1.54寸墨水屏跑出30帧 一个GameBoy模拟器背后的驱动魔法1.54-Inch E-Ink Screen Runs at 30fps: The Driver Magic Behind a GameBoy Emulator


AI free token/AI免费积分

本文来源:https://mp.weixin.qq.com/s/zGlxzVdwUXtEiU9hatIAhA
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。

一块 1.54 寸的黑白墨水屏,跑 Game Boy 模拟器,帧率 30 FPS。如果你用过 Kindle,就知道这个数字有多反直觉:翻一页都要闪一下的屏幕,怎么可能追得上马里奥的跳跃。

这是 Hackaday 本周报道的一个项目,作者在 YouTube 上叫 Bo Made It。它的趣味不在“墨水屏能玩游戏”,而在于作者怎么把电子纸公认的几个缺点,变成了为我所用的工具。

Game Boy 是个很巧妙的选择。任天堂 1989 年的那块灰绿色 passive matrix LCD,本来就慢、有残影、对比度低。这些毛病和墨水屏几乎一模一样,所以玩家对它的容忍度天然就高。用墨水屏去模拟一块“烂屏”,反而有种复古的贴切。

图片

硬件本身非常普通:一块 Waveshare 的 ESP32-S3-ePaper-1.54 开发板,属于“随手挑一块”的那种现成产品。屏幕自带 controller,里面预置了厂家的 waveform table,对开发者来说基本是一个黑盒。

熟悉这个圈子的人可能会问:今年 Wenting Zhang 不是已经做出 60 Hz 的墨水屏显示器了吗?确实如此,但两者的前提完全不同。Wenting 的项目直接驱动 panel 的裸接口,波形完全自己写;Bo 则是隔着一个闭源 controller 在做事,可动的旋钮少得多。

图片

打个比方,前者是拆掉发动机电脑重写程序,后者是只用油门和档位去跑出赛车成绩。Bo 也大方承认,这个项目的灵感直接来自 Wenting 之前做的墨水屏掌机 PaperBoy。

要理解 Bo 做了什么,先要知道墨水屏为什么慢。每个像素里是一堆带电的黑白颗粒,靠电压 pulse 推着上下移动。标准的驱动方式不管像素要变多少,都给它一整套完整的波形,稳妥但浪费。

Bo 的第一个发现是,即使隔着 controller,仍然可以单独寻址每一个像素。于是他给每个像素一个“预算”:最多 8 个 pulse,按它实际需要变化的幅度来分配。画面里大部分像素帧与帧之间根本不动,这部分时间就全省下来了,思路和视频编码只传帧间差异很像。

第二个技巧更有意思:用残影来做灰阶。Game Boy 有 4 级灰度,而墨水屏的经典烦恼之一,就是像素没有被完全推到位时会停在中间态,留下 ghosting。Bo 反过来用:如果一个像素只需要变亮或变暗一级,就故意不把它推完,让它停在那个灰色的中间位置。

这个做法的代价是转换变得不对称。从黑到“白”要走很长的路,从“白”到下一级灰却只是一小步;4 级灰之间有 12 种不同的转换,每一种的 pulse 数和时序都得单独确定。

确定这些参数的办法,是整个项目里最“工程”的部分。Bo 用摄像头拍下屏幕在不同驱动参数下的实际表现,从画面里读出真实灰度,再反推每种转换该用多少 pulse。这是一个闭环 calibration,让耗时的手工调参变成半自动的流程。

图片

Hackaday 评论区里有人点出了一个关键问题:温度。墨水屏颗粒在液体里的移动速度对温度非常敏感,商用 controller 的 waveform 本来就是按温度分段存储的。一套在室温下精心标定的 pulse 表,到了冬天的户外或被手捂热之后,很可能就不准了。

所以这个项目要走向实用,大概率还得加上温度测量,并把标定做成随温度插值的多维表。目前标定仍在进行,代码还没放上 GitHub,作者表示完成后会全部开源,并把这套 display driver 用在一个类似 PDA 的设备上。

那么,是不是任何墨水屏,换一个 custom driver 就能高刷?答案是否定的。墨水屏的 TFT backplane 扫描本来就能到几十赫兹,慢的是一次完整转换需要连续很多个 pulse。所谓高刷,就是不再等转换做完,每次扫描都接受新画面,代价是残影和拖尾。

要做到这一点,得同时满足几个条件。一是能拿到逐像素、逐扫描的控制权,要么直接驱动裸接口,要么 controller 开放自定义 LUT 和 partial update。二是接口带宽够用,小屏走 SPI 没问题,大尺寸高分屏就吃不消。三是墨水本身要快:黑白电泳屏和在其上叠彩色滤光层的方案可以,而靠多种颜料颗粒出色的彩色电子纸,一次刷新就要十几秒,软件无能为力。

再加上温度和面板寿命的约束,高刷模式通常还要定期插入一次完整刷新来清除累积残影。所以 custom driver 是钥匙,但门后有没有房间,要看面板的物理特性和 controller 开放到什么程度。

评论区还有另一种声音:墨水屏明明这么能打,货架上的产品却只拿来当价签,厂家是不是在挤牙膏式地挤利润。这个判断只对了一半。

一个爱好者可以对一块 panel 做精细标定,可以接受一点残影,也不必关心屏幕五年后的状态。量产却要面对 panel 之间的个体差异、从零下到三十多度的使用环境、长期驱动对寿命的影响,以及普通用户对残影几乎为零的容忍度。厂家的 waveform 偏保守,很大程度上是这些约束的结果。

但这类 hack 确实揭示了一个方向:按内容、按像素自适应地分配驱动,再配合自动化标定,墨水屏的刷新潜力远没有被挖尽。从 Wenting 的 60 Hz,到今年稍早那个 20 FPS 的 hack,再到这次的 30 FPS,一个开源社区正在一点点重写人们对“墨水屏只能显示静态内容”的印象。

等 Bo 的代码开源,最值得看的是他的 pulse 分配算法和标定流程。如果这套方法能在更多现成开发板上复现,那么更流畅的墨水屏小设备,可能比我们想象的来得更早。

(本文由 Claude Opus5.5 生成。文章prompt: 解读一下这个项目,写成一篇公众号文章 https://hackaday.com/2026/10/14/tiny-e-paper-display-delivers-30-fps-of-game-boy-fun/。文章里解读一下这个问题:是否任何墨水屏都可以做成高刷?是只要有所谓的 custom driver 即可吗?)

apov读者群管理员微信:weilaifuture65
利益相关:无
相关文章
离大谱,为电子纸开发游戏
墨水屏文字游戏机 Ink Console 停止众筹
Modos Flow 达成众筹目标/开发者介绍墨水屏显示器开发历程
在越南给阅星曈刷“鸡”?
M5 Paper S3 即将迎来复刻 + 升级?

 

Article source: https://mp.weixin.qq.com/s/zGlxzVdwUXtEiU9hatIAhA If you have any copyright concerns, please contact us at 13951431913#139.com or leave a comment on the right.

30 FPS on a 1.54-inch E-ink Display: The Driver Magic Behind a GameBoy Emulator

1.54-inch E-ink Screen Hitting 30 FPS: The Driver Magic Behind a Game Boy Emulator

Original Video Link: https://www.youtube.com/watch?v=XiZmM6hwwAU

A 1.54-inch black-and-white e-ink screen running a Game Boy emulator at 30 FPS. If you've used a Kindle, you know how counterintuitive this number is: a screen that flashes when turning a page — how could it possibly keep up with Mario's jumps?

This is a project covered by Hackaday this week, and the author goes by Bo Made It on YouTube. The fun isn't in "e-ink can play games," but in how the author turned the well-known drawbacks of electronic paper into tools at his disposal.

The Game Boy is a clever choice. Nintendo's 1989 gray-green passive matrix LCD was inherently slow, had ghosting, and low contrast. These flaws are nearly identical to e-ink, so players naturally have a high tolerance for them. Using e-ink to emulate a "lousy screen" creates a refreshing retro authenticity.

图片

The hardware itself is quite ordinary: a Waveshare ESP32-S3-ePaper-1.54 development board, the kind of pre-made product you'd grab off the shelf. The screen comes with its own controller, with the manufacturer's waveform table pre-flashed — essentially a black box for developers.

Those familiar with this scene might ask: Didn't Wenting Zhang already make a 60 Hz e-ink display this year? Indeed, but the prerequisites are entirely different. Wenting's project drives the panel's bare interface directly, writing waveforms from scratch; Bo works through a closed-source controller, with far fewer knobs to turn.

图片

To put it bluntly, the former is rewriting the engine computer's program after tearing it out, while the latter is racing to set records using only the throttle and gears. Bo freely admits that this project's inspiration came directly from Wenting's earlier e-ink handheld PaperBoy.

To understand what Bo accomplished, we first need to know why e-ink is slow. Each pixel contains a cluster of charged black-and-white particles that move up and down under voltage pulses. The standard driving method delivers a complete waveform sequence regardless of how much the pixel needs to change — safe but wasteful.

Bo's first discovery was that even through the controller, every pixel can still be addressed individually. So he gives each pixel a "budget": at most 8 pulses, allocated according to how much the pixel actually needs to change. Most pixels in a frame don't move at all between frames, saving that time entirely — the approach is similar to video codecs that only transmit inter-frame differences.

The second technique is even more interesting: using ghosting to create grayscale. The Game Boy has 4 levels of grayscale, and one of e-ink's classic headaches is that pixels not fully driven will stop in an intermediate state, leaving ghosting. Bo flips this around: if a pixel only needs to brighten or darken by one level, he deliberately stops short and lets it rest at that gray midpoint.

The cost of this approach is asymmetric transitions. Going from black to "white" requires a long path, while going from "white" to the next gray level is just a short step. There are 12 different transitions among the 4 grayscale levels, each requiring its own pulse count and timing to be determined separately.

The method for determining these parameters is the most "engineering" part of the entire project. Bo uses a camera to photograph the screen's actual performance under different driving parameters, reads the real grayscale from the images, then reverse-engineers how many pulses each transition should use. This is a closed-loop calibration, turning laborious manual tuning into a semi-automated process.

图片

Someone in the Hackaday comments section flagged a key issue: temperature. The movement speed of e-ink particles in liquid is extremely sensitive to temperature, and commercial controller waveforms are already stored in temperature-segmented tables. A pulse table carefully calibrated at room temperature may become inaccurate outdoors in winter or after being warmed by hand.

So for this project to become practical, it will likely need temperature sensing and calibration built into a multi-dimensional table with temperature interpolation. Calibration is still underway, and the code hasn't been pushed to GitHub yet. The author says he'll open-source everything once complete, and apply this display driver to a PDA-like device.

So, can any e-ink screen achieve high refresh rates with a custom driver? The answer is no. E-ink TFT backplane scanning can already reach dozens of hertz; what's slow is that a complete transition requires a continuous sequence of many pulses. True high refresh means no longer waiting for transitions to finish, accepting new frames on every scan — at the cost of ghosting and trailing.

Achieving this requires meeting several conditions simultaneously. First, you need per-pixel, per-scan control — either by driving the bare interface directly or having the controller expose custom LUTs and partial updates. Second, the interface bandwidth must suffice: SPI works fine for small screens, but large high-resolution panels can't keep up. Third, the ink itself must be fast: black-and-white electrophoretic displays and color-filter-over-black-and-white schemes are viable, while color e-ink relying on multiple pigment particles takes over ten seconds per refresh — software can't help there.

Add in temperature and panel lifespan constraints, and high-refresh modes typically need periodic full-refresh inserts to clear accumulated ghosting. So a custom driver is the key, but whether there's a room behind the door depends on the panel's physical characteristics and how much the controller exposes.

Another voice in the comments: e-ink is clearly capable, yet products on shelves only use it for price tags — aren't manufacturers squeezing利润 toothpaste-style? This judgment is only half right.

A hobbyist can do fine calibration on a single panel, tolerate some ghosting, and not care about the screen's condition five years from now. Mass production must face panel-to-panel variation, usage environments from below zero to over thirty degrees, long-term driving effects on lifespan, and ordinary users' near-zero tolerance for ghosting. Manufacturers' conservative waveforms are largely the result of these constraints.

But this kind of hack does reveal a direction: content-adaptive, per-pixel driver allocation combined with automated calibration shows that e-ink's refresh potential is far from fully挖掘. From Wenting's 60 Hz to the 20 FPS hack earlier this year, to this 30 FPS achievement, an open-source community is gradually rewriting people's impression that "e-ink can only display static content."

When Bo's code goes open-source, the most worth watching will be his pulse allocation algorithm and calibration pipeline. If this method can be reproduced across more off-the-shelf development boards, then smoother e-ink small devices may arrive sooner than we expect.

(This article was generated by Claude Opus5.5. Article prompt: Analyze this project and write it as a WeChat public account article https://hackaday.com/2026/10/14/tiny-e-paper-display-delivers-30-fps-of-game-boy-fun/. In the article, address this question: Can any e-ink screen be made into a high-refresh device? Is it enough to have a so-called custom driver?)

apov Reader Group Admin WeChat: weilaifuture65

Conflict of Interest: None

Related Articles

Outrageous: Developing Games for E-Ink

Ink Console E-Ink Text Game Device Ends Crowdfunding

Modos Flow Hits Crowdfunding Goal / Developer Shares E-Ink Monitor Development Journey

"Giving" Yue Xing Tong a "Chick" in Vietnam?

M5 Paper S3 Coming for Remake + Upgrade?

1.54-Inch E-Ink Screen Runs at 30fps: The Driver Magic Behind a GameBoy Emulator

暂无评论,快来发表第一条评论吧!No comments yet. Be the first!

📮 需求咨询📮 Inquiry