Hi

I've been using a 7805 for a while now as my voltage source, and the output accuracy has never been an issue, but now that I want to start using precision voltages it has become an issue...

When I connect my 7805 to my oscilloscope, im getting 400mv of fluctuation, If someone could point me in the direction of what I need to start learning and in what direction I must go to have less than 1mv ripple/noise in my output?

Thanks
Clauds

 

circuit, circuit, circuit.
You must show a circuit schematic so that we can help you.
A lot has to do with physical layout. Can you post a photo of your board?

 

Do you have a 0.33uF cap from the input terminal to the ground terminal, and a 0.1uf cap from the output terminal to the ground terminal, with the leads as short as they can be?

Those caps are not optional. Without them, it is very easy to wind up with oscillations. Having the 0.33uF on the input and 0.1uF on the output effectively dampens out those oscillations.

You have not mentioned the nature of your load. The 78xx series needs to have at least 5mA load in order to provide the guaranteed regulation specified in the datasheet. If your load is very low, you might find it convenient to add an LED in series with a current limiting resistor on the output as a "power on" indication.

The 78xx and 79xx series of regulators have been around for quite a while. While they are not "terrible", they still have a considerable amount of dropout voltage from the input to the output terminals. Many newer linear regulators have lower dropout voltage.

 

Wow, that was a rapid response...
Thank you

The circuit is simply supplying a pic 16f877a chip...

This is what im currently testing on... And am not getting the resolution i'd like...

Ive also tried the 1st circuit on page 7 on following pdf with the LM350AT:
http://docs-europe.electrocomponents.com/webdocs/0780/0900766b807800e7.pdf
It is better, around 200mv ripple...

But my final goal is to have the regulator supplying a circuit with a 470ohm resistor and led attached to a series of ball bearings and connection points in steel, when the contact patch of the steel decreases on ball bearing, the resistance should increase, thereby altering my voltage, which is read by the microcontrollers ADC...

Since this will measure 0.01mm change in contact patch, I need it to be quite accurate, hence the 1mv ripple or less.

Would using a voltage reference suit this project better?

Thanks
Clauds

 

Wow, that was a rapid response...
Thank you

The circuit is simply supplying a pic 16f877a chip...

This is what im currently testing on... And am not getting the resolution i'd like...

Ive also tried the 1st circuit on page 7 on following pdf with the LM350AT:
http://docs-europe.electrocomponents.com/webdocs/0780/0900766b807800e7.pdf
It is better, around 200mv ripple...

But my final goal is to have the regulator supplying a circuit with a 470ohm resistor and led attached to a series of ball bearings and connection points in steel, when the contact patch of the steel decreases on ball bearing, the resistance should increase, thereby altering my voltage, which is read by the microcontrollers ADC...

Since this will measure 0.01mm change in contact patch, I need it to be quite accurate, hence the 1mv ripple or less.

Would using a voltage reference suit this project better?

Thanks
Clauds

are these ball bearings supporting a load, and is it static or dynamic? I'm thinking that a constant current may better suite your needs.

 

Wow, that was a rapid response...
Thank you

The circuit is simply supplying a pic 16f877a chip...

This is what im currently testing on... And am not getting the resolution i'd like...

Ive also tried the 1st circuit on page 7 on following pdf with the LM350AT:
http://docs-europe.electrocomponents.com/webdocs/0780/0900766b807800e7.pdf
It is better, around 200mv ripple...

But my final goal is to have the regulator supplying a circuit with a 470ohm resistor and led attached to a series of ball bearings and connection points in steel, when the contact patch of the steel decreases on ball bearing, the resistance should increase, thereby altering my voltage, which is read by the microcontrollers ADC...

Since this will measure 0.01mm change in contact patch, I need it to be quite accurate, hence the 1mv ripple or less.

Would using a voltage reference suit this project better?

Thanks
Clauds

I think the cap after the rectifier is to small. Use say 4700uF. If you need an accurate voltage. Use a voltage reference. However they can not source much current. If this is just a lab setup. Maybe you can use LM317 and tune it in with an accurate voltmeter. I would think that would be stable during an experiment session

 

Maybe your transformer does not have enough voltage?

Using 7805 voltage should be pretty stable.

I use for instance 78L05 as voltage reference, 2x 470 Ohms to get 2.5V out.
But not for power supply, they get hot even for smaller currents.

 

are these ball bearings supporting a load, and is it static or dynamic? I'm thinking that a constant current may better suite your needs.

yes, a spring loaded plate with pins that will always be in contact with ball bearings, current might be the better way to go then...
What would be the best way to measure current on a microcontroller?
Shunt resistor, measuring each sides voltage and doing the math in the software?

t06afre, ill try a voltae reference as well

takao21203, transformer is giving me 18v unloaded, I also thought that the datasheet was showing a better accuracy, but the oscilloscope doesnt lie

Will have to do more tests and report back tom...
Thanks guys

 

That 10 uF cap on the 7805's output seems like a lot of cap... but at what frequencies?

Large electrolytic caps crap out fairly quickly. Ceramic caps work to much higher frequencies. So as counter-intuitive as it seems, add in a 0.1uF across the 10uF. Also adding a .001 cap will also help.

Also, I almost never use the VSS supply for the reference of my A2D. It just isn't stable for either the short or long term. A much better way is to get a 4.096 volt precision reference for the A2D.

Oh, and these things need a TON of cap to stay stable, 1 or even 10 uF, plus that 0.1 uF. Without those their output looks ghastly on a scope.

 

1mv accuracy on 7805 linear voltage regulator?

Accuracy and ripple are two very different things.

Assuming the input voltage is high enough to keep it out of dropout, the ripple will be whatever ripple is on the input DC source reduced by the loop gain at 120 Hz which is probably 60 dB or a shade better. That's a factor of about 1000X. So, you need enough capacitance on the input to make the ripple less than 1V. You also need some good caps on the output.

Any reg has some noise on the output, not sure what the 7805 spec is.

Accuracy is determined by the internal reference and the internal resistors. The output voltage does wander with temp change and also with output current change. Accuracy degradation from those effects is many times larger than 1 mV.

 

yes, a spring loaded plate with pins that will always be in contact with ball bearings, current might be the better way to go then...
What would be the best way to measure current on a microcontroller?
Shunt resistor, measuring each sides voltage and doing the math in the software?

t06afre, ill try a voltae reference as well

takao21203, transformer is giving me 18v unloaded, I also thought that the datasheet was showing a better accuracy, but the oscilloscope doesnt lie

Will have to do more tests and report back tom...
Thanks guys

Well I use a LM7805 (not 78L05) as voltage reference, PIC 16f884, LED display for voltage via A/D, power supply = LM2576 switcher which produces quite large ripple. Display is stable to 0.1V, I don't know what is causing the fluctations beyond that, simply don't need higher precision.

18V to 5V should not produce 400mV ripple. Be sure you have let say 100 or 220uF on the input (depending on load current), and also something like that on the output. If you only draw some mA you can get away with smaller caps. And be sure always to load the 7805 with some mA.

All loads will produce additional ripple currents and disturbances. But 7805 (78L05) only for reference should be quite stable. Maybe not 1mV but you could test with oscilloscope.

 

Maybe this could inspire you

 

Maybe this could inspire you

How is that any better then a single MCP1541 4.096V Precision Voltage Reference and a couple of caps?

 

takao: do you realize the 7805 has an Output Voltage Drift of -0.8 – mV/°C?

That means if it's temperature changes by more then 1°C you blow your 1 mV limit?

 

The 78L series has a ripple rejection ratio of 41 to 80 db. This tells you that the ripple voltage on the input will be reduced by x amount, but it won't become zero.
Reducing output ripple depends on reducing the input ripple.

 

The 78L series has a ripple rejection ratio of 41 to 80 db. This tells you that the ripple voltage on the input will be reduced by x amount, but it won't become zero.
Reducing output ripple depends on reducing the input ripple.

I tried to make sense of that matter in relation to the OP's reported output ripple voltage - i.e. 400mV [see post #1].

Looking at Fairchild's data on the LM7805 they quote at least 65dB rejection.

Even allowing for a rejection as low as 41dB this would lead to an input ripple of around 45V which makes little sense. I suspect the OP's observations aren't necessarily related to regulator input ripple problems. I'd be looking for other explanations.

 

Good point. There has to be another source of error.

 

Well I use a LM7805 (not 78L05) as voltage reference, PIC 16f884, LED display for voltage via A/D, power supply = LM2576 switcher which produces quite large ripple. Display is stable to 0.1V, I don't know what is causing the fluctations beyond that, simply don't need higher precision..

OK, I understand. You are feeding high frequency EMI from the output of a switcher into a sensitive circuit... and can't figure out why it fluctuates?

As I once tried to write an app note about when I worked at National Semiconductor...... and was threatened with death by the Marketing Director if I did:

Any linear reg (including a 7805) has a finite unity gain bandwidth of maybe 100 kHz. If you feed switcher ripple into it, the 7805 can only filter out the lower frequency content. The high frequency stuff shoots through like june bugs through a goose.

 

I tried to make sense of that matter in relation to the OP's reported output ripple voltage - i.e. 400mV [see post #1].

Looking at Fairchild's data on the LM7805 they quote at least 65dB rejection.

That applies to AC sine wave ripple with a frequency of about 100 Hz. At high frequency, where the loop gain has rolled off, it has ZERO rejection.

 

takao: do you realize the 7805 has an Output Voltage Drift of -0.8 – mV/°C?

That means if it's temperature changes by more then 1°C you blow your 1 mV limit?

Some idiot already said that on the first page of the thread.

Accuracy is determined by the internal reference and the internal resistors. The output voltage does wander with temp change and also with output current change. Accuracy degradation from those effects is many times larger than 1 mV.

Sincerely:

The Idiot