficient. If you want to go from scratch, use 1 diode as a reference, and another in a temperature compensation opamp circuit.
 

crutschow

Joined Mar 14, 2008
29,825
I would like to try generating a voltage reference from basic components.
It can be difficult to generate a voltage reference stable with temperature from discrete components.
What value of temperature stability are you expecting?

Zeners with a breakdown voltage of around 5V can be fairly stable with temperature (below).
Do you consider that a discrete component?

1648355611681.png
 

Ian0

Joined Aug 7, 2020
5,171
Why is U4 connected to pin 5 of the 555?
If you want to switch the 555 on and off, then use pin 4 (reset)
If you want to modulate the pulse width, then don't use a comparator, because its output can only be low or high, and the CV pin does not work with its voltage at the supply rails.
 

Thread Starter

k1ng 1337

Joined Sep 11, 2020
542
Updated circuit:
boost3.png

A small signal diode, like the ultra-common 1N914 or 1N4148, can be used as a simple 0.6 V voltage reference. However, it has a very well documented temperature coefficient. If you want to go from scratch, use 1 diode as a reference, and another in a temperature compensation opamp circuit.
Some averaged readings, I'll call the anode of D2 VREF. The voltage at this node likes to jump around under test.

for V+ = 1.25V
VREF = 520mV
Voltage at OUT = 3.3 (desired)

for V+ = 2.70V
VREF = 560mV
Voltage at OUT = 3.6V

During my trials I noticed I can significantly heat the diode with my hand and so the voltage across it drops. I didn't think it was this pronounced. If the circuit is kept at room temperature and draws a slowly degrading but steady current, how will the diode respond'? It is clear I need a PTC circuit to compensate but first I would like to understand the factors that affect NTC for a lone diode and resistor. With a variance of 0.3V as seen above, the reference at the non-inverting terminal is operating *not bad*.

It can be difficult to generate a voltage reference stable with temperature from discrete components.
What value of temperature stability are you expecting?

Zeners with a breakdown voltage of around 5V can be fairly stable with temperature (below).
Do you consider that a discrete component?

View attachment 263709
The lowest value Zener I have on hand is 2.9V. I do have plans to try out where the two effects cancel at 5V. I've been trying to make it work with a basic silicon diode with limited success. Concerning my schematic, what value for R4 is ideal as input to U4 with the least current draw possible to maintain regulation?

Why is U4 connected to pin 5 of the 555?
If you want to switch the 555 on and off, then use pin 4 (reset)
If you want to modulate the pulse width, then don't use a comparator, because its output can only be low or high, and the CV pin does not work with its voltage at the supply rails.
I have been revising a basic feedback circuit I found in internet land as real feedback circuits are currently discombobulating. What would you suggest in place of LM393? I assumed pin 5 would see an average voltage due to rapid switching. If this is not the case, why is regulation fairly tight at OUT providing RFEF is stable? I'm using basic test equipment so who knows what kind of noise is really going on. My analog/digital clock as a load doesn't seem to mind and keeps time.
 
Last edited:

Ian0

Joined Aug 7, 2020
5,171
I have been revising a basic feedback circuit I found in internet land as real feedback circuits are currently discombobulating. What would you suggest in place of LM393? I assumed pin 5 would see an average voltage due to rapid switching. If this is not the case, why is regulation fairly tight at OUT? I'm using basic test equipment so who knows what kind of noise is really going on. My analog/digital clock as a load doesn't seem to mind and keeps time.
This tells you how to make a feedback circuit that works:
https://www.ti.com/seclit/ml/slup340/slup340.pdf
You have a voltage-mode buck regulator.

By the way, if you have a ferrite toroid, it will saturate and lose most of its inductance - you need an iron powder toroid.
 

AnalogKid

Joined Aug 1, 2013
10,067
If you want to modulate the pulse width, then don't use a comparator, because its output can only be low or high,
Incorrect. That comparator is simply an opamp with 1/2 of the output stage missing (and less compensation). It is the (linear) error amplifier in the control loop. Referring to post #1, the circuit is operating as a fixed-gain amplifier with a variable input. The input is from the R8 wiper, and the negative feedback is from the R6-R7 node. R6-R7 set the gain at 11.

In post #14, things have changed. Now the circuit has a fixed input (D2 Vf) and variable gain (R6). But U4 still is operating as a linear amplifier.

NOTE: Powering the error amplifier from the output it is regulating is not a good idea. Also, a variable-gain control loop is more difficult to compensate.

ak
 
Last edited:

AnalogKid

Joined Aug 1, 2013
10,067
I'll call the anode of D2 VREF. The voltage at this node likes to jump around under test.
Add two noise filter capacitors across D2, a 0.1 uF ceramic and an electrolytic in the 10 uF - 100 uF range.

During my trials I noticed I can significantly heat the diode with my hand and so the voltage across it drops. I didn't think it was this pronounced.
The temperature coefficient is -2 mV / degree C. If the nominal Vf is 600 mV, it decreases1% for every 3 degree rise in temperature.

The tempco is so large and linear that this type of diode often is used as a temperature sensor.

ak
 
Top