Data Error (Cyclic Redundancy Check) on an External Hard Drive: What to Do First

You open the external drive, or you try to copy one folder off it, and Windows stops with a short sentence: Data error (cyclic redundancy check). The drive is visible, some folders may even open, yet the copy refuses to finish. In Microsoft's list of system error codes this is error 23 (0x17), named ERROR_CRC, and its entire official description is the sentence you already read on screen.

That brevity is the problem. The message tells you that a check failed. It does not tell you where. This guide is about finding out, in the right order, without making the situation worse.

What the check actually is

A cyclic redundancy check is a small value computed from a block of data. The sender computes it, the receiver computes it again, and the two are compared. If they differ, the block was altered somewhere in between and it is rejected.

On an external hard drive, that comparison happens at two different places, and they correspond to two very different faults:

On the link. Data travelling between the disk, the USB bridge board in the enclosure, the cable and the computer is protected by checksums. A damaged cable, an oxidised connector, a port that delivers unstable power or a failing bridge board corrupts blocks in transit. The disk itself is fine.

On the surface. Every sector on the platter is stored with error correction data. When a sector has degraded past what that correction can repair, the drive gives up after several attempts and reports a read failure. Windows surfaces it with the same sentence. This time the disk is not fine.

Same message, opposite diagnoses. One is solved with a cable that costs a few euros. The other means the drive is losing sectors and your files should leave it today.

Before anything else: three things not to do

Do not format. Windows sometimes follows the CRC error with an offer to format the drive, or shows it as RAW. Formatting does not repair a failing surface, and it replaces the file system that still describes where your files are. The RAW case has its own guide: recovering a RAW drive.

Do not initialise the disk. If Disk Management shows the drive as "Not initialized" and answers with the CRC error when you try, the drive could not even return its first sectors. Insisting changes nothing, and on a disk that did answer, initialising would write a new partition table over the old one.

Do not start with a repair. The reason deserves its own section, further down.

Step 1: rule out the path

Start here because it is free, fast and harmless.

1. Replace the cable. Use a short one, ideally the cable of another drive known to work.
2. Change port. On a desktop, use a rear port soldered to the motherboard. Avoid front panel ports, hubs and keyboard pass-through ports.
3. Try a second computer. If the error follows the drive to another machine with another cable, the path is cleared and the drive is the suspect.
4. For a 3.5 inch desktop drive, check its power adapter. These drives do not run on USB power alone.

If the error disappears with another cable, copy your important files off the drive anyway while it behaves, then keep an eye on it. If the drive is simply invisible on every machine, that is a different symptom, covered in external hard drive not showing up.

Step 2: let SMART tell you which side failed

Drives keep internal counters, and two groups of them separate the two faults quite cleanly:

A non-zero 199 with clean 197, 5 and 198 says the damage happened in transit. Pending or reallocated sectors say the opposite. Reading these values is a read-only operation and puts no stress on the drive. The commands, and the limits of what SMART can predict, are in our guide to checking bad sectors with smartctl.

One caveat: some USB enclosures do not pass SMART commands through. If your tool shows nothing, that is the enclosure's bridge, not proof that the disk is healthy.

Step 3: copy first, repair later

This is where most guides send you straight to chkdsk X: /r. Look at what Microsoft's own documentation says that switch does: it "locates bad sectors and recovers readable information", it includes the functionality of /f, and /f "fixes errors on the disk". The same page notes that checking for bad sectors reads every sector and that runtimes of many hours are not unusual on large hard drives.

So /r is two things at once: a full surface read, and a repair that writes to the file system. On a healthy disk with a logical glitch, that is fine. On a disk whose surface is degrading, it is hours of extra work for failing hardware, combined with changes to the structures that still locate your files. The GNU ddrescue manual puts the rule in one line: "Never try to repair a file system on a drive with I/O errors; you will probably lose even more data."

The safer order is the reverse:

1. Copy the irreplaceable files first, the smallest and most precious ones before the big ones. If a file stops with the CRC error, skip it and continue. Do not retry the same file in a loop.
2. If the errors are many, image the whole drive to another disk with a tool designed for damaged media. It copies the readable areas first, records the unreadable ones in a map file and returns to them later, so the bulk of your data is safe before the difficult sectors get hammered. The method is detailed in cloning a failing drive with ddrescue.
3. Work on the copy. Repairs, scans and file recovery all happen on the image or the clone, where a mistake costs nothing.

Microsoft's page hints at the same order: it describes chkdsk /b as a parameter to use "after imaging a volume to a new hard disk drive".

When the error touches a single file

Sometimes the whole drive copies cleanly except one file that always stops at the same percentage. That pattern means one or a few sectors under that file are unreadable. The rest of the disk may be perfectly usable for now, but a sector that died is information: note the SMART values, finish the backup, and treat the drive as one that has started to age.

For the file itself, an image made with a damaged-media tool will contain everything except the unreadable sectors, filled with zeros. A video or a large archive often survives that with a glitch. A compressed document may not.

When chkdsk is the right tool

There is a legitimate place for it:

• the cable or port was the cause, the error is gone, and SMART shows no pending or reallocated sectors;
• your files are already copied elsewhere;
• the file system reports inconsistencies after an unsafe removal.

In that situation chkdsk X: /f repairs the logical structure, and /r adds the surface scan if you want the full check. Microsoft notes that interrupting chkdsk is not recommended, so start it when the drive can stay connected until the end.

When to stop and hand it over

Stop powering the drive and consider a professional laboratory if any of these appear alongside the CRC error:

• the drive clicks, grinds or beeps;
• it disconnects and reconnects by itself during copies;
• pending sector counts climb between two readings;
• the copy speed collapses to a few kilobytes per second across the whole disk.

These signs describe a mechanical or head problem, and every additional hour of operation reduces what a lab can get back.

Summary

Data error (cyclic redundancy check) is Windows error 23: a block failed its integrity check. Swap the cable, the port and the computer to clear the path. Read SMART to see whether the interface counter (199) or the surface counters (197, 5, 198) moved. If the surface is involved, get the files off first, by direct copy or by imaging, and keep chkdsk for the moment when nothing on that disk exists in a single copy.