Voltage Reference

[gnif]

Hi,

I acquired some samples from Maxim for a reference supply to check my equipment against and I was wondering if I could get some feedback about the design. I ordered some free samples of the MAX6350CSA+ 5V voltage reference which advertises 0.5ppm - 1ppm tempco, this seems pretty decent for a simple low power solution. The example circuit in the datasheet shows a 2.2pf and a 2.2uf cap, one across the input, one on the output.

So my questions are:

1) Is this a good chip to use as a poor man's reference?
2) If I power this from a 9VDC battery, do I need to worry about the caps?
3) Does this device require a minimum turn on time to become stable?

Any feedback would be greatly appreciated, thanks.

[branadic]

Not the perfect choice but a good one. The MAX6350 is a good reference, except its plastic package, based on buried-zener. I've ordered MAX6350MJA (16,10€/pc), the CERDIP version. It is specified to have maximum 2.5ppm/°C tempco over a -55°C to +125°C temperature range (means less at room temperature) and is inherent to humidity, an important point due to longterm stability for a reference. This is why the LM399 and LTZ1000 are packaged into a metal can.
You should buffer the output with a zero drift low noise opamp or transistor. Don't forget to spend the 1µF noise reduction cap.

--> also interesting: AN82f (Linear Technology), AD586LQ, MAX6126, LTC6655-2.5LS8, MAX6126, ADR444 <--

[gnif]

Not the perfect choice but a good one. The MAX6350 is a good reference, except its plastic package, based on buried-zener. I've ordered MAX6350MJA (16,10€/pc), the CERDIP version. It is specified to have maximum 2.5ppm/°C tempco over a -55°C to +125°C temperature range (means less at room temperature) and is inherent to humidity, an important point due to longterm stability for a reference. This is why the LM399 and LTZ1000 are packaged into a metal can.
You should buffer the output with a zero drift low noise opamp or transistor. Don't forget to spend the 1µF noise reduction cap.

--> also interesting: AN82f (Linear Technology), AD586LQ, MAX6126, LTC6655-2.5LS8, MAX6126, ADR444 <--

Thanks for the reply, but do I need the noise reduction cap if I am powering it from a battery supply? And why is the buffer required if this is only going to be used to test the calibration of my meters?

I like to understand the reasons why these things are needed, my level of analog electronics is still fairly basic. I have set the goal of eventually building a high precision micro controlled switch mode bench PSU so I am currently in the research phase trying to learn about all the different building blocks I will need, how they work and how to correctly use them.

[psycho0815]

From what i understand from the datasheet and my general electronics knowledge, you shouldn't' need the caps nor the opamp. Batteries are usually very clean powersources and your not expecting any transients or current draw at all, so i guess you could get along without those. But including those things gives you more flexibility and some protection, like when you accidentally short out the output it won't kill the reference, but the propably cheaper opamp. And anyways those caps won't hurt, so i really don't see the point of skimping on like a Dollar worth of components.

That said, I'm not really an expert on analog stuff, so all that i said may be grossly wrong.

[gnif]

Thanks, it is not about skimping on cost, it is more that I am in a more remote area and getting these parts, mainly the op-amp, is a PITA. My local Jaycar is just a reseller and only stocks all the commercial rubbish and a few connectors.

What op-amp would you suggest?

[CarlG]

The noise reduction cap is for reducing internal noise of the reference, not the power supply noise. So is you want to minimize noise, add the NR cap.

[gnif]

The noise reduction cap is for reducing internal noise of the reference, not the power supply noise. So is you want to minimize noise, add the NR cap.

I took measurements on my CRO before and after adding this cap, and you are correct. Before hand I was seeing a 10uv voltage ripple every 4us, after adding the cap it is pretty much gone. It does however take about 200ms for the cap to charge and the output to become stable before the ripple goes away, I guess due to the charging of the cap.

With regards to the cap, I just used a 1uF electrolytic I had on hand, would a different type of cap be better suited here?

[BravoV]

Should be no problem, since the datasheet doesn't mention a specific type of cap.

My MAX6350 for quick & dirty DMM check, made while ago.

[gnif]

Nice, I like your probe inputs, I epoxied some banana jacks onto the board and wired them up. I got the small surface mount package also so I had to use a knife and cut some tracks up to make the pads for it to break it out. Not very neat, but works . I did add a trim-pot for adjust also, but I think I may remove it and just label its voltage so there is no doubt about what it should be reading.

[mzzj]

Not the perfect choice but a good one. The MAX6350 is a good reference, except its plastic package, based on buried-zener. I've ordered MAX6350MJA (16,10€/pc), the CERDIP version. It is specified to have maximum 2.5ppm/°C tempco over a -55°C to +125°C temperature range (means less at room temperature)

I would be really careful before making such a conclusions...
Sometimes tempco is specified between endpoints of the temperature range and it actually does not specify absolute maximum for tempco slope at every point.
Ie. roughly S-shaped tempco curve has nice looking ppm/°C value if you calculate it based to extremes. In reality it may have nastiest part of the curve slope right at room temperature.
Maxim specification Note 1: Temperature coefficient is measured by the box method; i.e., the maximum ?VOUTis divided by ?T x VOUT. for sure leaves some room for more than absolute 2,5ppm/°C  tempco.

I got my "fingers burnt" once with Vishay precision resistors, specified as 0,2ppm/ °C over entire temp range. In reality some of those buggers had 3ppm/°C tempco at +23°C !
After contact with Vishay rep. and reading every line of the tiny print in the datasheet I found out that this was more or less expected. 

[BravoV]

Nice, I like your probe inputs, ....

Thanks, that is made from approx 0.5 mm thick copper wire from stripped cable, then coated with solder and have been wound at dmm probe's pin to form the spiral shape, and also this to ensure the inside spiral's diameter is at correct common probe's pin diameter. So far its effective to hold the probes firmly when checking the voltage without the need for hands holding the probes, and ductile enough to be shaped if needed.

[gnif]

I installed a pin header so that the pot can be jumpered into circuit. My Agilent DMM reads 4.993-4.994v, seems the longer it is on the more stable it is at at 4.994v. Would be nice to have an additional DMM (or 20 like Dave) to check this against

[BravoV]

The fun experiment with voltage reference, if you have 6 or at least 5.5 digits dmm, hook it up, after its stabilized, take note on room temperature and also voltage readings on both input and output. Then use a hair dryer to blow the chip that its almost too hot to touch by finger for a while, and watch the drift too during heating up. Finally lets it cool down back to room temp and measure again, you will see the effect of thermal hysteresis that even its cooled down to the same exact previous temp, the voltage won't be the same again ever.

On crappy or cheap vref chip, the effect of thermal cycling sometimes can make huge deviation in multiple folds more than specified in the datasheet. 

[mzzj]

The fun experiment with voltage reference, if you have 6 or at least 5.5 digits dmm, hook it up, after its stabilized, take note on room temperature and also voltage readings on both input and output. Then use a hair dryer to blow the chip that its almost too hot to touch by finger for a while, and watch the drift too during heating up. Finally lets it cool down back to room temp and measure again, you will see the effect of thermal hysteresis that even its cooled down to the same exact previous temp, the voltage won't be the same again ever.

On crappy or cheap vref chip, the effect of thermal cycling sometimes can make huge deviation in multiple folds more than specified in the datasheet. 

And that is (one of) the reason(s) why serious cal lab grade voltage references have large built-in battery to keep the reference heaters (ie LTZ1000) activated even during transportation.
Some others use "reference conditioning" by cycling the reference temperature back and forth with a decreasing temperature amplitude. (similar to magnetic "degaussing")

[bingo600]

Guyzz

I have some (100+) MAX6198AESA , 4.096v/500uA - 5ppm/c Refs
I got them cheap off "the bay" some time ago , and saw the "buffer opamp" mentioned above.

As i'm an analog noob , i was wondering if anyone could show/"point me to" a reasonable circuit ($$) that might deliver 100mA or so ?

I know stability/tempco costs $$ , and i'm not talking "Voltage Nut" stability here ...
I have a Geller-10v ref , some AD587KN's and an Agilent 34401A for that stuff.

But i'd like to be able to supply ie. an ACT14 that i'd then PWM from an uC , to create a cheap DAC.

Could i use a PNP and build a "Mini PSU" for it or is an opamp the way to go ?

Edit: Add DS URL for the 6198
http://datasheets.maximintegrated.com/en/ds/MAX6190-MAX6198.pdf


TIA

/Bingo

[ecat]

Coincidently my poor man's reference decided to arrived today, three AD588BQs from Hong Kong ...
http://www.ebay.co.uk/sch/i.html?_from=R40&_trksid=p2047675.m570.l2736&_nkw=AD588BQ

Genuine? They look new and the top markings look the part and carry a 0517 date code. No markings on the bottom unlike some pictures on the net.

My 1997 cal. 7150 reads the following

1) 10.0007
2) 10.0017
3) 10.0021

That's 1.4mV variation across 3 samples so I guess it is within  the +/- 1mV tolerance of the part and my meter could well be out by 1mV. I'll cross check with my Datron later, but since it's also many years out of calibration 

If I'm feeling keen I'll try to devise some sort of temperature related test to check the ppm/C, sounds like a whole lot of fun and games.

I could say the basic idea was having some sort of long term cross check reference but to be honest I just wanted to mess around so genuine or not it's no great loss.


Edit...

A temperature probe as close to the chip as possible and a quick but gentle blast with a hot air gun - this is how you do temperature tests But take note of BravoV's comment in reply #12 before trying this.

At stable 22C
3) 10.0021

At about 42C
3) 10.0026

And at around a stable 27C as the temperature fell away
3) 10.0023

So, 500uV over 20C = 25uV/C and 200uV over 5C = 40uv/C

Since the probe heats quicker than the chip 25uV could be an under estimation, the probe also cools a lot quicker than the chip so 40uV could be an over estimation.

The chip is rated at 3ppm/C, so at 10V nominal output that's 30uV/C. Looks pretty good really

[branadic]

As i'm an analog noob , i was wondering if anyone could show/"point me to" a reasonable circuit ($$) that might deliver 100mA or so ?

Could i use a PNP and build a "Mini PSU" for it or is an opamp the way to go ?

Refer to AN82 page 9.

I have a LM399 running on a 34410 since a few days. The circuit is the simple "10V Buffered Reference" from the Linear Technology datasheet (with LT1001, 9k, 20k, 7.5k). Instead of 9k/0.1% I had to use 18k/0.1% (TC=25ppm) and in parallel a 20k Vishay 534 pot with 10 turns (TC=20ppm). The LT1001 is a normal plastic  package DIL8 type up to now, I'm awaiting the CERDIP type.

The LM399 is thermal isolated with an additinal PP box and some foam plastic inside.

Under lab conditions (reasonably constant temperature) the voltage drifts within ±20µV, thanks to the resistors, need to find better ones. By touching the contacts of the Vishay pot the drift increases to about ±200µV.
For home lab measurements it's quite good, but could be more stable, I'm working on that but without getting a volt nut

[BravoV]

And that is (one of) the reason(s) why serious cal lab grade voltage references have large built-in battery to keep the reference heaters (ie LTZ1000) activated even during transportation.
Some others use "reference conditioning" by cycling the reference temperature back and forth with a decreasing temperature amplitude. (similar to magnetic "degaussing")

This reminds me of my self while ago, I used to be such a naive dreaming to become a volt nutter  , and almost pulled a trigger on buying those expensive LTZ1000 and LT1013 (direct from LT for genuine parts) and quality discrete components to build my own at home for multiple LTZ1000 for multiple base reference points. .

But the more I read and understand about this world, realizing the process of taking care and baby sitting these "babies" scared me like hell. All the investments, efforts, energy and time wasted are not worth it, at least for me which is just an enthusiast. Its just too overwhelming to have this kind of long time commitment. 

[bingo600]

[QUOTE}Could i use a PNP and build a "Mini PSU" for it or is an opamp the way to go ?

Refer to AN82 page 9.
[/quote]

Thanx 

Now all i need is to understand it ...
And adjust the resistors for ie. 100ma

/Bingo

[branadic]

Now all i need is to understand it ...
And adjust the resistors for ie. 100ma 

LTSpice is your friend

Today my LT1001ACJ8 received and I replaced it in the circuit. Additional I made the ice spray test and confirmed my problem with the used resistors, damn how the voltage drifts.

I also had the luck to find a perfect styrofoam box with top cover today, inner dimesions: 310mm in depth, 285mm in width, 205mm in height and 85mm of wall thickness, pretty big. Perfect for the home bench and the developments I make. It only needs an inner stainless steel encasement and a heat regulation.

[branadic]

I have now replaced the resistors to SMD type 0805 with max. 5ppm/K and it seems the circuit is now much more stable with ambient lab conditions. Otherwise noted in the datasheet I had to recalculate the resistor value (original 9k/0,1%) in the feedback path for selected references to match my 10.00x V requirements.

1: 6.84409V        LM399H (National Semiconductor) --> 15k || 24k
2: 6.92224V        LM399AH (National Semiconductor)
3: 6.88884V        LM399H (Linear Technology)
4: 6.87789V        LM399AH (National Semiconductor)
5: 6.88650V        LM399AH (National Semiconductor)
6: 6.86366V        LM399AH (National Semiconductor)
7: 6.91630V        LM399AH (National Semiconductor) --> 15k || 22k

Number 1 is (still) running under lab conditions, the HP 34410A has warmed up for more then one week:

12.03.2013: 10.00340V / 10.00341V
13.03.2013: 10.00338V / 10.00339V
14.03.2013: 10.00337V / 10.00338V at morning & 10.00338V / 10.00339V at closing time
15.03.2013: 10.00337V / 10.00338V at morning

It is planed to run a "Burn-In" (168h @ 125°C), but that will take some preparing time. I've also ordered some crystal heaters, that could be mounted on top of the resistors:

http://www.kuhne-electronic.de/en/products/crystal-heater/qh-40-a.html

By now I'm fine with the results, but will watch out the further drift.

P.S. It's the nature of an engineer to ask what is possible, not was is required, to reach the limits and try to pass them. We search for absolut accuracy and are unhappy with relative accuracy, unhappy with physical limitations and defacto standards, knowing that we are living in a relative Einstein's universe, where constants are variable

[Neganur]

Coincidently my poor man's reference decided to arrived today, three AD588BQs from Hong Kong ...
http://www.ebay.co.uk/sch/i.html?_from=R40&_trksid=p2047675.m570.l2736&_nkw=AD588BQ

Genuine? They look new and the top markings look the part and carry a 0517 date code. No markings on the bottom unlike some pictures on the net.

I couldn't resist my curiosity since I bought two genuine BQ from AD directly a while ago ($$$ ). Maybe a long shot but the CDERDIP I received from the link above looks like it has been sanded/ground with a very fine substance and reprinted. You can hint a very faint print on the bottom side of the chip.

Will see if I can find the microscope and shoot some pictures.

[branadic]

Will see if I can find the microscope and shoot some pictures.

Any news about your references?

[amspire]

Just noticed this ebay seller is selling used LM199 refernces for $8. They also have used Vishay foil resistors at varying prices.

I was put onto this site by this post from casinada.

https://www.eevblog.com/forum/buysellwanted/0-01-resistors-or-verified-0-1-resistors-wanted/msg220347/#msg220347

[Huluvu]

I have now replaced the resistors to SMD type 0805 with max. 5ppm/K and it seems the circuit is now much more stable with ambient lab conditions. Otherwise noted in the datasheet I had to recalculate the resistor value (original 9k/0,1%) in the feedback path for selected references to match my 10.00x V requirements.

1: 6.84409V        LM399H (National Semiconductor) --> 15k || 24k
2: 6.92224V        LM399AH (National Semiconductor)
3: 6.88884V        LM399H (Linear Technology)
4: 6.87789V        LM399AH (National Semiconductor)
5: 6.88650V        LM399AH (National Semiconductor)
6: 6.86366V        LM399AH (National Semiconductor)
7: 6.91630V        LM399AH (National Semiconductor) --> 15k || 22k

Number 1 is (still) running under lab conditions, the HP 34410A has warmed up for more then one week:

12.03.2013: 10.00340V / 10.00341V
13.03.2013: 10.00338V / 10.00339V
14.03.2013: 10.00337V / 10.00338V at morning & 10.00338V / 10.00339V at closing time
15.03.2013: 10.00337V / 10.00338V at morning

It is planed to run a "Burn-In" (168h @ 125°C), but that will take some preparing time. I've also ordered some crystal heaters, that could be mounted on top of the resistors:

http://www.kuhne-electronic.de/en/products/crystal-heater/qh-40-a.html

By now I'm fine with the results, but will watch out the further drift.

P.S. It's the nature of an engineer to ask what is possible, not was is required, to reach the limits and try to pass them. We search for absolut accuracy and are unhappy with relative accuracy, unhappy with physical limitations and defacto standards, knowing that we are living in a relative Einstein's universe, where constants are variable

Hey branadic,

your board looks good 
Please consider the thermoelectric voltage effect on your jacks, connectors, cables ...  and avoid mixing the metals.

huluvu