Non-volatile memory with leak tests
A non-volatile memory is configured to perform multiple leak tests integrated into a pre-erase process for a set (e.g., block) of non-volatile memory cells after the set of non-volatile memory cells have received programming. An inference circuit is configured to use results of the leak tests with a pre-trained model to predict whether the set of non-volatile memory cells will fail.
1 . A non-volatile storage apparatus, comprising:
a non-volatile memory structure comprising a first set of non-volatile memory cells and signal lines connected to the first set of non-volatile memory cells; and
a control circuit connected to the non-volatile memory structure and configured to program and erase the first set of non-volatile memory cells, the control circuit comprises an inference circuit, the control circuit is configured to perform multiple leak tests for the signal lines, the multiple leak tests are integrated into a pre-erase process for the first set of non-volatile memory cells after the first set of non-volatile memory cells have received programming, the inference circuit is configured to use results of the leak tests with a model to predict whether the first set of non-volatile memory cells will fail, the control circuit is further configured to perform an erase process for the first set of non-volatile memory cells including applying one or more erase voltage pulses to the first set of non-volatile memory cells while using the inference circuit to predict whether the first set of non-volatile memory cells will fail.
2 . The non-volatile storage apparatus of claim 1 , wherein:
the control circuit is further configured to perform a countermeasure to prevent loss of data prior to a failure in the first set of non-volatile memory cells if the inference circuit predicts that the first set of non-volatile memory cells will fail.
3 . The non-volatile storage apparatus of claim 1 , wherein:
the control circuit is further configured to prevent the first set of non-volatile memory cells from further storage of valid data if the inference circuit predicts that the first set of non-volatile memory cells will fail.
4 . The non-volatile storage apparatus of claim 1 , wherein:
the model is a pre-trained model received pre-trained from a source external to the non-volatile storage apparatus; and
the model is stored in the non-volatile memory structure.
5 . The non-volatile storage apparatus of claim 1 , wherein:
the control circuit is configured to perform an erase process on the first set of non-volatile memory cells while the inference circuit uses results of the leak tests with the model to predict whether the first set of non-volatile memory cells will fail.
6 . The non-volatile storage apparatus of claim 5 , wherein:
the control circuit is further configured to prevent the first set of non-volatile memory cells from further storage of valid data if the inference circuit predicts that the first set of non-volatile memory cells will fail or the erase process fails.
7 . The non-volatile storage apparatus of claim 6 , wherein:
the control circuit is further configured to allow the first set of non-volatile memory cells to store future valid data if the inference circuit predicts that the first set of non-volatile memory cells will not fail and the erase process succeeds.
8 . The non-volatile storage apparatus of claim 1 , wherein:
the control circuit is configured to independently enable and disable any of the multiple leak tests; and
the inference circuit is configured to use results of whichever leak tests are enabled to predict whether the first set of non-volatile memory cells will fail.
9 . The non-volatile storage apparatus of claim 1 , wherein:
the control circuit is configured to perform multiple erase processes on the first set of non-volatile memory cells; and
the control circuit is configured to perform the multiple leak tests before only a subset of the erase processes.
10 . The non-volatile storage apparatus of claim 1 , wherein:
the results of the leaks tests include operational data measured during each of the leak tests;
the inference circuit is configured to use the operational data to predict whether the first set of non-volatile memory cells will fail; and
the control circuit is further configured to:
allow the first set of non-volatile memory cells to store valid user data in response to determining that the erase process passed and the inference circuit predicting that the first set of non-volatile memory cells will not fail, and
prevent the first set of non-volatile memory cells from storing valid user data in response to determining that the erase process failed or the inference circuit predicting that the first set of non-volatile memory cells will fail.
11 . The non-volatile storage apparatus of claim 10 , wherein:
the operational data includes information other than a binary indicator of whether any particular leak test failed.
12 . The non-volatile storage apparatus of claim 10 , wherein:
the control circuit includes a charge pump configured to generate a voltage for one or more of the leak tests;
the control circuit comprises a current mirror in communication with the charge pump; and
the operational data comprises measurements of current from the current mirror.
13 . The method of claim 10 , wherein:
the control circuit includes a charge pump configured to generate a test voltage for one or more of the leak tests;
the control circuit comprises a counter to count clock cycles needed to drive the test voltage; and
the operational data comprises a number of clock cycles needed to drive the test voltage.
14 . The method of claim 10 , wherein:
the erase process and the leak tests are performed during a garbage collection process while the non-volatile storage apparatus is otherwise idle.
15 . The non-volatile storage apparatus of claim 1 , wherein:
the model is a pre-trained model received pre-trained from a source external to the non-volatile storage apparatus;
the first set of non-volatile memory cells comprise a block of non-volatile memory cells arranged in NAND strings;
the signal lines comprise data word lines connected to the block, source side select lines connected to the block, and drain side select lines connected to the block;
the non-volatile memory structure further comprises additional non-volatile memory cells storing the pre-trained model; and
the control circuit is configured to perform multiple leak tests for the signal lines by:
performing leak tests for the data word lines by applying high voltages to some of the word lines and low voltages to other of the word lines and sensing a first set of one or more measurements indicative of current flowing,
perform leak tests for the source side select lines by applying high voltages to source side select lines and low voltages to data word lines and sensing a second set of one or more measurements indicative of current flowing, and
perform leak tests for the drain side select lines by applying high voltages to drain side select lines and low voltages to data word lines and sensing a third set of one or more measurements indicative of current flowing;
the inference circuit is configured to use the pre-trained model with the first set of one or more measurements indicative of current flowing, the second set of one or more measurements indicative of current flowing, and the third set of one or more measurements indicative of current flowing to predict whether the block will fail; and
the control circuit is further configured to:
perform an erase process on the block while the inference is working to predict whether the block will fail,
add the block to a bad block pool in response to the inference predicting that the block will fail or the erase process failing, and
use the block to store new data in response to the inference predicting that the block will not fail or the erase process succeeding.
16 . A method, comprising:
performing multiple different leak tests for a set of signal lines connected to a first set of non-volatile memory cells, including measuring operational data during each of the leak tests;
using an inference engine with a pre-trained model and the operational data measured during each of the leak tests to predict whether the first set of non-volatile memory cells will fail;
while using the inference engine to predict whether the first set of non-volatile memory cells will fail, performing an erase process including applying one or more erase voltage pulses to the first set of non-volatile memory cells and performing erase verify after at least a subset of the erase voltage pulses until the erase process is determined to have passed or failed;
allowing the first set of non-volatile memory cells to store valid user data in response to determining that the erase process passed and the inference engine predicting that the first set of non-volatile memory cells will not fail; and
preventing the first set of non-volatile memory cells from storing valid user data in response to determining that the erase process failed or the inference engine predicting that the first set of non-volatile memory cells will fail.
17 . The method of claim 16 , wherein:
the operational data includes information other than a binary indicator of whether any particular leak test failed.
18 . A non-volatile storage apparatus, comprising:
a memory die assembly comprising:
a non-volatile memory structure comprising a block of non-volatile memory cells arranged in NAND strings, data word lines connected to the block, source side select lines connected to the block, drain side select lines connected to the block and additional non-volatile memory cells storing a pre-trained model received pre-trained from a source external to the non-volatile storage apparatus; and
control circuit connected to the non-volatile memory structure and configured to program and erase the block of non-volatile memory cells, the control circuit comprises an inference circuit positioned in the memory die assembly, the control circuit is configured to:
perform leak tests for the data word lines by applying high voltages to some of the word lines and low voltages to other of the word lines and sensing a first set of one or more measurements indicative of current flowing,
perform leak tests for the source side select lines by applying high voltages to source side select lines and low voltages to data word lines and sensing a second set of one or more measurements indicative of current flowing,
perform leak tests for the drain side select lines by applying high voltages to drain side select lines and low voltages to data word lines and sensing a third set of one or more measurements indicative of current flowing,
use the inference circuit with pre-trained model, the first set of one or more measurements indicative of current flowing, the second set of one or more measurements indicative of current flowing, and the third set of one or more measurements indicative of current flowing to predict whether the block will fail,
perform an erase process on the block while using the inference circuit to predict whether the block will fail,
add the block to a bad block pool in response to the inference circuit predicting that the block will fail or the erase process failing, and
use the block to store new data in response to the inference circuit predicting that the block will not fail or the erase process succeeding.
19 . The non-volatile storage apparatus of claim 18 , wherein:
the memory die assembly comprises a memory die; and
the non-volatile memory structure and the inference circuit are both positioned on the memory die.
20 . The non-volatile storage apparatus of claim 18 , wherein:
the memory die assembly comprises a memory die bonded to control die;
the non-volatile memory is positioned on the memory die; and
the inference circuit is positioned on the control die.