Просмотр сообщений
1
Просмотр сообщений
В этом разделе можно просмотреть все сообщения, сделанные этим пользователем.
3
Картриджи / диски / Re: FakeGirl (852 in 1)
« : 30 Ноябрь 2025, 14:18:07 »А была такая версия 1.2 ? Я собирал версию 2.
(Ссылка на вложение)
i'd like to say: the pcb shape is like a pig
4
Картриджи / диски / Re: помогите добавить игры
« : 16 Ноябрь 2025, 17:37:44 »
// Mapper #446, #468 - Mindkids, BlazePro
// Mapper #468 - EPM240
// Mapper #446 - EPM570
// Mapper #468 - EPM240
// Mapper #446 - EPM570
5
Картриджи / диски / Re: Дампинг, обсуждаем, делимся ромами своих дампов
« : 29 Октябрь 2025, 08:52:00 »
Life on Mars
Life on Earth
Secret of the Four Winds
Coloco DX
These are actually the same mapper type, which is Kai Magazine(or Kaiser Wave).
$3F0000 - $3FFFFF are self-programming area
$000000 - $00FFFF are CFI command area
They use 45 bytes of CFI query data as copy protection. Writing $98 to ($000055 << 1) enables CFI query, and writing $F0 to ($000000 << 1) disables CFI query.
If a different FLASH is used or if writing commands to FLASH CFI is disabled, a warning will appear during boot.Of course, there is a method to bypass this mechanism by providing a set of fake CFI query data to deceive the protection.As for reading/saving data to $3Fxxxx, there is also a method to map it to SRAM, ignoring CFI write commands in this address space. Since CFI commands are in the $3F0XXX address range and the saved data is in the $3FFXXX address range, mapping $3FXXXX directly to SRAM without any processing is perfectly fine.
Life on Earth
Secret of the Four Winds
Coloco DX
These are actually the same mapper type, which is Kai Magazine(or Kaiser Wave).
$3F0000 - $3FFFFF are self-programming area
$000000 - $00FFFF are CFI command area
They use 45 bytes of CFI query data as copy protection. Writing $98 to ($000055 << 1) enables CFI query, and writing $F0 to ($000000 << 1) disables CFI query.
If a different FLASH is used or if writing commands to FLASH CFI is disabled, a warning will appear during boot.Of course, there is a method to bypass this mechanism by providing a set of fake CFI query data to deceive the protection.As for reading/saving data to $3Fxxxx, there is also a method to map it to SRAM, ignoring CFI write commands in this address space. Since CFI commands are in the $3F0XXX address range and the saved data is in the $3FFXXX address range, mapping $3FXXXX directly to SRAM without any processing is perfectly fine.
6
Картриджи / диски / Re: Дампинг, обсуждаем, делимся ромами своих дампов
« : 22 Октябрь 2025, 16:05:40 »
A new English translated version of Daiku no Gen San 2(AKA: Hammerin's Harry 2) which was published by Retro-Bit recently.
The cart must be manufactured in some Chinese company - maybe mindkids - by analyzing its mapper
Mapper #446, so familiar, and I extraced it and got the orginal ROM of mapper #65.
Here u r:
The cart must be manufactured in some Chinese company - maybe mindkids - by analyzing its mapper
Mapper #446, so familiar, and I extraced it and got the orginal ROM of mapper #65.
Here u r:
7
Игры на консолях 1-4 поколений / Re: [SMD] Paprium, новая игра для мегадрайва
« : 27 Сентябрь 2025, 06:27:47 »
Maybe many people are trolling you:
https://www.neogaf.com/threads/paprium-reverse-engineered-dumped-and-made-playable-on-retroarch-the-scam-is-over.1686093/
Often, things that are believed to be true are actually wrong.Most people think that level shifter has much lower noise than resistor array, but it is not true. Level shifters generate surge at the moment of output flipping, which can interfere with the data bus. These surges are sometimes unavoidable. In FPGA design, there are usually "slew rate" options to reduce switching noise, but for level shifters, there's nothing to do.
Although resistors will limit the maximum frequency the system can respond to. However, for retro game consoles, which are relatively low-speed platforms (<10 MHz), this is not an issue. Using resistor array actually has much lower noise.This explains why on the real hardware, Paprium can ealily experience program crashes and jump to error register display screens after running for a period of time (about half an hour).I assume some people are only proficient in software but not really good at hardware design.
On the other hand, exceeding the 32Mb addressing range is also one of the causes of compatiblity issues that I have mentioned before.
https://www.neogaf.com/threads/paprium-reverse-engineered-dumped-and-made-playable-on-retroarch-the-scam-is-over.1686093/
Often, things that are believed to be true are actually wrong.Most people think that level shifter has much lower noise than resistor array, but it is not true. Level shifters generate surge at the moment of output flipping, which can interfere with the data bus. These surges are sometimes unavoidable. In FPGA design, there are usually "slew rate" options to reduce switching noise, but for level shifters, there's nothing to do.
Although resistors will limit the maximum frequency the system can respond to. However, for retro game consoles, which are relatively low-speed platforms (<10 MHz), this is not an issue. Using resistor array actually has much lower noise.This explains why on the real hardware, Paprium can ealily experience program crashes and jump to error register display screens after running for a period of time (about half an hour).I assume some people are only proficient in software but not really good at hardware design.
On the other hand, exceeding the 32Mb addressing range is also one of the causes of compatiblity issues that I have mentioned before.
8
Картриджи / диски / Re: Дампинг, обсуждаем, делимся ромами своих дампов
« : 21 Сентябрь 2025, 18:56:25 »
Mapper #268 - Coolboy, Mindkids AA6023
Submapper 0/1 - AA6023A
Submapper 2/3 - AA6023B
The only differences between the two are:
prg_base[22:20]
prg_mask[14]
If there is no menu loader at all, or even if there is a menu loader but it does not access the register 5001/6001 (default power-on state), then the two are compatible.
Submapper 0/1 - AA6023A
Submapper 2/3 - AA6023B
The only differences between the two are:
prg_base[22:20]
prg_mask[14]
If there is no menu loader at all, or even if there is a menu loader but it does not access the register 5001/6001 (default power-on state), then the two are compatible.
9
Картриджи / диски / Re: Дампинг, обсуждаем, делимся ромами своих дампов
« : 20 Сентябрь 2025, 20:30:41 »Попалась на глаза мне вот эта одноигровка:
https://aliexpress.ru/item/1005002165797973.html
(Ссылка на вложение)
(Ссылка на вложение)
и значит решил я её сдампить. Смотрю на плату а на ней квадратная микросхема SMD133 и еще две микросхемы CS18LV20483 и S29GL064N.
Начинаю гуглить SMD133 это маппер MMC3, S29GL064N - 64 mbit флэш-память, CS18LV20483 - 256k SRAM-память. При попытке сдампить под MMC3 ничего не работает, помучался немного и полез гуглить как эта микросхема SMD133 работает и наткнулся на маппер 268 на сайте nesdev.org, еще раз внимательно разглядываю картридж и вижу надпись на нём Mindkids, вижу в таблице, что это маппер 268.0 или 268.1, дамплю как --mapper 268.0. Открываю его в Fceux 2.6.6 и он работает! Надо было додуматься еще одноигровку на нём сделать.
Here you are
Rush'n Attack (U) [Title Conversion][hM4][t1] (Extracted).zip (59.7 КБ - загружено 257 раз.)Another that I've stripped the coolboy loader, also compatible with standard mapper #4
Rush'n Attack (U) [Title Conversion][hM268][t1] (Trimmed).zip (238.19 КБ - загружено 266 раз.)10
Картриджи / диски / Re: Проблемы с клоном флэшкатриджа everdrive md
« : 19 Сентябрь 2025, 16:28:39 »
Try removing the battery and wait one minute before reinstalling it.Some setting parameters are written into NVRAM (SRAM). For some reason, parameters in the NVRAM may be changed, which can cause abnormal startup or failure to load games or other faults.The battery discharge method is usually used as a debugging method to clear abnormal NVRAM and restore the loading of default parameters.It's just like when a computer motherboard's BIOS is modified and cannot restart, requiring the battery to be removed for discharge.
If the above can solve the problem, it indicates a soft fault in the NVRAM that can be restored.
However, it is also possible that an error occurred during the erasing and programming procedure, causing the system to freeze and it will never reach the completion of the progress bar. As a result, the user forces a shutdown.Once this happens, the entire BIOS of the flashcart has been corrupted, leading to partial functionality failure or even completely bricking the device.
If the above can solve the problem, it indicates a soft fault in the NVRAM that can be restored.
However, it is also possible that an error occurred during the erasing and programming procedure, causing the system to freeze and it will never reach the completion of the progress bar. As a result, the user forces a shutdown.Once this happens, the entire BIOS of the flashcart has been corrupted, leading to partial functionality failure or even completely bricking the device.
11
Картриджи / диски / Re: Famicom Cartridge Schematics & PCBs
« : 12 Сентябрь 2025, 11:49:22 »Теперь понял, никакой магии значит нет
Дальше видимо китайцы будут продавать пустые корпуса с наклейками.
Хотя, в принципе решение интересное. Если на консоли большая часть встроенных игр - хлам, но есть поддержка ммс3, то можно записать нормальные ромы.
Но вопрос по перезаписи такого картриджа пока открытый, вернее даже по сборке рома под него.
Based on VR Tech's built-in mapper specifically for famiclone, it integrates the mapper chip and the CRAM into the console. Of course, it is also possible to embed the cartridge ROM into it, which is no longer surprising (for example, the p&p handheld console).Perhaps a blank PCB could be sold, with only one line connected to a specific slot pin. By connecting Vcc/Gnd, it can switch between the internal cart and the external cart.
12
Ромхакинг / Хаки на любой вкус
« : 03 Сентябрь 2025, 19:22:59 »Я бы и рад, но на моём флеш картридже она нивкакую не хочет запускаться. У меня не оригинальный эвердрайв, а китайский аналог. Я поэтому и подумал что мало ли у кого то найдется хороший ром. В любом случае спасибо за ответ! Хотя бы не буду тешить себя надеждами)
Unfortunately, edn8 can only support a maximum PRG size of 512K, and edn8 pro also has a limitation of PRG size of 1M
It seems that it can currently run on a homebrew cartridge.
14
Картриджи / диски / Re: FakeGirl (852 in 1)
« : 28 Август 2025, 12:00:36 »
Street Fighter II/Master Fighter II can not be converted to standard MMC3(#4), but thers is a variant that mapper #208 should be all right
15
Картриджи / диски / Re: FakeGirl (852 in 1)
« : 26 Август 2025, 17:41:06 »
Can EPM240 really contain mapper1-4, 7, 9-11(???), VRC2-4(???), Taito(???), Sunsoft(???)
??? - I can not even belive it
??? - I can not even belive it
16
Картриджи / диски / Re: Дампинг, обсуждаем, делимся ромами своих дампов
« : 23 Август 2025, 10:54:58 »doragasu
S29GL064: Spansion Nor-flash 64Mbit
YX5200: 悦欣 sigle chip mp3 decoder(www.yxin18.com)
HCT377: Latch, Octal D-type flip-flop
The dumpped 4MB ROM may not be the full data, the full ROM I guess should be the size of 8MB. Despite having almost half of the data redundancy, the data at certain addresses is different. In principle, a RAM chip is needed, but by doubling the redundant data and using a latch to control the switching of the highest address bit, cost reduction can be achieved.
17
Игры на консолях 1-4 поколений / [SMD] Paprium, новая игра для мегадрайва
« : 24 Июль 2025, 07:18:53 »
Yes, as I have always said, the original Paprium is fragile in terms of compatibility with consoles: it can only run on a very limited number of (MD2???) models. Unfortunately, this is determined by the CART_CE generation logic of the console, which may not be fixable through the cartridge, and the same goes for flashcarts.
The fundamental reason for all this is that Watermelon tried to use the entire 80Mb address space, while different models of consoles have inconsistent logical decoding for address spaces larger than 32Mb. Neofid Studio, is much more cautious, they access only 32Mb and use bank switching techniques.
The fundamental reason for all this is that Watermelon tried to use the entire 80Mb address space, while different models of consoles have inconsistent logical decoding for address spaces larger than 32Mb. Neofid Studio, is much more cautious, they access only 32Mb and use bank switching techniques.
18
Игры на консолях 1-4 поколений / [SMD] Paprium, новая игра для мегадрайва
« : 14 Июль 2025, 17:10:13 »
These nipple robots are perverts dressed in latex bodysuits, quite vulgar. 
19
Картриджи / диски / Re: Flash-картридж для SEGA из Китая
« : 12 Июль 2025, 14:16:50 »https://aliexpress.ru/item/1005007853927832.html
Китайцы уже торгуют клоном MEGA EVERDRIVE X7 от KY технолоджи
Мда, только вот на шифтерах они решили таки сэкономить, добавив вместо них тупо резисторы для согласования уровней.
Ну и куда же без подсветОчки
10 white LEDs, averaging at least 10mA current each, according to spec, they can operate at a current of 100mA. Specifically check the output current of the LED driver chip. In principle, there are two ways to connect: series or parallel; series is the traditional backlight for LCD screens, while parallel would require multiple current source chips. It is estimated that the most likely implementation would be in series, inputting the instruction signal of SD card accessing, and then driving 10 white LEDs. I don't understand what benefits this approach has besides causing more light pollution and unnecessary current consumption.
20
Игры на консолях 1-4 поколений / [SMD] Paprium, новая игра для мегадрайва
« : 12 Июль 2025, 04:31:39 »
Hacking the original ROM to strip everything from the ARM processing seems to involve a lot of work, but it is not impossible. Music cannot be decoded, waveform sound effects, voice, etc., also seem to be unachievable or require address remapping. Sprites and tiles also need address remapping. The scaling of sprites also needs to be removed.
A block of RAM needs to be added to handle all of these, because on the original cartridge it is written by ARM and read by the console. If we strip away ARM, which will perform the writing? Alternatively, we can add a loader to preprocess the writing of this RAM. Or we could directly expand banks, treating this already processed block of data as a static bank section, and then switch banks via monitoring addresses.
A block of RAM needs to be added to handle all of these, because on the original cartridge it is written by ARM and read by the console. If we strip away ARM, which will perform the writing? Alternatively, we can add a loader to preprocess the writing of this RAM. Or we could directly expand banks, treating this already processed block of data as a static bank section, and then switch banks via monitoring addresses.
21
Игры на консолях 1-4 поколений / [SMD] Paprium, новая игра для мегадрайва
« : 11 Июль 2025, 13:41:05 »ROM INFO
=====================================
GameID:1618760926
Name: PAPRIUM (c) WATERMELON GAMES
Console Name: SEGA MEGA DRIVE_
Type: GM
Version: T-574120-00
Copyright: (c)T574 2018.DEC
Countries: JUE
Description: DT__ HDC+MEGAWIRE 4.0 JUE!
Checksum: 42491
IO Support: JC64
box cover:
22
Игры на консолях 1-4 поколений / [SMD] Paprium, новая игра для мегадрайва
« : 10 Июль 2025, 21:15:21 »
Some foolish blind followers on the nesdev forum even believe that combining 3.3V flash memory with level shifters COULD/SHOULD provide better reliability.
Just look at this board of Pap-rium(a lump of pap)
, it perfectly meets their requirements and taste, but the compatibility is terrible. The electrical performance is even worse than Watermelon’s previous work - Pier Solar - at least the latter didn’t use any level shifter at all, just resistor array.
Just look at this board of Pap-rium(a lump of pap)
, it perfectly meets their requirements and taste, but the compatibility is terrible. The electrical performance is even worse than Watermelon’s previous work - Pier Solar - at least the latter didn’t use any level shifter at all, just resistor array.23
Игры на консолях 1-4 поколений / [SMD] Paprium, новая игра для мегадрайва
« : 10 Июль 2025, 13:44:54 »
MAME's indirect emulation of music playback provides an idea for the reproduction of physical carts, it is not necessary to completely reverse engineer, forgetting the original design where ARM acts as a music synthesizer (VGM player???), and directly playing back recorded tracks.
Replicas may soon be available.
Replicas may soon be available.
24
Игры на консолях 1-4 поколений / [SMD] Paprium, новая игра для мегадрайва
« : 10 Июль 2025, 07:32:11 »
Why do some retro game cartridges need to add internet access features?
To put it nicely, it's for uploading scores and global rankings, and for downloading updated content.
To put it bluntly, it's to install backdoors for modifying your cartridge, for user location data, and to identify who the user is.
Why do developers need this data? Ask Boeing.
To put it nicely, it's for uploading scores and global rankings, and for downloading updated content.
To put it bluntly, it's to install backdoors for modifying your cartridge, for user location data, and to identify who the user is.
Why do developers need this data? Ask Boeing.
25
Игры на консолях 1-4 поколений / [SMD] Paprium, новая игра для мегадрайва
« : 09 Июль 2025, 15:07:46 »Do you want to tell, that is custom audio chip is fiction too and it just plays wawe files directly to the cart port?
So, that makes current emulation state is almost perfect.
The ARM is used to decode MP3 and then output it to the audio interface of the cart socket, of course, a port address must be defined to give instructions to the ARM to play which track. That's all. It's not difficult to guess. The low-bit encoded music is stored in a 64M SPI flash.
Those audio-out interfaces of the genesis/md clone without any connecting wires will have no background music. This further increases the return rate. The developer must be a tech enthusiast and quite stubborn, completely disregarding the differences among the numerous hardware available in the market.
ARM does two things:
1. Acts as a flasher to write the ROM image from SPI flash into PSRAM, where different data will be written according to different STAGES.
2. Acts as an MP3 player.
The mini-game also does two things:
1. Acts as a bootloader. To speed up game loading, this bootloader is written first, and after loading the bootloader, ARM acts as a flasher, then marks the status flag. When the next reset occurs, it directly jumps to the formal game entry vector.
2. Prevents being dumped. Only on physical consoles can the bootloader complete the data loading of PSRAM; if a dumper is used, it will only be able to dump this bootloader, which is this low-resolution mini-game.
26
Игры на консолях 1-4 поколений / [SMD] Paprium, новая игра для мегадрайва
« : 09 Июль 2025, 11:08:25 »
The hardware partner of the French clearly have great ambitions; they even took full advantage of ARM by connecting the UART port and coined a specialized term called "mega-wire" to confuse players.In reality, it's just a communication cable, possibly UART to USB, connecting to a computer to read and write to SRAM, such as uploading high scores or death counts for global rankings. It could also be used to replace MP3 audio files, allowing for the playback of custom BGM.The hardware design is full of fantasies, but cost control is terrible, and the compatibility is a nightmare. This has led to serious issues in later mass production. The higher the cost, the heavier the after-sales burden.
The compatibility issues remain unresolved, which stems from the use of 64Mbit and 3.3V flash. Many Genesis/MD systems exhibit different logic when handling addresses over 32Mbit (for example, the cart CS logic of model1 and model2 is slightly different). As a result, they dare not proceed with mass production and can only supply in small quantities. Even so, they still received a large volume of returns and exchanges, and the high exchange rate further pressured them to refrain from continuing supply. Consequently, angry buyers began requesting refunds, and PayPal froze their receiving account.
The compatibility issues remain unresolved, which stems from the use of 64Mbit and 3.3V flash. Many Genesis/MD systems exhibit different logic when handling addresses over 32Mbit (for example, the cart CS logic of model1 and model2 is slightly different). As a result, they dare not proceed with mass production and can only supply in small quantities. Even so, they still received a large volume of returns and exchanges, and the high exchange rate further pressured them to refrain from continuing supply. Consequently, angry buyers began requesting refunds, and PayPal froze their receiving account.
27
Игры на консолях 1-4 поколений / [SMD] Paprium, новая игра для мегадрайва
« : 09 Июль 2025, 10:17:43 »
The initial $70 investment, if you were lucky enough to receive the physical product, you would find that the physical cost is $20, plus the necessary development, marketing expenses, and profits, bringing a total of $60 which is reasonable. Pricing at $70 is not unacceptable; for a design with an integrated co-processor (ARM), it is also worth buying for hardware enthusiasts.
However, they seem very timid and insecure in processing the $10 difference (70-60=10). They were afraid that the hardware would be cracked and the cartridges could be copied, so guess what they did? They ground the chips and covered them with epoxy resin (oh, it's not even hardened epoxy resin, but a kind of black hot melt glue) to make the overall appearance look beautiful and elegant, and then covered it with a piece of white copper, resembling the design of a CPU lid, deceiving players into thinking it was a custom chip. Due to the hot melt glue, many players soon experienced issues with the lids falling off, and even the hardened glue melted and overflowed from the PCB when heated during operation. Unless you treat it as an investment piece, never play it, keep the original packaging and unopened, and then sell it at a high price on auction sites.
The Datenmeister chip? It is just a lump of black poop surrounding the ARM and CPLD, with a piece of white copper attached. If they could use better glue to keep the lid from falling off, at least 'The Emperor's New Clothes' would have a garment.
However, they seem very timid and insecure in processing the $10 difference (70-60=10). They were afraid that the hardware would be cracked and the cartridges could be copied, so guess what they did? They ground the chips and covered them with epoxy resin (oh, it's not even hardened epoxy resin, but a kind of black hot melt glue) to make the overall appearance look beautiful and elegant, and then covered it with a piece of white copper, resembling the design of a CPU lid, deceiving players into thinking it was a custom chip. Due to the hot melt glue, many players soon experienced issues with the lids falling off, and even the hardened glue melted and overflowed from the PCB when heated during operation. Unless you treat it as an investment piece, never play it, keep the original packaging and unopened, and then sell it at a high price on auction sites.
The Datenmeister chip? It is just a lump of black poop surrounding the ARM and CPLD, with a piece of white copper attached. If they could use better glue to keep the lid from falling off, at least 'The Emperor's New Clothes' would have a garment.
28
Картриджи / диски / Re: Everdrive N8 (China ver) JACK DIY
« : 06 Июль 2025, 04:19:44 »
If your N8 is the classic short board version, absolutely do not upgrade to 1.26.Upgrading to 1.26 will cause the nvram of the OS to be mismatched, resulting in abnormal settings and even making it unable to boot. You will only have to open the case, remove the battery, and wait for the capacitor to discharge in order to clear the incorrect nv settings that have been overwritten.
29
Картриджи / диски / Re: Everdrive N8 (China ver) JACK DIY
« : 04 Июль 2025, 06:07:20 »The digital audio output amplitude of the FPGA is 0-3.3V, but in most cases, in order to achieve good resolution, it is always below 3.3/2. The APU outputs a DC voltage of around 5/2V. When there is only a small impedance between the two, the output of the APU is pulled down by the FPGA, which may result in a large current, and both interfere with each other, leading to deterioration in sound quality. Specifically, 3.3/2 continuously pulls down the 5/2 level, which also lowers the static operating point of the audio amplifier transistor in the subsequent stage.
What if we increase the resistance between the two? Replacing the 2.2ohm resistor with a larger one? For example, using a Mohm or Kohm level resistor would greatly reduce the interference with the APU output, but it could not solve the problem of the subsequent static operating point being pulled down. At this point, the FPGA still dominates the APU output, and the static operating point is similarly lowered, leading to weakened output from the audio amplifier transistor.
By programming the software, we can increase the FPGA digital output, raising the equivalent output volume of the expansion audio from the FPGA, which can significantly reduce interference with the APU output (this is similar to the approach where the software OS upgrade only retains the hi status option); however, this only improves the situation and does not completely resolve the issue.
Using an operational amplifier for matching of the FPGA digital audio output resolves the problem.
What if we do not use an operational amplifier? The simplest circuit approach would be to simply add an electrolytic capacitor to isolate the lower output level of the FPGA.
If you are very resistant to destructive modifications to hardware, then the following parameter changes may be suitable for you. Change the values of two resistors to isolate each other. However, these two values have certain computational considerations; their balanced voltage center should be closer to 5/2, so as not to cause additional effect on the original famicom/clone. The series resistance for the FPGA audio output should be larger, with a ratio of about 1:2 being appropriate, and values in the range of >470 Ohm, <10K Ohm should be acceptable.
30
Картриджи / диски / Re: Everdrive N8 (China ver) JACK DIY
« : 04 Июль 2025, 03:15:55 »
Between pin 45 and 46 of the cartridge, a direct connection is fine in a 5V system. However, on the flashcart platform, pin 46 has a digital output from the FPGA, and since the FPGA operates at 3.3V, a simple connection would pull down the 5V input at pin 45, causing clipping distortion. Therefore, there is usually a balancing resistor between pins 45 and 46, but this is not sufficient, as this resistor is typically designed to be quite small. It is better to add an electrolytic capacitor, with the R-C combination designed for a bandwidth of 10K. After adding the capacitor, it will no longer cause interference in volume and waveform from the subsequent stage to the previous stage.
The audio amplification designs of the Dendy and Famicom are different, and if pin 45 and 46 are directly connected (or isolated by a small resistor), the linearity will be very poor, resulting in significant differences on the two different platforms. The 3.3V not only pulls down the front-end APU audio output but also decreases the static operating point of the back-end audio amplifier transistor. Thus, the electrolytic capacitor is necessary. Unfortunately, many Dendy audio amplification circuits were not designed with an electrolytic capacitor to isolate the different levels of the front and back stages (at that time, a unified 5V system would not have caused any issues).
The audio amplification designs of the Dendy and Famicom are different, and if pin 45 and 46 are directly connected (or isolated by a small resistor), the linearity will be very poor, resulting in significant differences on the two different platforms. The 3.3V not only pulls down the front-end APU audio output but also decreases the static operating point of the back-end audio amplifier transistor. Thus, the electrolytic capacitor is necessary. Unfortunately, many Dendy audio amplification circuits were not designed with an electrolytic capacitor to isolate the different levels of the front and back stages (at that time, a unified 5V system would not have caused any issues).