low drift / low noise 100mV Source

[Andree Henkel]

Hallo, I´m asking for advice, if there is room for Improvement.
The voltage source is made to deliver input signal (SMA-Connector, terminated with 50Ohm, input voltage 100mV..5V) to a laser designed for microscopy.
To control laser during temperatur chamber test which measure stability of optical power.
I choose the components to later build a version with 100mV / 5V / adjustable 0-5V that can deliver 100mA at 5V into the 50Ohm
Driver: LT3040 low noise, high PSRR precision buffer, eval board DC2783A used as base, operated at +12V
Reference: MAX6350 (I used this for 5V, reasonable low drift and low noise - in combination with the Trim option to compensate for the cable loss caused by the 100mA at 5V) with 3 Capatitors (In, out, noise reject)
I did glue/solder the Ref to the Evalboard, Output fed to a divider 4,99k 0.01% 5ppm, 95R3 0.1% 25ppm and a 10R Trimmer.
Adjusting it using T-piece at laser input to DMM Keysight 34461A, i ca adjust to +/-5µV to 100mV display, so the accuracy is mainly determined by DMM.

Operating it with a channel of a HMC8043 Powersupply i got noticable ripple, so I added a 20mH common mode choke.

Noise now measured with scope R&S RTA4004 (500µV/Div, 10ms/Div, 20MHz) to 30µV, (25µ measured with 50Ohm terminator connected), so corrected sqrt(30^2-25^2)= 16,6µV noise.

Plan is now to mount into metal box, with isolated BNC Socket, box connected to board ground at point.
Any suggestions to the internal / external wiring, is very susceptible, probably triboelectricity, microphonics of capacitors is so far minor problem.

[TexasRanger]

I guess you can compensate for DC cable losses by connecting the V_FB Pin of the LT3040 directly at the Load, but put some capacitor directly between V_OUT and V_FB for loop stability.
The resistance of BNC connectors can vary dramatically, so I don't think thats the best option if you want high long term stability and repeatability.

Edit: you might want to take a look at "power supply remote sense" if you want to go full complicated:
https://www.edn.com/power-supply-remote-sense-mistakes-remedies/

[Andree Henkel]

What would be best way for cabling?
Maybe I switch to Twinax / Triax connection at Box output and need to improvise the change to sma at load side of cable.
Remote sense is an idea, but it would in case of the v_fb at lt3040 only remove the +side cable drop, not the -side.

So far the plan was to always use same cable between box and load, so I adjust one single time measuring at load. Maybe readjust after a longer time like 6 month.

The accuracy requirements are not that strong, +/-0,1mV at start of measurement cycle is good enough.
Important is that the voltage does change little during one temperature chamber cycle, which may last up to 72h. Also lot of vibrations.
Box is outside chamber in climatized room, cable goes inside chamber to DUT (Laser)

[RandallMcRee]

If you need high precision across a connector there are some voltage references that have force/sense on both output and ground, e.g. MAX6126.

[maat]

Is this some kind of laser diode?
Do you need a current source or voltage source?

[Andree Henkel]

It´s a diode laser module with fiber output. Inside the unit is a power regulation loop to control laser power, by regulating set value for laser current source. Another regulation loop controls laserdiode temperature via Peltier.

The SMA Input gives set value for Laser Power to the unit, 5V: Pmax; 100mV: Pmax/500.

We until recently did perform the temperature chamber tests using a hp 6114a, which worked fine for the task, but is broken now.

[Andree Henkel]

@RandallMecRee, thank you for the Max6126, might be better for 5V into 50Ohm to try it instead of trimming MAX6350 the ref for the cable drop. To try if it works with LT3040 inside the cable drop regulation loop of the Max6126.

however for the 100mV source it won´t help. Also at 5V I don´t need by far as accurate as at the 100mV. Target measurement is laser power stability, to be able to measure 1% laser power stability at the 1/500 point, I need stability better 10mV/100 -> 10µV.

We do log the voltage at laser input during temp-chamber run using 6,5 digit dmm.
I didn´t do that test yet, wanted to do reasonable cabling in before

[Vgkid]

Is the input actually 50ohms , or high impedance. As the loading effect will ruin your dividers accuracy. Now if the input is high impedance , and you are using the 6126 reference. Using the 100ohm resistor as the botyom leg of the divider. I would use a 2k with an appropriately calced fixed resistor and trimmer resistor in parallel to get it to 1948ohms. About 74.923k total parallel resistance.
Ideally you would use a resistance less than 10ohms as the bottom leg of the divider.

[Andree Henkel]

Input is 50Ohm.
Yes, was to concentrated on beeing later able to switch source between 100mV / 5V / adjustable 0 to 5 V. That´s why there is the lt3040 buffer.
However for just 100mV actually the buffer only can harm the accuracy. And 5V Ref with divider where the 50Ohm load is part of the divider as you suggest is simpler and most probably better solution, the 2mA load current needed here the reference can provide itsselv.