Low noise circuit filter for LTC6665 article from LT

[Overspeed]

Hello

In my quest to built the best current source possible I have located this article on ultra low noise ref voltage base on LTC6665 , can perhaps even be used for a REF5025 or other ?

Regards
OS

[Kleinstein]

Much of the circuit is the buffer to drive the ADC. The actual filter is only the simple RC combination. Here the tricky part is finding a suitable capacitor: low leakage and large capacitance. Of cause the filter can also be used with a different reference.

It depends on the use case how much filtering is actually useful. The filter helps with the higher frequency noise, but it adds some drift and very low frequency noise (though possibly not much). Instead of the effort for a filter the alternative could be a better (or mulitple) reference.

[EC8010]

Despite the fuss LT made about it when introduced, the LT6655 isn't all that quiet (it's a bandgap reference). A REF102 (subsurface Zener) is quieter and a MAX6126 (XFET) with a 100uF electrolytic on its NR pin is even quieter. My measurements were down to 0.1Hz; if you need low noise below that, you may need to experiment.

[Overspeed]

Hello

Thanks for the MAX6126     possibility I often miss the Maxim products

MAX6126 can supply 10 mA max and is available in 2.048V, 2.500V, 2.800V, 3.000V, 
3.300V, 3.600V, 4.096V, and 5.000V output voltages

Displayed performances are nice

 Ultra-Low 1.3μVP-P Noise (0.1Hz to 10Hz, 2.048V
Output)
 ● Ultra-Low 3ppm/°C (max) Temperature Coefficient
 ● ±0.02% (max) Initial Accuracy

Another good point the price as 12 Euro ( Mouser) for a MAX6126 0.02 % 1 PPM / C that not bad at all

Regards
OS

[maat]

Hello Overspeed,

what kind of current source are you trying to build? I have some experience in doing so. My application is for laser drivers and temperature controllers.

Do you already have target properties in mind?

- Max output current?
- Output impedance?
- Noise?
- Linearity?
- Sink, source, or bi-drectional?
- Long-term stability? Tempco?
- Digital or analog?

There is not much that beats an AD797+ MOSFET + ADR1399, regarding both noise and stability. I got my (unipolar) design of a 500 mA current source down to < 70 pA/sqrt(Hz) and 28 nA_rms (10 Hz-1 MHz). For details have a look here: https://tuprints.ulb.tu-darmstadt.de/27607/3/Dissertation_Baus_Patrick.pdf

[Overspeed]

Hello Overspeed,

what kind of current source are you trying to build? I have some experience in doing so. My application is for laser drivers and temperature controllers.

Do you already have target properties in mind?

- Max output current?
- Output impedance?
- Noise?
- Linearity?
- Sink, source, or bi-drectional?
- Long-term stability? Tempco?
- Digital or analog?

There is not much that beats an AD797+ MOSFET + ADR1399, regarding both noise and stability. I got my (unipolar) design of a 500 mA current source down to < 70 pA/sqrt(Hz) and 28 nA_rms (10 Hz-1 MHz). For details have a look here: https://tuprints.ulb.tu-darmstadt.de/27607/3/Dissertation_Baus_Patrick.pdf

Thanks for your post and the link an impressive and useful work

I work on pulsed DC (in the 5 Hz range with possible ratio change in the pulse duration)
-   1 amp
-   Current source
-   unipolar
-   analog
-   better than 0.05% at 1 Amp target is 50 ppm , concern is more the current value is high , difficult is to reach high precison


I need a repeatable source as tests need to be done on several months, a test duration is in the 10 minutes
I have spent lot of time on tempco and resistor setup base on VPR220 /221 and other TO220 resistor
I try to build something better than Keithley 2430 and also more affordable because:
- I dont need all the range of adjustment of the Keithley 2430 (6220 don t reach 1 Amp and the only pulsed one is the 2430) 
- I need something simple to use, Keithley 2430/ 2440 as there is a lot of possible setup is complex to use
- I don t want to spend thousand Euro in an over complex equipment for the purpose

Right now I have a basic circuit with a TI REF5025 / various amp op / VPR220 and Alpha resistors with 4 wires connection built to validate the power effect / thermal effect on the measuring resistor and heatsink design I use 5 V PSU .

Purpose is metallurgy testing for ultra-low resistance measurement

Question: As your circuit work and the design have been demonstrated  , does that possible to use it in parallel? 2 in parallel will generate 1 Amp

Regards
OS

[zlymex]

Purpose is metallurgy testing for ultra-low resistance measurement

Well, ultra-low resistance and very high precision are very difficult to achieve at the same time.
If only ultra-low resistance is required, a large current is also required just like DCCB with current extender.
If only high precision is required, one needs ‘comparison’ instead of measurement.
Either way, current reversal is a must just like DCCB and HIOKI RM3545 instead of uni-polar current.

Measurement is comparison.
In the case of RM3545, comparison is performed between the unknown and the very poor internal 0.1R resulting 610ppm accuracy spec. However, this can be greatly improved to within 2ppm by external comparison i.e. substitution method.

[Overspeed]

Purpose is metallurgy testing for ultra-low resistance measurement

Well, ultra-low resistance and very high precision are very difficult to achieve at the same time.
If only ultra-low resistance is required, a large current is also required just like DCCB with current extender.
If only high precision is required, one needs ‘comparison’ instead of measurement.
Either way, current reversal is a must just like DCCB and HIOKI RM3545 instead of uni-polar current.

Measurement is comparison.
In the case of RM3545, comparison is performed between the unknown and the very poor internal 0.1R resulting 610ppm accuracy spec. However, this can be greatly improved to within 2ppm by external comparison i.e. substitution method.

Hello
I agree , right now my project is on progress and I have started to identify first , solve after uncertainties or limit them by technical solutions as :

- Voltage source stability even go with a LTZ1000 but that request a voltage divider which generate another layer of error , REF80 have the same problem that can be interesting to compare side by side high side voltage source with divider to direct value voltage source IC as 6665 or MAX6126
- Thermal contact effect : by design and material selection
- Amp op performance that an open subject ... I use LT1012 for thermal test but I am open to proposal as AD797 as far the cost is not a NASA cost ...
- PCB design will revied at the end after all other paramters / concerns solved
- resistor ref and thermal load / heat think effect including warmup on dummy load

Note : I don t need 1 Amp so 1.00000 but I need to generate a very stable and repeatable know value as the purpose is to measure material changes with time , so trimming is not the first concern or request to adjust to 1.0000.... and all current source precision decrease with the level of current , easy at 1 mA far more difficult at 1 Amp .

Measurement by comparizon is proven solution I agree than current reversal is must , pulsed current  is used right now as I have no metrology solution for reversal setup .

RM3545 is a very nice instrument but don t go in the 0.001 range and HIOKI is a impressive brand

Regards
OS

[Kleinstein]

For 1 A even the 2 V from max6125 or similar reference is a bit on the high side. So one has the choice of looking for a really high power resistor and use some 2 V or have a voltage divider possibly already with a 2.5 V reference.
Besides having super low TC resistors one could also consider matched resistors / resistors sets in a kind of current booster / current mirror with gain. Such a circuit would be about matching and not the absolute TC.

For the OP-amp it depends a bit on the voltage - with a higher voltage like 2.5 V it would be not so much of an issue (good OP-amps are often better than the voltage reference). I would consider something like OPA205 or an AZ type like ADA4522 or OPA189 / OPA388. The LT1012 would be a choice for rather low currents, like < 100 µA.

The laser driver linked has the requirements also for the higher frequency noise and thus the AD797. For a more near DC case it would be more about the LF noise and drift.

One can connect current references in parallel if needed.

[Overspeed]

For 1 A even the 2 V from max6125 or similar reference is a bit on the high side. So one has the choice of looking for a really high power resistor and use some 2 V or have a voltage divider possibly already with a 2.5 V reference.
Besides having super low TC resistors one could also consider matched resistors / resistors sets in a kind of current booster / current mirror with gain. Such a circuit would be about matching and not the absolute TC.

For the OP-amp it depends a bit on the voltage - with a higher voltage like 2.5 V it would be not so much of an issue (good OP-amps are often better than the voltage reference). I would consider something like OPA205 or an AZ type like ADA4522 or OPA189 / OPA388. The LT1012 would be a choice for rather low currents, like < 100 µA.

The laser driver linked has the requirements also for the higher frequency noise and thus the AD797. For a more near DC case it would be more about the LF noise and drift.

One can connect current references in parallel if needed.

Hello

Thanks , I have already worked with OPA182 which is drop in improved OPA189 , that a very interesting IC regarding performance and cost

Solution to design a 500 mA and set them in // is an interesting way to investigate

Regards
OS

[GigaJoe]

for my budgeting current source 0-3a i use 15ppm regular 0.25w resistors ,  3 decades 30ma 0.3 3a ,  and each current sense has 10-15  15ppm resistors in parallel glued by thermal glue on thick (15mm ) aluminum plate 
( requirement mostly for  stability after warm-up period)
 
depend on you budget you can pick even for harder tolerance,  but USD$40 for 5ppm, current sense resistor was over my budget ..
probably large SMD 5ppm possible to find