LTC6655LN-4.096 DC noise appears 20x higher than spec

Hi Adamvs,

      Few questions -

What are the feedback values around your low noise opamp?

Whats the noise of the opamp?

Regards,

Jeff

Hi Jeff,

The 1000x is being provided by a LNA10 made by AlphaLabs. (This board prevents me from posting links but you can find it's specs and a teardown by googling.)

I can't comment on the noise of the LNA directly except to say that with the same setup (probe->HP filter->LNA10) and the probe attached to a 9V battery I get about 5x lower noise than with the same setup tied to the test point of the LTC6655LN.

It has an internal output LP filter which I set to 10 Hz. I also put a 16 Hz LP filter on its output to cut down on the fuzz seen on the scope. LNA10 is on its DC coupled setting. My HP filter (C= 80 uF, MLCC, and R=10 kOhm) mentioned above provide the DC blocking needed so that the 100x gain does not rail the output.

When you say - apply the same technique to a battery - you mean measures its noise? Have you tried measuring the noise of a100KPOhm resistor? Also, is the board you are evaluating the LT6655 the eval kit?

Jeff

Battery:

Yes. I am measuring its DC (0.1..10Hz) noise. Under the assumption that the battery has essentially no noise in the 0.1..10 Hz band this establishes the noise floor of the measurement apparatus.

Measuring a resistor:

I can try this. So just a standard 100kOhm metal foil resistor across the In+, In- terminals? Are you looking to see if I can observe the flicker noise? If so, do you expect it to be in the few uV p-p range?

The Board:

No, I am measuring a reference mounting on my own PCB. I provided the schematic and layout in case someone sees a mistake I made in this regard.

I'm still looking at this. I reduced my noise floor further and can say with additional confidence that LF noise appears at intervals. It can be quiet (< 1uV p-p, i.e. <2x the spec) for sometime, then start oscillating a bit. (up to 10 uV deviations over a minute or longer, ~20x the spec)

Open questions:

1) is there an Eval board with an example of or LTC6655LN with recommended components?

2) What is a recommended capacitor for  the LN pin?

I discussed this with a engineer here: (ez.analog.com/.../output-bypass-capacitor) but ever got a full answer. Because I can see large (disturbances caused by mechanical noise (e.g. tapping the board) i wouldn't be surprised if the MLCCs i used could cause some noise. But the requirements for the cap to be low/stable and yet get 100 uF are difficult to achieve.

Hi, there is not an eval board for LTC6655LN. The capacitor values are discussed in the datasheet, and depend on the settling behavior you need. If the noise is changing over time as you show, I would look at airflow and vibration as causes. The change in thermocouple voltages at the solder joints due to periodic airflow can be mistaken for low-frequency noise, as can vibration that affects ceramic capacitors. Try covering the board so that airflow is minimized.

I'll look into the airflow, thanks.

The value of the cap is discussed yes, but not the type. Type is only discussed for bypass and load.

It says "use a low-leakage type". Now, I don't actually care about the exact reference value, just stability over 0.01 ... 10 Hz

What type do you use for doing the design validation and proving the 0.12 ppm spec? MLCC? film? Maybe other types are ruled out because of leakage, but if you don't care about very slow drift maybe a polymer is a good choice?

I see this question asked and discussed repeatedly on this board and others without a consensus. I am perplexed that there isn't a reference design or recommendation.

I'll note that your suggestion about the airflow was spot on. This appears to have been the main remaining contributor to my noise. With the chip and its caps covered (just a scotch tape dome for now) the remaining noise is below my approximate measurement floor (~1 uV p-p).

So for future readers this my setup for the LTC6655LN-4.096

Board:

1) 5V power generated by LT3045

2) Input and output caps: C3225X7R1C226M250AC (22 uF X7R) and CL10B104KO8NNNC (0.1 uF C0G)

3) NR caps: 2x ZRB18AR60J476ME01L (X7R, 47 uF with built-in interposer to reduce piezo noise)

4) Cover IC and capacitors to prevent airflow and (even gentle airflow will cause 10 uV of noise)

5) isolate board from mechanical vibrations (yes I can se taps to the table)

Measurement:

1) a normal probe on 1x setting

2) Probe output to HP filter consisting of 2x C4AQLBW6100M3MK and a ordinary 1K metal foil resistor

3) HP filter output to LNA 10 scope preamp on 100x gain, 10 Hz corner (DC coupled)

This setup is only sensitive to ~0.5..10 Hz.

Consider increasing 1K-> 5K in the HP filter to get down to 0.1 Hz but characterize the noise if so.

Sorry that there isn't a single answer here, but it really depends on the application. We try to give as many options as possible, and most are fine for most applications. Further, components change over time, so specific recommendations tend to become outdated.

Thanks for the follow-up, and I'm happy that you were able to get good results.