Part Number: LM4040A
I am seeing a high number of LM4040AIM3 10V references failing with higher than tolerance voltage even when tested fresh out of the tube with no soldering.
This thread has been locked.
If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.
We use these as a reference voltage supplied from 24V±10% via a 20k resistor and they are only loaded by a TL082 opamp input used as a voltage buffer. The test is being performed in an ambient of approx. 22 degrees C. We are seeing approximately 8% fail, even when tested without soldering in a SOT23 socket, with no load other than a 6.5 digit DVM. The devices that fail are always high.
Mike,
The LM4040A10 has a minimum cathode current of 75uA typical, 125uA max. With a 20k resistor you get between 0.58mA to 0.82mA of current through Rs. According to the TL082 datasheet, it can pull up to 2.8mA (1.3mA typical) of current with no load connected, so the resistor needs to be decreased.
I recommend decreasing the Rs value to at least 4kOhm so that you have enough current for operation under worst case scenario.
Regards, Diego Lewis
Hello Michael,
I have now also looked at the cathode of the devices on an oscilloscope and they appear fairly quiet, approximately 2mV of ripple that I can attribute to the DC power supply. The results for the 50 samples we tested fresh out of the packaging without any soldering are below:-
sample | voltage | notes |
1 | 9.99844 | |
2 | 10.00122 | |
3 | 10.00039 | |
4 | 10.006 | |
5 | 10.00768 | |
6 | 10.00528 | |
7 | 9.9981 | |
8 | 10.0046 | |
9 | 10.00472 | |
10 | 10.01515 | voltage drifted upwards during test |
11 | 9.99248 | |
12 | 10.00634 | |
13 | 10.00058 | |
14 | 9.99215 | |
15 | 10.00459 | |
16 | 10.00681 | |
17 | 10.00866 | |
18 | 10.00609 | |
19 | 9.99757 | |
20 | 10.00176 | |
21 | 9.99688 | |
22 | 10.00692 | |
23 | 9.99957 | |
24 | 9.99394 | |
25 | 9.99953 | |
26 | 10.01074 | voltage drifted upwards during test |
27 | 10.00502 | |
28 | 9.99873 | |
29 | 9.99645 | |
30 | 10.00184 | |
31 | 10.00336 | |
32 | 9.99911 | |
33 | 9.9933 | |
34 | 9.9989 | |
35 | 10.00454 | |
36 | 10.00286 | |
37 | 10.00607 | |
38 | 10.01026 | voltage drifted upwards during test |
39 | 9.9958 | |
40 | 10.0054 | |
41 | 10.0134 | voltage drifted upwards during test |
42 | 9.99846 | |
43 | 10.00109 | voltage drifted upwards during test |
44 | 10.00229 | |
45 | 10.00479 | |
46 | 9.99872 | |
47 | 10.00139 | |
48 | 9.99796 | |
49 | 9.99863 | voltage drifted upwards during test |
50 | 10.00185 |
I hope this info is helpful, I have more data from larger samples that show a similar distribution, measured with traceable calibrated meters.
Best Regards,
Mike
1 | 9.99844 |
2 | 10.00122 |
3 | 10.00039 |
4 | 10.006 |
5 | 10.00768 |
6 | 10.00528 |
7 | 9.9981 |
8 | 10.0046 |
9 | 10.00472 |
10 | 10.01515 |
11 | 9.99248 |
12 | 10.00634 |
13 | 10.00058 |
14 | 9.99215 |
15 | 10.00459 |
16 | 10.00681 |
17 | 10.00866 |
18 | 10.00609 |
19 | 9.99757 |
20 | 10.00176 |
21 | 9.99688 |
22 | 10.00692 |
23 | 9.99957 |
24 | 9.99394 |
25 | 9.99953 |
26 | 10.01074 |
27 | 10.00502 |
28 | 9.99873 |
29 | 9.99645 |
30 | 10.00184 |
31 | 10.00336 |
32 | 9.99911 |
33 | 9.9933 |
34 | 9.9989 |
35 | 10.00454 |
36 | 10.00286 |
37 | 10.00607 |
38 | 10.01026 |
39 | 9.9958 |
40 | 10.0054 |
41 | 10.0134 |
42 | 9.99846 |
43 | 10.00109 |
44 | 10.00229 |
45 | 10.00479 |
46 | 9.99872 |
47 | 10.00139 |
48 | 9.99796 |
49 | 9.99863 |
50 | 10.00185 |
Mike,
I compiled the data for the 100 samples you provided (in chart at bottom)
You seem confident in your test equipment, so I won't challenge that.
ZKA could add 1/2 to 1mV positive shift based on your current source. Not a significant factor.
Temperature shift from 25C typical (20ppm/C) would be 0.2mV/C with some devices being significantly more. The ones that you have seen drifting up. This is a factor, but not the whole reason.
Thermal hysteresis is typical 0.08% which is a typical 8mV shift. This is significant. The devices were not soldered, but they may have been in different temperatures between the factory test and today.
So despite a 10mV factory test window other factors will widen the actual results.
Ron,
Thank you for the reply. I had only provided data for 50 samples, and tried to add a similar chart of distribution which ended up as a repeat of the sample data, however the spread is still the same.
When I have examined a batch of >300 units soldered into boards using the recommended soldering profile I see a similar distribution centred around 10V but with approx. 8% above the 10mV spec. the highest being 10.037V. The data suggests there may be something different about this 8%.
Is there a process that will bring the devices back inside the published 0.1% specification?
Is the drifting voltage on some of the samples typical in some devices? I would have expected them to be consistent.
It does seem a bit misleading that when we buy parts that to the best of our knowledge have been shipped, stored and assembled within your specified limits then apear to fail your standard specification.
Regards,
Mike
Hello Mike,
Unfortunately the thermal hysteresis, in addition to other forms of variation previously mentioned, can cause this drift as applying a high temperature to the device while soldering can increase the die temperature and ultimately affect its accuracy.
It is possible that you could force temperature cycles on the device, say take the device down to -40C or up to 125C and let it soak for a little bit, then bring the device back to room temperature and see how the reference has shifted. You can possibly run multiple cycles of cold or hot to get the reference back into 0.1% specification.
We test all units to our specification before they are shipped, so that when the devices are shipped, they are guaranteed to meet our datasheet specification. If you feel that something is abnormal about these 8% of devices, then you could also consider submitting a return either to your local distributor or sales person.
Please let me know if you need anything else or if I can help.
Best,
Michael