Matching Resistors Using Wheatstone Bridge Or Other

[Shadders]

Hi,
I am looking to match resistors with a very high degree of accuracy/tolerance.

One can purchase 0.01% resistors quite easily, but with regards to a project such as an R-2R ladder DAC, which has 24bit depth, then the accuracy/tolerance would have to be 0.000005%.

Since such a tolerance is not available, then the next option is to match resistors using a Wheatstone bridge.

What i am not sure of is whether it is possible to achieve the matching to such tolerance, or if there is another method of achieving the accuracy such as paralleling the resistors (0.01% + trim) to match the target resistance. This would require a fair amount of physical testing to trim the resistor to match the selected target.

Is this method doomed to fail, or are there better alternative methods ?

Thanks and regards,

Shadders

 

[KlausST]

Hi,

One can purchase 0.01% resistors quite easily, but with regards to a project such as an R-2R ladder DAC, which has 24bit depth, then the accuracy/tolerance would have to be 0.000005%.

Where do you have this information from?
I allways thoght the one benefit of the R-2R ladder techniques us, that they don't need high precision resistors.

And I don't know how a wheatstone bridge could help you here.

But you may find out that R-2R ladder precision is also influenced by ON resistance of the used analog switches. But
ON resistance depends on supply voltage, signal voltage, production tolerance, temperature...
How do you handle this?
Indeed I assume you can't easily produce better performance with discrete devices, than with a cheap ready to buy DAC.

*****
Another technique is to use a delta-sigma DAC.

Klaus

 

[FvM]

I always thought the one benefit of the R-2R ladder techniques us, that they don't need high precision resistors.

I fear, R-2R ladder definitely needs precision resistors on the MSB side.

I didn't yet see a 24-bit ladder DAC, if there are any, they'll utilize trimming and/or digital correction tables. Making the DAC with discrete resistors sounds difficult just considering TC matching.

 

[Shadders]

Hi Klaus,

Thanks for the reply.

This is a hobby based project - so some of the issues you have raised are not an serious issue.

For the ON resistance - i located a low on resistance analogue switch, and the voltage used is relatively low which is 5volts. So i will not be too bothered by the variations. Audio people have a preference to signed magnitude DAC - so this is just to meet this criteria.

I know that one cannot provide better performance with discrete components than the usual IC fabrication process with laser trimming etc.

The reason for initiating the discussion with the Wheatstone Bridge is that it is very good at indicating changes in a component.

The process i will have used is to match one resistor with selection using the Wheatstone Bridge, by trimming the test resistance with a parallel resistor of a higher value.

The tolerance required is the reciprocal of (2^bit depth), converted to a percentage to determine the accuracy required of the resistors.

I understand that existing silicon solutions are preferred - but i was hoping that someone already has a method for matching resistors to such a precision.

Regards,

Shadders.

 

[Shadders]

I fear, R-2R ladder definitely needs precision resistors on the MSB side.

I didn't yet see a 24-bit ladder DAC, if there are any, they'll utilize trimming and/or digital correction tables. Making the DAC with discrete resistors sounds difficult just considering TC matching.

Hi FvM,

I recall Philips used to produce a sign magnitude DAC - 24bit with laser trimmed resistances - but they are not produced anymore and remaining stock is sold for in excess of £50.

Digital correction tables - had not considered that before - but i assume that you will need precision Digital Volt Meter (would need 7 significant digits) - or use precision THD measurement mechanism ?

Either way - i am looking for a mechanism that can match resistors with high accuracy, that does not rely upon accuracy of the other circuit components to be the same as the target accuracy.

Thanks and regards,

Shadders.

 

[KlausST]

Hi,

did you do some calculations?

With a 10k resistor and +/-0.000005% tolerance ...
--> then you talk about +/-50mOhms.

Lets assume your wheatstone bridge is optimized for those 10kOhms (symmetric) and it is powered with 5.0V
--> then you talk about a tolerance of +/-1,25nA or 12,5uV bridge mismatch.

Do you have the knowledge and the equippment to measure those signals without introducing higher errors (like with thermocouple effects).

To be honest: I don´t have..

Klaus

Errors in the resistors will cause:
* gain errors
* offset errors (depends on circuit)
* INL errors (mainly causes intermodulation distortion in audio signals)
* DNL errors (mainly causes noise in audio signals)

 

[Shadders]

Hi,

did you do some calculations?

With a 10k resistor and +/-0.000005% tolerance ...
--> then you talk about +/-50mOhms.

Lets assume your wheatstone bridge is optimized for those 10kOhms (symmetric) and it is powered with 5.0V
--> then you talk about a tolerance of +/-1,25nA or 12,5uV bridge mismatch.

Do you have the knowledge and the equippment to measure those signals without introducing higher errors (like with thermocouple effects).

To be honest: I don´t have..

Klaus

Hi Klaus,

Thanks - yes i looked at a precision galvanometer and the best i could locate is one with 270uV/div sensitivity - so as you have indicated - insufficient.

The best i could attain is approximating a lower bit depth - 18 or 19 bits.

Thanks for your replies - it seems that it is not possible to achieve a higher precision to support more than 18bits. Unless there is another method ?

Thanks and regards,

Shadders.

 

[crutschow]

I think 24-bits accuracy is impossible with discrete parts, not matter how carefully you screen them.
Generally sigma-delta techniques are needed to get much over 18-bits.