Building a 7 decade voltage calibrator

[Jay_Diddy_B]

Hi,

One of the best solution DIY solutions is to use a high resolution DAC.

You can view this as set of matched resistors and precision switches. A micro can be used to supply the SPI signal to set the output voltage.

A good example of a DAC is the LTC2756

http://www.linear.com/product/LTC2756#

This part has softspan. and can generate + and - outputs.

This opens up the possibility of making an accurate, low frequency AC reference.

This will be much cheaper than the precision resistors and switches needed to build a Kelvin Varley divider.

Jay_Diddy_B

[branadic]

Why using 2 LTCs for a factor 1,5 multiplication?
with 1 LTC1043 you can: Divide by 2 + add the divided voltage to the input voltage.

The question is, what is more noiseless, a summing amp or an additional LTC1043 multiply by 3 circuit.

The inverted voltage for 10V will not work due to limited power supply range. (maximum 18V).
For what do you need it?

Is this your suggestion or have you simulated this? The -10V are needed for the negative reference voltage supply on the AD5791.

[quarks]

EDN had a video on building a 5-decade KV reference - neat as a starting point:

http://video.linear.com/28

Very good hint, I also thought of this one. If up to 10Volt is what you are after, it is a very nice design. And you can easily change the design to the reference of your choice (LTZ1000, LM399, ...). I have played around with it (on breadboard), with a Geller SVR and my KVDs. If I remenber right, there were a few minor errors in the schematic, so watch out if you build it. Also I had trouble to get the LT1881, so I finally ordered directly from LT in US.

Besides "only" 10V and the nice/easy free choice of reference you want to use, all depends on the KVD. If you are after great accuracy, a DIY KVD is a challenge.

[sync]

I would use a capacitive voltage multiplier * 1.5 having a 10.5V maximum range.
For the LTC2400 I would use a voltage divider * 0.666 having a 4.6V reference and up to 5.1V measurement range.

Thanks for your response!

10.5V is a good value. What calibrator resolution do you want to use?

I have some questions about the capacitive multipliers/ dividers. Are there alternatives to LTC1043?
In the LTC1043 data sheet they show typical applications and their precision, e.g. multiply by 2 has +-5ppm. Is this error stable?
I had planed to use a capacitive multiplier at the output of the AN86 to get a >=10V range. Do they work at low voltages (near 0V)? What's about the precision then?

The LTC2400 linearity is given as "best straight fit" so if you do a "end point calibration" of the range the actual value is practically doubled against the data sheet value. So 4 ppm are a 40uV missing to the 2.5 V mid range value.
(You can find this in carefully analyzing the assembler code of AN86).

Is this “best straight fit” somewhere noted in the data sheet? I'm looking for terms or descriptions so i can detect these data sheet optimizations.

The LTC2400 has a noise voltage of 10uVpp (= 0.3ppm, eff in best case). So the settling time to the final value below 10uV will need averaging many measurement values. So the much finer resolution is only valuable when you can use a integration time of 1 minute or longer.

Is there a rule of thumb between the noise and the settling time?
The AN86 says 1.4s settling time to 1ppm.

btw: Do you have an opinion about the DAC1220? It's not as good as the AD5791. But much cheaper. Or do you know other suitable DACs?

[sync]

The question is, what is more noiseless, a summing amp or an additional LTC1043 multiply by 3 circuit.

What is more stable?

The -10V are needed for the negative reference voltage supply on the AD5791.

No. The data sheet specify it with 0V for the negative Vref too.

[Andreas]

Why using 2 LTCs for a factor 1,5 multiplication?
with 1 LTC1043 you can: Divide by 2 + add the divided voltage to the input voltage.

The question is, what is more noiseless, a summing amp or an additional LTC1043 multiply by 3 circuit.

The inverted voltage for 10V will not work due to limited power supply range. (maximum 18V).
For what do you need it?

Is this your suggestion or have you simulated this? The -10V are needed for the negative reference voltage supply on the AD5791.

Why a summing amp? On a charge pump you can simply put the negative output pins on the positive input pins.
You only need a high impedant voltage follower (in any case).

The maximum voltage supply for the LTC1043 in the datasheet is given with 18V.

The negative reference supply (-10V) for the AD5791 is only needed if you want to output negative voltages (but in this case the resolution is 20uV between the steps). I would use +/-5V as reference in this case and use the internal multiply by 2 of the AD5791.

By the way you could also get 7V * 3/4 = 5,2V out of one full LTC1043. (3.5 + 1.7V)

With best regards

Andreas

[Andreas]

What calibrator resolution do you want to use?

I have some questions about the capacitive multipliers/ dividers. Are there alternatives to LTC1043?

In the LTC1043 data sheet they show typical applications and their precision, e.g. multiply by 2 has +-5ppm. Is this error stable?

I had planed to use a capacitive multiplier at the output of the AN86 to get a >=10V range. Do they work at low voltages (near 0V)? What's about the precision then?

Is this “best straight fit” somewhere noted in the data sheet? I'm looking for terms or descriptions so i can detect these data sheet optimizations.

Is there a rule of thumb between the noise and the settling time?
The AN86 says 1.4s settling time to 1ppm.

btw: Do you have an opinion about the DAC1220? It's not as good as the AD5791. But much cheaper. Or do you know other suitable DACs?

Many questions.

I still have not fully planned my calibrator. Perhaps 2 interleaved 16 Bit DACs which I have already here together with a LTC2400 for self calibration or a interleaved PWM version (2*16 Bit PWM). In the first case I will use a resistive voltage multiplier * 1.5 after the summing point of the DACs and adjust the full scale (and temperature drift) with a 2:1 LTC1043 divider and the LTC2400.

The error of LTC1043 mainly depends on leakage currents (PCB + Capacitors) and the injection currents.
At least in the divider version the value is very stable over temperature and time (below my noise limit of around 1-2uV)
The charge injection will give some (offset) error. But the OP-Amps which you use also have some error. So you will have to do some offset (+ full scale) calibration in any case.

Stability: Generally you can say the stability is better if a lower count of LTC1043 (and flying capacitors) is used.
A division of voltages is better since the output impedance is lower than the input impedance in this case.
A multiplication has more errors. (higher output impedance).

For the best straight fit look DS of LTC2400 on page 4 footnote 6. But the real meaning of the value can only be calculated by analyzing the software given in AN86.

The settling time in this case is mainly determined by the conversion rate of the LTC2400.

For the DACs I generally look for low TC and good linearity.
E.g. the AD5545BRUZ which I have already here before the AD5791 came up.

With best regards

Andreas

[sync]

Many questions.

Yes, sorry about that. Many thanks for your answers!

[branadic]

I would use a capacitive voltage multiplier * 1.5 having a 10.5V maximum range.

with 1 LTC1043 you can: Divide by 2 + add the divided voltage to the input voltage.

On a charge pump you can simply put the negative output pins on the positive input pins.
You only need a high impedant voltage follower (in any case).

By the way you could also get 7V * 3/4 = 5,2V out of one full LTC1043. (3.5 + 1.7V)

Sorry, I don't get it, maybe because I'm not familiar with switched-capacitors.

[BravoV]

Sorry, I don't get it, maybe because I'm not familiar with switched-capacitors.

Though I'm just an enthusiast, I think he means like this below using LTC1043 to get quite high precision voltage multiplier or divider, and quite reliable in term drift/stability against temperature.

I used this once while ago, very accurate, though I only have 5.5 digits bench DMM which too overkill for my lame dmm.  cmiiw

Example applications quoted from LTC1043 datasheet and there are others as well :

[branadic]

I'm able to read datasheet and found the several circuits for /2, /3, /4 and *2, *3, *4, thats not the problem, but what I didn't get is how Andreas wants to realize

multiplication *3/2
multiplication *3/4

with only one LTC1043. So I was asking for that.

[Marco]

For 3/2 can't you just use the second part of the chip as a buffer to add 1/2 to Vin?

[Andreas]

I'm able to read datasheet and found the several circuits for /2, /3, /4 and *2, *3, *4, thats not the problem, but what I didn't get is how Andreas wants to realize

multiplication *3/2
multiplication *3/4

with only one LTC1043. So I was asking for that.

Its easy ;-)

I show the example for *3/2 the *3/4 is then the homework.

With best regards

Andreas

[Jay_Diddy_B]

Hi,

I did a little simulation on the circuit proposed by Andreas.

I want to draw attention one of the challenges with this approach.

I have modified the LTspice model to step one of the capacitors through a range of values. You see that the output changes. This means that capacitors have to be matched.



The results:



There was also about 100uV of ripple at the LTC1043 oscillator frequency.


Jay_Didddy_B

[sync]

I have modified the LTspice model to step one of the capacitors through a range of values. You see that the output changes. This means that capacitors have to be matched.

You missed the connection to the S4A pin of U2.

[Jay_Diddy_B]

Hi,
It works a lot better !!

My mistake !!

Jay_Diddy_B

[Andreas]

Hi,
It works a lot better !!

Hello,

for a "perfect" LTC1043 switched capacitor design the tolerances of the components have only influence on the settling time of the output value, but not on the final output value itself.

To reach this it is necessary that all the capacitors which are switched in series in the one state are switched to parallel in the other state. Only with this approach charge differences can be equalized.

With best regards

Andreas

[Spikee]

This looks promosing: (use google translate)
http://bbs.38hot.net/thread-39900-1-1.html

[Bryan]

Just presents a log in screen?

[Spikee]

It worked on my PC (win7) but it also shows login on Android ....
Ill fix it when i am home.

[mrflibble]

Worked earlier on, now I get the login. On the same PC + browser. From the translated messages I think they may be doing some site maintenance.

[senso]

I have seen it earlier also, nice schematics, can't make a thing of the text..
Now its very slow and doesn't show much more than the top advertising.

[Spikee]

I have seen it earlier also, nice schematics, can't make a thing of the text..
Now its very slow and doesn't show much more than the top advertising.

You could try if my ripped version works :

[old rip - see edit]

EDIT:
Now as pdf:
https://dl.dropboxusercontent.com/u/100819959/ching%20chang/Reference%20devices%20Voltgen%20%28self-calibration%20of%20the%20six%20half-voltage%20source%29%20TR_V1.pdf

Edit 2:

His setup:


Main schematic:
https://dl.dropboxusercontent.com/u/100819959/ching%20chang/scheme1.png

Communication schematic:
https://dl.dropboxusercontent.com/u/100819959/ching%20chang/scheme%202.png

Self cal schematic:
https://dl.dropboxusercontent.com/u/100819959/ching%20chang/self%20cal.png

LTC2400 adc schematic:
https://dl.dropboxusercontent.com/u/100819959/ching%20chang/ltc2400.png

[branadic]

the *3/4 is then the homework.

Okay, got it. Here's the answer:

[Jebnor]

Hello Gerry,

While you have likely already seen this video, it might be useful as a basis for a buffer circuit.


I think a KV divider is an obvious choice, then using it in the feedback loop of an Opamp circuit. 

It's a good challenge. I like it.

Good luck.