32-bit ADC playground for precision measurement tasks.

[Andreas]

Hello,

what I want to know is the mechanical setup of the measurement.
Ok one thing is that they messed up the frequency response of the amplifier.

But there is still a large gap between the 0.25 ppmpp and the about 0.6-0.7 ppmpp that I measure.

With best regards

Andreas

[Echo88]

MisterDiodes mentioned that there are LTC6655 out there which dont behave noise-wise like they state in the datasheet. Maybe you have one of those Andreas?

[Andreas]

Hello,

shurely not.

the first sample of LTC6655 I bought end of 2011 in MSOP8-Package. (have not noted the date code).
the 2nd was a LS8-Package with date code R91434.

So both samples (being very similar at the same power supply voltage)
are not "recently".

By the way: even AD586 bought "recently" (Datecode of early 2016) show
a higher noise than usual. (3-4 uVpp instead of 2-3 uVpp).
And many of the AD587 of early 2016 show popcorn noise.
So there seems to be a general problem with some silicon wafers ...

Would be interesting which datecode MisterDiodes has on his LTC6655.

With best regards

Andreas

[RandallMcRee]

To add to the general malaise with LTC I just bought two LTC6655BHLS8-5 from Linear. Were backordered and just shipped from Malaysia.

I don't have the means to measure noise exactly but I estimate they are in the range mentioned in the noise thread.

But one is out of spec: it shows a voltage of 5.0602 volts the other 5.00008, in spec. First time this has happened to me (seven samples, total).

Sad. Not even in spec for the non-B version.

[MisterDiodes]

Would be interesting which datecode MisterDiodes has on his LTC6655.

With best regards

Andreas

This was an order received in late March / April, dates I believe were third week Jan, direct from LT.  They were replaced though with in spec parts.  LT  seemed to be happy to exchange, and it seemed to be no surprise.  They didn't offer any story that I know of, but that something slipped past QC testing.  This is not the usual routine for them, so it seems like something is off.

It is harder to keep the noise down with diffused voltage setting resistors built on the die - a lot can go wrong there and it's not the substrate necessarily.

The bottom line is it's always a good idea to spot-check all incoming shipments of the  before assembling a whole batch of product.

Off topic but: Even the distributors are not paying close attention recently-  Today I reviewed an incoming order from DigiKey with about 135 line items - three bags were shorted (for instance we got 45ea. 18 pin sockets instead of 50), and 2 bags had the wrong IC, and one IC bag was packed up with the wrong package type (It was supposed to be TO-220 but we got SMT's) and so on.  The order BOM was correct, the packing list was correct - it was just stuff mis-packed at their end.  That has happened on every incoming order from them in the last few weeks...Take a look at your goods carefully.  Especially if you order higher grade Vrefs or amps, make sure you receive the quality grade you paid for.

[branadic]

Hey TiN,

what about the results of your comparision?

-branadic-

[RandallMcRee]

Was reading this app note regarding using the LT6655 in conjunction with the LTC2500-32. It seems that the LT6655 is herein given a more realistic noise figure:

DN568: http://www.analog.com/media/en/reference-design-documentation/design-notes/dn568f_web.pdf

The reference used in this example is the LTC6655-5
(U2). The LTC6655-5 offers high accuracy (±0.025%
Max), exceptionally low noise (0.67ppm RMS Typ) and
drift (2ppm/ºC Max) performance.

So 0.67ppm in this app note, not 0.25ppm as in the datasheet.

Vindicates the experience of Mister Diodes, Andreas and the measurements in the Low Noise thread.

[Andreas]

Was reading this app note regarding using the LT6655 in conjunction with the LTC2500-32. It seems that the LT6655 is herein given a more realistic noise figure:

DN568: http://www.analog.com/media/en/reference-design-documentation/design-notes/dn568f_web.pdf

The reference used in this example is the LTC6655-5
(U2). The LTC6655-5 offers high accuracy (±0.025%
Max), exceptionally low noise (0.67ppm RMS Typ) and
drift (2ppm/ºC Max) performance.

So 0.67ppm in this app note, not 0.25ppm as in the datasheet.

Vindicates the experience of Mister Diodes, Andreas and the measurements in the Low Noise thread.

Hello,

the 0.67 ppm RMS (typ) in the app-note are for the wide band noise.
The data sheet lists both values:
0.25 ppm (peak-peak) for the 1/f (0.1 ..  10Hz) noise and
0.67 ppm (RMS) for the 10Hz .. 1kHz wide band noise.

What I measured is the 1/f noise.
But it seems also that this value is dependant on battery voltage
https://www.eevblog.com/forum/metrology/diy-low-frenquency-noise-meter/msg986718/#msg986718

Further the evaluation method is somewhat different.
I usually calculate the average of around 15 measurements with 10 second measurement time.
Whereas on the LTC6655 LT introduced some kind of "median" evaluation.
So they sort the resulting values and take the "middle" of the sorted values.

Further Alex has found out that the AN124 amplifier which is used to characterize the LTC6655
has a too low bandwidth and thus delivers too low noise values.

https://www.eevblog.com/forum/metrology/diy-low-frenquency-noise-meter/msg939106/#msg939106

with best regards

Andreas