Counterfeit ATtiny/ATmega microcontrollers in DIP housings

RomanWorkshop 20 Jul 2024 13:01 23 4911 Cool? (+15)

📢 Listen (AI):

TL;DR

Counterfeit ATmega88PA-PU microcontrollers in DIP28 housings from AliExpress were examined after delivery, with one chip damaged and the other showing suspicious markings.
The fake parts had sanded tops and reapplied lettering; acetone removed the top print, while the original laser-engraved markings on genuine parts would not dissolve.
One working chip returned signature 0x1E930A, and both samples carried the same top date code “1910,” suggesting a single production batch.
Testing pointed to an ATmega88A-D variant: it flashed an LED at 4 MHz/1.8V, and a refund dispute succeeded without returning the chips.

Generated by the language model.

While shopping on AliExpress, I noticed that a certain shop also has ATmega88PA-PU microcontrollers in DIP28 cases, so I ordered 2 pieces (price under 6 zł. per piece). After 8 days of dispatch, the package arrived. The writing on the microcontrollers looked quite good, but the signature read in the programmer no longer. One piece had two pins internally damaged and the signature could not be read at all - to be thrown away. The other one was used - it had the signature 0x1E930A correct for ATmega88(A/V), some program was uploaded and the fuse bits changed. Evidently a fake with the top of the case sanded down and new lettering applied, which had worn off when exposed to acetone. The original lettering is laser-engraved, so it is impossible to remove with acetone or any other solvent. Both pieces had identical dates ("1910" = week 10 of 2019) suggesting they came from a single production batch. .
.

Counterfeit top lettering with false
model and date of manufacture of the system.

Upper inscriptions rubbed off under the influence of
acetone (grinding marks visible).

Non-counterfeit lower-case lettering with a true-
version/revision lettering and date of manufacture of the
layout.

.
This also turned out to be a hoax after looking at the lettering on the bottom of the case, which was not forged (perhaps it needs too much work). The ATtiny/ATmega microcontrollers in the DIP cases, have a second production date (green box) and a letter on there to indicate the version/revision of the chip (red box). In the pieces ordered, these dates were different: "0936" (36th week of 2009) and "0808" (8th week of 2008).

At this point I knew that I had 1 working piece of the ATmega88/88A/88V. Now I wanted to find out which specific model I had. The ATmega88V model can be identified by the fact that it should not operate at clock frequencies above 10 MHz (e.g. with a 20 MHz quartz at supply voltage Vcc = 4.5-5.5V). If the microcontroller is operating at this frequency, two possibilities remain: ATmega88 or ATmega88A. The ATmega88A model can be recognised by the fact that it should operate at a clock frequency of 4 MHz and a supply voltage of Vcc = 1.8V (a normal ATmega88 should only operate from 2.7V). The second feature of the ATmega88A model is its lower current consumption, which, at 8 MHz and Vcc = 5.0V, is: about 5mA for the ATmega88 and about 4mA for the ATmega88A (the measurement conditions are described in the datasheet note, but in practice they are difficult to obtain and measure the current with such accuracy).

In summary, after testing, all indications are that I have the "D" version of the ATmega88A (it works at 4 MHz/1.8V - it flashes the LED, while, for example, the ATmega48-20PU does not work under such conditions). I have set up a dispute for a refund (I have attached pictures of the microcontroller, screenshots of the programmer and a description of the inconsistencies in the markings), which were received several hours later. I did not have to return the received chips. Despite the scam, I am still ahead by one "paint job", although it is not entirely clear whether it is fully operational.

Useful links:
http://romanworkshop.blutu.pl/avr/tips.htm - page with information about AVR microcontrollers (designations, parameters, signatures).
http://romanworkshop.blutu.pl/proj/avrdb.htm - program with basic information (parameters, signatures, default fuse bit values).

About Author

RomanWorkshop wrote 218 posts with rating 416 , helped 2 times. Been with us since 2013 year.

Comments

Add a comment

Sam Sung 20 Jul 2024 15:11

Well, interesting. And can anything be said about my ATtiny13A for less than a dollar? https://obrazki.elektroda.pl/7095142100_1721480212_thumb.jpg https://obrazki.elektroda.pl/7187523800_1721480212_thumb.jpg... [Read more]

RomanWorkshop 20 Jul 2024 15:53

The signature for the ATtiny13/13V/13A microcontrollers is identical, so no specific model can be determined from it. A test with acetone would clarify whether the inscriptions are original, as they look... [Read more]

gregor124 20 Jul 2024 16:02

. Not quite, I have new chips purchased directly from Microchip with a production date of 2021 with the Atmel logo. [Read more]

RomanWorkshop 20 Jul 2024 16:13

. Of course this may be the case, Atmel's trademarks belong to them, so they use them on their 'new' chips. This way it's even easier to distinguish AVR microcontrollers from PICs, and users also have... [Read more]

TechEkspert 20 Jul 2024 16:44

Interesting investigation, and it's quite a mine you can get into if you buy a larger series from such a source.... [Read more]

zgierzman 21 Jul 2024 00:25

Respect to the Author for the meticulous investigation. Chinese repainting is a known problem, especially with hard to find and expensive parts. Repaint a transistor for 10 cents and sell it for $90... [Read more]

szeryf3 22 Jul 2024 09:24

@romanworkshop I thought the Chinese had stopped this practice already. But they still dispose of electro-waste. How much does it cost them to do an hour's work if they are still repainting the circuits... [Read more]

elektryku5 22 Jul 2024 12:57

As if to give anyone pause for thought, how is an Arduino from China cheaper than the Atmega itself used in it. Likewise, by the way, with 32F103s sitting in various modules.... [Read more]

Sam Sung 22 Jul 2024 14:21

I have read on avrfreaks that the signature and calibration area is actually an extra page of flash. How to look at the ISP commands: 0010 H000 0aaa aaaa bbbb bbbb - Read Program Memory 0011 0000... [Read more]

RomanWorkshop 22 Jul 2024 17:53

. An interesting topic, requiring further investigation on various microcontroller models. I was particularly intrigued by the serial number from the case, stored in a "hidden" area of FLASH memory,... [Read more]

gregor124 22 Jul 2024 19:34

These extra bytes after the 3 signature bytes and the calibration byte are Serial Number bytes otherwise (unique device ID). They can be read when reading the signature or, for example, from the program... [Read more]

RomanWorkshop 22 Jul 2024 20:00

. It's just that these concepts were only introduced in the documentation of the ATmega48/88/168/328PB microcontrollers, which are completely new chips (they have additional features), although compatible... [Read more]

gregor124 22 Jul 2024 20:52

@romanworkshop It's not as if a given layout is created once and for all. Errors happen not only in the editing of instructions, but also in the designed layouts. And this is why, for example, Silicon... [Read more]

RomanWorkshop 22 Jul 2024 21:20

. The information about the additional data in the "hidden" FLASH memory, has been completely concealed by the manufacturer. I suppose that the serial number from the case is also stored in the memory... [Read more]

RomanWorkshop 23 Jul 2024 11:48

However, the signature, calibration data and other 'hidden' information, are stored in ordinary FLASH memory, which can be erased even accidentally during serial programming. In this topic someone... [Read more]

gregor124 23 Jul 2024 15:59

. How does it not appear? After all, it is in this document: https://ww1.microchip.com/downloads/aemDocuments/documents/MCU08/ProductDocuments/Errata/ATmega8A-SilConErrataClarif_DS80000853A.pdf.pdf to... [Read more]

RomanWorkshop 26 Jul 2024 20:40

I have looked through the AVR microcontroller catalogue notes I have and find that the manufacturer has not always completely suppressed information about the 'hidden' side in FLASH memory, which contains,... [Read more]

RomanWorkshop 27 Jul 2024 22:03

I read the 'hidden' memory of the counterfeit microcontroller from the first post and found that the serial number from the bottom of the case ('8J6671') matched the one stored in its memory: 921EFF93FF0AFFFF0CEEF7FFFFFF384A 3636373107FF0724061719121913FFFF FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF FFFFFFFFFFFFFFFFFFFFFFFFFFFF2590... [Read more]

zgierzman 27 Jul 2024 23:13

Could it be, however, that this is no "fake" atmega, but a full-fledged prcesor from which the factory markings have been removed for some reason? For example, it was prepared to be sold in some device... [Read more]

FAQ

TL;DR: One buyer ordered 2 cheap DIP chips and found one dead part and one remarked part; as the tester put it, "a fake with the top of the case sanded down." This FAQ helps AVR hobbyists and repairers verify suspicious ATtiny/ATmega DIP microcontrollers using markings, signatures, voltage/clock tests, and hidden serial data. [#21161408]

Why it matters: Very cheap DIP AVRs can be electrically usable yet still be remarked, misidentified, pre-programmed, or partially damaged, which can break prototypes, kits, and small-batch repairs.

Check	Suspicious remarked chip	More credible genuine chip
Top marking	Rubs off with acetone	Stays intact; laser marking resists solvent
Top vs bottom date	Dates conflict, e.g. 1910 vs 0936/0808	Dates and revision data align
Programmer state	Wrong model, changed fuses, old code present	Expected signature and clean state
Low-voltage behavior	May reveal older silicon family	Matches datasheet behavior

Key insight: A readable AVR signature alone does not prove authenticity. The strongest practical check is to compare package markings, solvent resistance, electrical limits, and any hidden lot/serial data as one set.

Quick Facts

The reported purchase price was under 6 zł per chip, and the package arrived 8 days after dispatch; one part was internally damaged and one was usable but remarked. [#21161408]
In the suspicious ATmega88 case, the top showed date code 1910, while the lower markings showed 0936 and 0808, indicating inconsistent origin data on the same DIP package. [#21161408]
The working chip matched signature 0x1E930A, which fits the ATmega88/88A/88V family, so signature alone could not identify the exact variant sold in the listing. [#21161408]
A practical discriminator from the thread: ATmega88V should fail above 10 MHz, while ATmega88A should still work at 4 MHz and 1.8 V; plain ATmega88 should require at least 2.7 V. [#21161408]
For ATtiny13-family parts, one test example reported signature bytes 1E 90 07 and suspicious top text 2309SKY, yet the thread notes that top branding alone is not decisive because genuine Microchip-sold parts may still carry the Atmel logo. [#21161543]

How can I tell whether an ATmega88PA-PU bought on AliExpress is a remarked or counterfeit chip?

You can tell by checking whether the package markings, signature, and electrical behavior agree. In the reported case, the bought part was sold as an ATmega88PA-PU with top date 1910, but its readable signature was 0x1E930A and the lower-case markings showed much older dates, 0936 and 0808. One chip was also dead, and the other had code already loaded and fuse bits changed. That combination points to a used or re-marked AVR rather than a clean new ATmega88PA-PU. [#21161408]

Why do the top and bottom case markings on an ATtiny or ATmega DIP microcontroller sometimes show different production dates?

They can differ because the top was repainted while the lower markings were left original. The thread’s suspect ATmega88 showed a top code of 1910 but lower markings of 0936 and 0808, which cannot describe one fresh 2019 package consistently. A second explanation exists for newer genuine parts: some later Microchip-sold AVRs no longer carry the same bottom information style, so absence of bottom printing alone does not prove fraud. [#21161408]

What is the AVR signature row, and what kind of data is stored there besides the 3 signature bytes?

The AVR signature row is a separate memory area that stores identification and factory data beyond normal user flash. "Signature row" is a memory region that identifies the MCU, stores calibration or manufacturing data, and can expose bytes outside the usual 3-byte signature field. In the thread, participants described extra bytes that map to lot-number characters, wafer number, and X/Y die coordinates on some AVR families. Datasheet tables for some parts explicitly describe these addresses. [#21169245]

How do I use acetone and visual inspection to check whether ATmega or ATtiny package markings are original or repainted?

Use acetone as a quick surface-marking test, then inspect for sanding marks. 1. Wipe a small area of the top text with acetone. 2. Check whether the printed legend smears or disappears. 3. Look for grinding traces under the removed paint and compare top and bottom markings. In the reported case, the false top legend wore off under acetone, while the original lower marking remained and grinding marks became visible. The thread states original laser text should not dissolve in solvent. [#21161408]

What tests can distinguish an ATmega88, ATmega88A, and ATmega88V when the programmer reports the same signature family?

Use clock-and-voltage limits, because the shared signature family does not separate those variants. The thread says ATmega88V should not run above 10 MHz, so a stable 20 MHz test at 4.5–5.5 V rules out the V version. Then test 4 MHz at 1.8 V: ATmega88A should still work, while plain ATmega88 should only work from 2.7 V upward. Current draw can also help at 8 MHz and 5.0 V: about 5 mA for ATmega88 versus about 4 mA for ATmega88A under datasheet-like conditions. [#21161408]

Which voltage and clock tests help identify whether a DIP AVR is really an ATtiny13, ATtiny13A, or ATtiny13V?

Use the operating-limit tests, because the signature 1E 90 07 is shared across ATtiny13, ATtiny13V, and ATtiny13A. The thread recommends a 20 MHz / 5 V test first to rule out the V version. Then run a 4 MHz / 1.8 V test: ATtiny13A should work there, while plain ATtiny13 should require at least 2.7 V. That makes low-voltage behavior a stronger identifier than signature bytes when the package text looks suspicious. [#21161583]

Why can extra bytes appear when reading beyond the documented AVR signature and calibration addresses, and what do those bytes mean?

They appear because some AVRs expose a larger hidden flash or signature area than the short programmer commands suggest. One participant reconstructed bytes from an ATtiny2313 and found ASCII data matching the package string "6F7011" stored at offsets 14–19. Another participant said those extra bytes are serial-number bytes and can also be read through I/O addresses 0xF0–0xF8 on some devices. So the extra data is not random in every case; it can hold manufacturing identity fields. [#21164069]

What is a silicon errata document, and how can it help when checking AVR microcontroller revisions or hidden memory behavior?

A silicon errata document lists known chip-revision faults, hidden behaviors, and workarounds that the main datasheet may not explain clearly. "Silicon errata" is a revision-specific manufacturer document that records design bugs, undocumented behavior, and approved workarounds for a chip family. In the thread, errata was used to confirm that hidden signature-related bytes could disappear during serial programming on older AVR parts, and to show that revision differences do not automatically mean a chip is fake. [#21164667]

How can the serial number or lot number stored in hidden AVR flash memory be compared with the case markings to spot suspicious chips?

Read the hidden signature-area bytes, decode the lot string, and compare it to the lower package marking. In one counterfeit-looking ATmega from the thread, the lower case showed 8J6671, and the hidden memory contained the matching sequence "8J6671", which suggested the lower marking was original. In another ATmega328P-PU from a local shop, the package showed A9B98A but hidden memory showed 7U5533. A mismatch like that is suspicious, but the thread does not prove it always means fraud. [#21170551]

ATmega88A vs ATmega88PA vs ATmega328P — which differences matter most when buying old DIP AVRs from unknown sellers?

The important differences are operating limits, hidden identification consistency, and marketability of the part being resold. The thread’s suspect device was sold as ATmega88PA-PU, but electrical testing suggested an older ATmega88A die, not the advertised part. For ATmega88-family chips, 1.8 V, 2.7 V, and 10 MHz limits help separate variants. For ATmega328P, hidden lot data can add another check, although one local-shop chip still showed a serial mismatch without clear proof of counterfeiting. [#21161408]

Why do some genuine Microchip-sold AVR chips still carry the Atmel logo years after the acquisition?

They still carry it because Microchip kept using Atmel branding on some AVR parts after the acquisition. One participant explicitly reported genuine chips bought directly from Microchip with a production date from 2021 that still had the Atmel logo. Another reply noted that the trademark still belonged to the company and helped users distinguish AVR parts from PIC devices. So an Atmel logo on a later chip is not, by itself, evidence of a fake. [#21161592]

What does the 'Device Signature Imprint Table' in AVR datasheets describe, and how do lot number, wafer number, and X/Y coordinates map there?

It describes how bytes in the hidden signature area map to manufacturing-identification fields. The thread says some datasheets, including ATtiny24/44/84 and ATtiny441/841, include a Device Signature Imprint Table listing addresses for signature bytes, calibration bytes, and internal-use fields. A later thread update cites official descriptions where 0x0E–0x13 hold the 6-character lot number, 0x15 holds wafer number, and 0x16–0x17 hold Y and X coordinates. That mapping explains why package markings can sometimes be cross-checked against memory. [#21187371]

How could signature, calibration, or hidden serial-number bytes in older AVR chips be erased during ISP programming?

They could be erased by a timing mistake around chip erase in serial programming mode. The thread quotes an errata note stating that if signature bytes are read before a Chip Erase command fully completes, the signature may be erased, causing device ID and calibration bytes to disappear. Another post also mentions an undocumented AC C0 command, one bit away from Chip Erase = AC 80, that can erase the extra flash page. That makes hidden-ID checks useful, but not infallible. [#21164667]

What are the risks of buying very cheap DIP AVR microcontrollers from AliExpress instead of Mouser or official Microchip distributors?

The main risks are receiving dead parts, remarked old stock, pre-programmed chips, and devices with altered fuses. In the thread, a purchase priced at under 6 zł per piece yielded 1 chip with internal pin damage and 1 used chip with unexpected code already loaded. Another comment notes that the cheap listing was about half the price of Mouser, which creates a strong incentive to relabel slow-moving or unwanted stock. Official distribution costs more, but it sharply reduces that uncertainty. [#21161408]

Why might a genuine-looking ATmega328P-PU from a local shop have a hidden serial number that does not match the number printed on the package?

It might be a production mismatch, a revision difference, or a suspicious part that still passed retail channels. The thread reports an ATmega328P-PU bought from a Polish shop whose top and bottom dates agreed, yet the lower case showed A9B98A while hidden memory showed 7U5533. A later reply warns that revision and documentation differences can exist without proving a fake. So the mismatch is a red flag, but not a standalone verdict without other checks. [#21170551]

Generated by the language model.