New: LTC6655LN

[splin]

Just when you thought you were getting to grips with measuring the 0.25ppm p-p LF noise of the regular LTC6655, up pops the LTC6655LN with half the noise at 0.12ppm!



Just happened to stumble across this, which doesn't seem to have been officially announced yet, but appeared in an update to the 6655 datasheet (revision note dated 19th Feb 2019):

https://www.analog.com/media/en/technical-documentation/data-sheets/LTC6655-6655LN.pdf

Only 2.5V plastic MSOP version listed as yet, but hopefully the range will be extended gradually.

[ycui7]

I was once hoping to build a stable voltage reference with LTC6655 LS8 package by regulate the PCB temperature.

However, the one board I assemble has always been drifting at ~1ppm/day rate for three months. One the other side, the tempco is almost un-observable on 34401A with the board temp being actively controlled (no oven, the temperature regulation is for the PCB itself)

Eventually, I gave up on continuing that routine and went the LM399.

[Magnificent Bastard]

This is very useful if all you need is a very low noise reference with good short term stability.  There are lots of applications where this reference is "Good Enough".

[splin]

The LTC6655LNBHMS8-2.5 and LTC6655LNCHMS8-5 are now available. The datasheet hasn't been updated; as above it shows 0.1 to 10Hz noise of .12ppm p-p.

However, the product information page here:

https://www.analog.com/en/products/ltc6655.html#product-reference

shows the headline features as:



That is very oddly worded and doesn't exactly fill me with confidence that the LF noise is any better than the standard LTC6655 which would be a shame. The overview implies that the main difference is that the LN version has an NR pin to reduce HF noise by adding a capacitor. Perhaps the hoped for LF noise improvement didn't work out?

Even more odd is that there seems to have been no news releases at all apart the new LN version which seems very out of character.

Anybody tried one out?

[Andreas]

Hello,

the 100uF gives a cut off frequency of ~ 5 Hz.
So why should it filter any relevant portion of the LF-noise?

And caution: the LTC6655 noise is very dependant on the supply voltage.
For low noise you need a rather low supply voltage.
Otherwise the noise is in the range of a buried zener (~3uVpp for a 5V reference).

with best regards

Andreas

[splin]

Hello,

the 100uF gives a cut off frequency of ~ 5 Hz.
So why should it filter any relevant portion of the LF-noise?

I wouldn't expect it to. However the datasheet clearly states LF noise .12ppm for the LN version compared to .25ppm. I assumed this was due to design oir process changes beyond the NR pin addition, if somewhat surprising. Sadly it looks like the datasheet is simply wrong.

Thanks for the heads up on the noise/Vcc relationship.

[Andreas]

to VCC dependancy see also here: (one sample)

https://www.eevblog.com/forum/metrology/diy-low-frenquency-noise-meter/msg986718/#msg986718

with best regards

Andreas

[Kleinstein]

Capacitance at the NR pin can reduce the higher frequency noise. If the cap is large enough, this higher frequency band can start at some 1-5 Hz and thus also reduce the 0.1 - 10 Hz noise. So the reduced noise specs could be just from the NR pin and filtering. 

For a normal just display oriented meter however the interesting band is more like 0.01 - 1 Hz and here the extra filter would not help much.

[David Hess]

Filtering flicker noise is tough; it is almost always better to remove it by starting with less.  Long time constant filters take longer to settle and errors due to leakage are difficult to control.  Longer integration time for measurements helps less than expected because extending the measurement time picks up more noise at lower frequencies where flicker noise increases.

There is a clue to its pernicious nature in its name which comes from the "flicker" in the least significant digit of the measurement.  I have seen many systems where that was the only clue needed to know something was not right.

[RandallMcRee]

Just in case you have don't visit Analog's site every day...here is an apropos article:

https://www.analog.com/en/analog-dialogue/articles/why-does-voltage-reference-noise-matter.html

[branadic]

Meanwhile the low noise version is also available in LS8 package.

-branadic-

[macaba]

That's getting interesting... for industrial DAQ the LTC6655LN in LS8 is very appealing, thanks for the alert.

[RandallMcRee]

That's getting interesting... for industrial DAQ the LTC6655LN in LS8 is very appealing, thanks for the alert.

Not too long ago AD also upgraded their ADR4550, ADR4525 to the D grade. No announcement for that one, either, but I did get a few samples. It is in the LS8 package and has 0.8ppm TCR. Very impressed with the stability.

https://www.analog.com/media/en/technical-documentation/data-sheets/adr4520_4525_4530_4533_4540_4550.pdf

[Kleinstein]

Strang that they use the D-grade for best performance.

The different case seems to make quite some difference.

[Echo88]

Did someone actually measure and confirm the specced 0.1-10Hz noise of 0.12ppm-pp for the LTC6655LN?
So far it seems to be the lowest noise reference in the low frequency region apart from the ADR1000 with its real ~0.5µVpp at 6.6V, giving 0.08ppm-pp.
Like Andreas said, the datasheet-noise-graphs of the LTC6655LN dont continue to the necessary lower frequency regions to prove the spec.
Meanwhile we also have the REF70 from TI as a contender, claiming "industry’s lowest noise (0.23 ppmp-p)".

The LTC6655LN isnt available right now and id like to use it to test the noise specs of the new AD4030-24 ADC at different voltage input, not only with a shorted input. So if one of you guys have some 5V or 4.096V-versions that they want to sell within the EU write me a PM

[Andreas]

to test the noise specs of the new AD4030-24 ADC at different voltage input, not only with a shorted input.

Hello,

I would use some new NiMH (LSD) AA cells (kept at constant temperature) for a low noise voltage source.
Thats how I check the noise floor under bias of my LNA.

with best regards

Andreas

[Echo88]

Which voltage noise results did you get for the NiMH cells Andreas? Did you also measure different chemistries/battery sizes?

[branadic]

Which voltage noise results did you get for the NiMH cells Andreas? Did you also measure different chemistries/battery sizes?

At least that was measured and covered in multiple paper out there.

-branadic-

[Andreas]

Which voltage noise results did you get for the NiMH cells Andreas? Did you also measure different chemistries/battery sizes?

Hello,

I usually check the noise floor of my 1/f LNA with a pack of 8 NiMH AA cells in series before doing measurements.
Total noise should be 120-200uVpp (most of it from the LNA leakage current of the input capacitor).
For me that is "good enough" for measuring the typical noise of a LTZ1000/ADR1000 (Ratio > 3:1) so I did no further checks.

https://www.eevblog.com/forum/metrology/lowest-noise-op-amps-for-low-frequency-low-level-ac-coupled-signals/msg2218644/#msg2218644

https://www.eevblog.com/forum/metrology/characterising-low-frequency-noise/msg936355/#msg936355

with best regards

Andreas