Precise 5v ref(REF02EP+ vs MAX6126AASA50+) for Programmable PSU

Go To Last Post
17 posts / 0 new
Author
Message
#1
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Hi freaks

 

I am designing an Atmega based precision(upto 2 decimal places) 0-35v power supply.I designed the circuit and need a stable 5v ref.I have both REF02EP+ and MAX6126AASA50+ in hand.I have tested and found that the REF02 is giving stable 4.996v(-4 mV off) whereas the MAX6126 is getting off upto -1mv to +15mV(4.999v to 5.015v) by varying the temperature.I measured it with a Fluke 87V in MIN-MAX mode.

But from datasheet,I can see that both ICs have same tempco(3ppm/degree centigrade).Why the MAX6126 behaving strangely??Any suggestions??

Which will be the best choice for my power supply??Please give your feedback.....

 

Rgds//Sharanya

Last Edited: Tue. Aug 2, 2016 - 04:44 PM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

You did not say over what temperature range the change above occurred, and how does that compare with the expected temperature range your PS will be subjected too? 

 

 

 

 

 

 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Doing temperature drift tests are tricky. How have you identified other sources of error in your test? Thermocouple effects?

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

When i was testing,the initial room temp was about 35 degree celsius(In India,it is very hot normally).At that very moment,the MAX6126(with 0.1uf at both input and output) was giving around 5.010 volt and then it was dropping and finally got stabled at 5.002v.At the same moment,the REF02 was giving 4.996v(totally stable).

After that,i turned on the AC and the temp dropped to almost 25 degree celsius.At that time,the MAX6126 was giving 4.999/5.000v and the REF02 was stable at 4.996/4.997.Then i breath out some hot air from my mouth on the ICs itself.At that time,the MAX6126 changed it's output upto 5.015v and again getting stable at 4.999/5.000 when it got cold again.But the REF02 was n effect on it.

 

But today when my normal room temp is 31 degree celsius,the MAX6126 is flickering between 4.999/5.000v.The REF02 is again at 4.996/4.997v.Although i didn't conduct the hot air test on them!!

 

The temp range is nearly 18 degree celsius to 44 degree celsius at my place.So the ambient temp of the power supply will be nearly +10 degree celsius of this(due to various heating effects).

 

 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

What happens if your Fluke is temperature sensitive?

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

But why only MAX6126 seems to change it's output??If my FLUKE was temp sensitive(Maybe it is),then the REF02 output must be logged incorrectly in Fluke.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

How do you know that the max is changing it's output? If you test the two chips at different times, then that introduces uncertainty. Get a peltier effect device and make a small test chamber. Rule out the temperature effects on the measuring device. Have you considered that different wire, solder and type of ic socket might cause differences?

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Ok...the good news is changing the output cap to 10uF electrical(highest supported by the device) seems to eliminate the problem at higher temperature.Now i conducted the test starting from 29 degree celsius and gone upto 72 degree with a hot air gun.Now the MAX6126 output was flickering between 4.999/5.000.Now i will conduct some more tests and will let you guys know the result.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

(upto 2 decimal places) 0-35v power supply

?

 

Maybe I need more sleep, or more caffeine? 

 

If your power supply is "accurate" to 1/100th's of a volt, why are you even worried about a couple of mV's?

 

You would have to drift 50 mV to possibly alter your 1/100th output digit.

 

More importantly, perhaps, is that designing a "good" power supply is rather challenging.

As the load experiences step changes in current draw, the circuit will have to respond, and the "ringing" in the circuit might far exceed inaccuracies due to the small error in the Vref source.

 

I'd pick a reference and get on with your design!

 

JC

 

blush

Opps, off by a factor of 10...

 

Definitely, need more sleep, followed by caffeine!

 

 

Last Edited: Wed. Aug 3, 2016 - 08:59 PM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

DocJC wrote:
Maybe I need more sleep, or more caffeine? If your power supply is "accurate" to 1/100th's of a volt, why are you even worried about a couple of mV's? You would have to drift 50 mV to possibly alter your 1/100th output digit.

I'd say both on the sleep and caffeine.  But that is just me.  (1/100 of a volt is 10mV, isn't it?  So a 50mV change would be 5x LSD right?)

 

Now, we don't really know the purpose and use of this supply.  If OP really wants/needs that last 0.01 to stay steady over full range of load, temperature, and supply it isn't a trivial task (IME).  That would be less than 1 part out of 3500 -- about 0.03%.  In my common apps, industrial and commercial, about 1% is usually good enough, as long as stable and linear and such.  Typically without great ado an AVR8 app with ADC reading ans such ends up at a few tenths of 1%.

 

When you get (again IME) requirements of say 0.1% and lower, now you are in the realm of bench/lab quality and one cannot just have a near-single-chip AVR8 solution.  I cannot recall tackling many (any?) of these.

 

NB:  Talking of temperature range and power range and similar, some apps just don't lend themselves to hard numbers such as 12.34V.  An example is conductivity readings on a fluid to determine the strength of "concentrate" (e.g. acid, caustic) in the solution.  There are too many variables such as the hardness of the water and how "dirty" the fluid is.  So the readings become unitless -- set up the system and create the solution of the correct concentration, measured via titration or other means.  Then say my controller reads "67 on a unitless 1-100 scale.  67 is then entered as the setpoint.  Entirely different mindset, eh?

 

 

You can put lipstick on a pig, but it is still a pig.

I've never met a pig I didn't like, as long as you have some salt and pepper.

Last Edited: Wed. Aug 3, 2016 - 07:29 PM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

This supplies up to 35V or 7X the reference voltage. Hence, a 5mV variation in the reference will produce a 35mV output change at 35V.

 

Jim

Jim Wagner Oregon Research Electronics, Consulting Div. Tangent, OR, USA http://www.orelectronics.net

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

theusch wrote:
When you get (again IME) requirements of say 0.1% and lower, now you are in the realm of bench/lab quality and one cannot just have a near-single-chip AVR8 solution.
AVR XMEGA A1U has a 12-bit ADC noise of about 500micro-volt-rms (typical) (about 200ppm or 0.02%).

XMEGA E5 has a 12-bit ADC with an over-sample and decimate mode to 16 bits but there's no noise spec or data in its datasheet; would need some design considerations to reach 16 bits.

http://www.atmel.com/images/Atmel-8385-8-and-16-bit-AVR-Microcontroller-ATxmega64A1U-ATxmega128A1U_datasheet.pdf (go to page 183 for Figure 38-128.Noise versus Reference Voltage)

 

"Dare to be naïve." - Buckminster Fuller

Last Edited: Thu. Aug 4, 2016 - 05:11 AM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

ka7ehk wrote:

This supplies up to 35V or 7X the reference voltage. Hence, a 5mV variation in the reference will produce a 35mV output change at 35V.

 

Jim

 

That's why i was worried about the ref voltage.Currently I am using a 12bit DAC and I can't get the desired resolution.Current resolution is 35/4096=8.544mV/step(In the 7x range).

I ordered a 18 bit DAC(DAC8574) to minimize the error.So the resolution would be 35/ 262143=0.133mv/step

Currently with LT1013 op amp and TIP35+MJE15032 darlington, I am getting nearabout 40mv off at my output.Also i don't have precision resistors(I am using 1% MFR).I am using 12k and 2k resistors and I think that may be the problem!!

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

DocJC wrote:
As the load experiences step changes in current draw, the circuit will have to respond, and the "ringing" in the circuit might far exceed inaccuracies due to the small error in the Vref source.

 

Still now,i didn't face any type of ringing in my test setup in CV mode while using DC DUMMY load.(I varied the current from 20mA to 4.57A).

Can you please tell me what is the formula to calculate compensation cap value??I am using 100nF currently.Although there is no significant change observed with that removed.

Regulation is still very good(As i observed).There is only 0.01v drop in the range of 20mA to 5.0A current consumption range.Although i didn't tried any Inductive load.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

100nf seems excessive for a comp cap.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 1

OK GUYS....so here is the thing......My voltage divider was causing the problem.I have added a 5k 10 turn pot(that is what i have this time) in series with the 12k+2k resistor divider and adjusted the  divider to get a precise factor of 7.Now I am getting  17.50/17.49v at output while the 12 bit DAC output is 2.501v(Although the MAX6126 output is 5.000v now).laugh

 

Now i will implement the CC part(and change the DAC to 18 bit) and connect it to this CV part and will disturb you guys once again if any problem occurs!!!!!!!!wink

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

gchapman wrote:
AVR XMEGA A1U has a 12-bit ADC noise of about 500micro-volt-rms (typical) (about 200ppm or 0.02%).

But I said "AVR8"...

 

And even in the AVR8 with 10-bit results from the much-maligned ADC, many apps with decent signal drive and a bit of averaging can get about 1000 counts, stable, monotonically increasing, no "missing codes".  That approaches the 0.1%.  And IME that is good enough for nearly all "general" signals in industrial/commercial apps.  Certainly there are other signals (load cells come to mind) where we use an external device.

 

Indeed in those well-running apps, supply/signal/ref variations with changes in load/temperature/supplyV would disturb the above. Often/usually some shift affecting absolute accuracy isn't critical as long as things stay linear and so forth.  So maybe it takes the nominal 1 count in 1000 0.1% to a few more counts -- still below 1% and usually "good enough".

 

===================

OT:  All above may become moot with the apparent doubling in AVR8 prices.  At least one 'Freak that used to apply a lot of $1/$2/$3/$5 AVR8s is looking long and hard at e.g. Cortex M0/M0+ at $1/$2/$3/$5 that might offer 12-bit or more ADC and more SRAM and more speed and ... at about the same price point.

You can put lipstick on a pig, but it is still a pig.

I've never met a pig I didn't like, as long as you have some salt and pepper.