The LaST Upgrade Part 47 - DMA SURVIVAL GUIDE
Last updated December 18, 2023
The Atari STE computer, released in 1989, has been the subject of a persistent belief within the Atari community that some early models were manufactured with a faulty DMA (Direct Memory Access) chip, leading to issues with data transfer between the computer's memory and peripherals such as floppy disk and hard drives. However, the only evidence for this "Bad DMA" theory is a news snippet claiming that the first 200 machines out of the factory had a faulty chip, which reportedly never made it into users' hands.
There is no credible evidence to support the existence of a faulty DMA chip in the Atari STFM or later STE models. While some systems may have experienced issues with data transfer or other problems, these were likely due to other factors rather than a faulty DMA chip.
Despite this lack of evidence, the belief in a "Bad DMA" chip has persisted within the Atari community, with some users attributing all hard drive problems to this supposed issue. However, research conducted by Exxos suggests that DMA-related problems with hard drive corruption can be resolved through means other than replacing the "Bad DMA" chip, which is often identified by the number C025913-38 with the "Good DMA" with number C398739-001A. Exxos found that multiple factors can lead to hard drive corruption, and faulty DMA chips are rarely the cause. Changing the DMA chip to the later C398739-001A version does not necessarily fix all issues is commonly assumed.
Atari made a number of revisions to the STE hardware during its production run, and any issues with early models were generally addressed through software updates or hardware revisions. Many users have successfully applied these fixes and used an -38 DMA chip without experiencing hard drive problems.
It is essential to understand that hard drive corruption can stem from various sources, such as a faulty hard drive or memory card, a flawed power supply, bad connections, or cables. Even if the hard drive itself is in excellent condition, an improperly functioning power supply in an Atari can still cause hard drive problems. It is also problematic because simply changing the ROM chips themselves can make or break stability issues along with simply changing the brand of processor, Even on earlier STFM machines.
The presence of "DMA problems" is not limited to the STE series, despite the output buffering chips it employs. Hard drive issues can also occur in the original STFM and STM machines. It is essential to recognize that these problems are not exclusive to the STE series, as commonly assumed. While it is commonly believed that writing to the drive causes corruption, this issue is not restricted to write failures, as the odd read failure is less noticeable but still problematic.
In summary, while subsequent STEs featured the upgraded C398739-001A DMA chip, effectively addressing the majority of issues, it's crucial to acknowledge that early STE machines also necessitate supplementary fixes, as duly documented by Atari & Exxos. Simply swapping out the DMA may not suffice on its own. Furthermore, obtaining the C398739-001A DMA chip proves challenging and often entails a substantial cost, along with a requirement for users to possess soldering expertise for safe removal and replacement, avoiding irreparable damage. Exxos's DMA investigation identified multiple points of system failure, and their objective is to offer straightforward and cost-effective solutions to rectify these issues.
WHAT IS THE "DMA ISSUE" ?
The well-known classic DMA issue manifests when attempting simple operations like copying files within a hard drive partition, creating folders, or saving the desktop. Typically, users can create multiple folders and save the desktop, and after opening and closing the window several times, the files may become corrupted. To test this issue more comprehensively, it is advisable to continuously generate hundreds of files and folders, verify the names for correctness, delete them, and observe the outcomes. Alternatively, users can copy specific test files to an SD card on a PC and then transfer these files from one partition to another, eventually deleting them. In this process, it is common to encounter corruption or unexpected errors during copying or deletion.
Similar faults may arise with the floppy drive, leading to intermittent messages indicating that the drive is write-protected when it is not. While this issue also falls under the umbrella of DMA problems, it generally differs from the previously mentioned hard drive-related DMA issue. It's crucial to recognize that there are multiple failures within the DMA system. Consequently, it is important to systematically identify, address, and fix each issue individually. Even if users currently do not experience these problems, it is still recommended to implement these modifications, as they have been tried and tested by many community members over the past decade, ensuring a proactive approach to maintaining the system's stability.DOWNLOAD TEST FILES These files contain many directories within directories and also contain various sizes of text files filled with random characters.
It's crucial to recognize that while there are "free" HD drivers like the ICD or AHDI drivers, these drivers are essentially three decades old and have inherent issues of their own. I typically recommend utilizing well-known and actively maintained drivers, such as the one found HERE . This particular driver is consistently updated and exhibits compatibility with both old and modern hard drives in general. While it may not be free software, it is undoubtedly a worthwhile investment for those seeking trouble-free hard drive operation with the latest and most reliable driver software.
KNOW YOUR HARDWARE
Given the abundance of Atari-compatible hard drives available for sale, it cannot be assumed that all projects have undergone thorough testing before release. Numerous projects have been cloned and replicated multiple times, making it challenging to ascertain whether modifications, upgrades, or changes have been implemented by competent individuals or rigorously tested. Exxos has certified only the BBaN drive in recent years, conducting extensive tests and personally verifying its reliable functionality without issues.
It is crucial not to assume that a purchased hard drive, regardless of the source, works as intended. Jumping to blame the computer in case of malfunctions may be futile. It is advisable to double-check the drive's reliability not only on a STE machine but also on an STFM-type machine. If the hard drive exhibits malfunctions on both machines, there is a likelihood that the issue lies with the drive itself or the power supply, emphasizing the importance of thorough verification.
Servicing the power supply in your Atari machine is equally crucial. A malfunctioning power supply can lead to stability issues throughout the entire system. It has been demonstrated repeatedly that when the power supply is faulty, the DMA system tends to suffer corruption first. Therefore, regular servicing of your machine is essential, regardless of whether issues are immediately apparent or not. This proactive approach helps ensure the overall health and reliability of the Atari system.
It's important to avoid jumping to the incorrect assumption that the DMA is faulty, as this is rarely the case. Exxos has provided an in-depth article outlining various ways a circuit can function and create the appearance of a defective chip. This highlights the complexity of diagnosing issues in electronic circuits, emphasizing the need for a thorough understanding of potential factors that can mimic the symptoms of a faulty DMA chip. Patience and careful analysis are key when troubleshooting Atari systems to accurately identify and address the root cause of any problems.
It is crucial to understand that Atari issued several motherboard revisions over the years, addressing various issues, incorporating new chips, and continuously updating them as time progressed.
CA4003290 - "BETA BOARDS" 1989
The above image displays one of the initial popular series of motherboards produced in the factory, often referred to as beta boards. Notably, at the center of the board, there are makeshift wires (bodge wires) and a chip soldered on top of another one, serving as a standard factory fix. Some earlier revisions even feature the older SIPP (pinned) style SIMMs or have black SIMM sockets.
CA4003290 - 2nd REVISION
As evident in the image above, the third chip is now soldered directly to the motherboard. Ironically, despite being a later revision, this board still retains the use of older SIPP sockets.
CA4003290 - 3nd REVISION
This revision reveals the elimination of the external blitter chip, integrating it into the large surface mount chip positioned just to the left. It is probable that around this time, the updated DMA C398739-001A chip began to emerge. However, pinpointing the exact transition during production batches is challenging to discern from the provided images, as people have been replacing these chips over time. From my personal experience, very few actually have the updated DMA chip from factory.
CA401177 - FINAL/LAST REVISION 1991
The revision above sees sockets removed and surface mount CPU,Shifter, DMA. Because there are not many of this provision out in the wild it is difficult to ascertain if this board has DMA issues or not. But considering it is one of the latest board revisions, it is assumed it does not have the issues.
There may well be many other variants of the above motherboards but the ones listed are generally the main ones seen.
ATARI'S OWN FIXES
One of the earliest documented Atari remedies involves replacing the resistor pack P100, switching from a 2.2k type to a 1.2k type. This adjustment is also mentioned in the referenced article HERE. Where this increases the pull-up strength of signals, mainly BR,BG,BGACK which are part of the DMA hand-off between the CPU and other DMA devices. The 33pF modification will be mentioned later.
Atari also changed the resistor packs on the address bus P101,P102,P103 from the generic 10K to 4.7K. Likely to address stability issues.
Exxos's tests on the STE at 32MHz revealed that even in 8MHz mode, the 4.7K and 10K pullups were inadequate, leading to repeated crashes after extended runs. To address this, both address and data bus resistors were replaced with a lower value of 2.2K. This not only resolved the 8MHz stability problems but also addressed issues when running the bus at a faster speed of 32MHz. It was also noted that in doing this modification it greatly resolved DMA (hard drive) related issues as well.
Returning to the 33pF modification intended to address DMA issues, Exxos has raised concerns about a potential typo in the article. He observed that the capacitor is placed directly across two data lines, a configuration that is highly unusual. Despite testing variations such as connecting the capacitors to GND, Exxos did not observe any significant improvements. It remains unclear where this capacitor modification originated, but there is a possibility that these capacitors were introduced to mitigate bus noise, similar to the resistor pack changes.
While it is not uncommon for a signal to experience considerable noise leading to issues, Exxos documented particular problems with D4 and A1 on some STFM machines. These issues were generally addressed by changing the pull-up resistors to the lower 2.2K values.
Less recognized yet equally significant are other bus-related issues. Notably, Exxos discovered that merely switching ROM chips from one brand to another could lead to stability problems, potentially resulting in DMA corruption. Similarly, altering the CPU brand could be a determining factor in DMA-related issues, with the potential to not only disrupt the DMA circuit but also cause system malfunctions, manifesting as random ROM errors, TOS crashes, and other unexpected issues. Changing the pull-up resistors to the lower 2.2K values also alleviated such problems, even on earlier STFM machines.
EXXOS DMA FIXES INTRODUCITON
The frustration surrounding "DMA issues" often stems from a narrow focus on the DMA chip itself, which is rarely faulty on its own. An example is swapping a perfectly working C025913-38 from an STFM into a suspected faulty STE, resulting in the STE working fine. Incorrectly assuming the removed chip from the STE was faulty can lead to misguided conclusions. If that supposedly faulty chip is then placed in the STFM, it often functions without issues.
Swapping the DMA chip and seemingly solving the problem can be attributed to chance. It's crucial for individuals to understand that each machine has its own unique characteristics, and addressing specific quirks is necessary. Determining the specific faults on particular motherboards is challenging. Exxos has dedicated considerable time to testing various motherboard revisions, identifying faults that can lead to hard drive or floppy corruption, and devising fixes that not only address the problems but are also verified by the community members. This approach emphasizes the importance of a comprehensive understanding of each machine's intricacies for effective troubleshooting and resolution.
As mentioned earlier, the updated C398739-001A DMA chip, as demonstrated by Exxos in THIS video, does not ensure a guaranteed solution, as success in resolving the issues can be influenced by various factors. While users may instinctively turn to the updated DMA chip, it's important to note that it is often challenging to locate and can be expensive. Recognizing the potential risks associated with users attempting to remove and replace the DMA chip,potentially causing damage to their boards,Exxos sought alternative solutions that don't require soldering. This approach aims to offer safer alternatives that work effectively at a lower cost, aligning with Exxos's objective of providing accessible and secure options for Atari system enthusiasts. Regrettably, due to the array of different faults, soldering is occasionally necessary. Fortunately, in most cases, this doesn't involve the removal of chips, making it a safer procedure to undertake.
An additional, less commonly known or observed issue involves the 1772 floppy controller, which, for reasons not yet fully understood, has been associated with causing hard drive corruption. Substituting the 1772 chip with a different one appears to resolve the problem, although the exact cause remains speculative. One possibility is that certain 1772 chips may exhibit slow deactivation of the bus, potentially leading to conflicts. However, this is currently speculative, and further research is needed to gain a better understanding of this issue. As of now, this problem has been observed in only 2 machines, making it an uncommon occurrence.
DMA BUS ISSUES, FLOPPY & HDD
A primary concern involves problems with floppy drives, as detailed in THIS video by Exxos, where he highlights peculiar faults associated with floppy drives. These issues may manifest as intermittent read or write errors, along with random error messages indicating that the disk is write-protected or that data in the drive is corrupted. It's crucial to acknowledge that, in addition to system-related faults, floppy drives themselves may require servicing. Common issues include dirty heads or perished belts, emphasizing the importance of regular maintenance for optimal performance.
The fix as outlined in the video are represented below.
To address the issues outlined in the video and various related problems, soldering a 10K "common bus" 9-pin resistor array onto the DMA/1772 DATABUS proves effective. The data bus, indicated by red arrows on the right, can have the resistor pack soldered onto either the DMA or 1772. However, it is generally more convenient and consistent to solder the resistor pack onto the 1772 itself. One end of the resistor pack can be connected to the 5V side of the capacitor on the motherboard, and 5V can be sourced from multiple locations, wherever it is most convenient.
It's worth noting that although Atari introduced buffer chips between the DMA and DMA socket, Exxos observed no noticeable impact with or without these buffers. The original ST models, which did not have these buffers, functioned well with the resistor mod added. Exxos posits that the buffer chips, while serving an apparent purpose, may be buffering problematic signals rather than effectively resolving the issues Atari intended to address by adding the buffers.
1772 PULLUP FIX
Similar to the problems above the floppy drives, for some reason some drives suffer from other issues as described in THIS article.
The signals listed below by the red dots are generally those with pull-up resistors on the floppy drive itself. However, for reasons not entirely clear, these resistors can be insufficient in certain cases. Some drives may experience malfunctions even with a 10K pull-up built onto the floppy drive, while others operate perfectly with a 500K pull-up. Additionally, some drives function well with a 4.7K pull-up, while others do not.
In light of these variations, Exxos recommends adding pull-ups directly on the 7406 to address intermittent floppy drive problems. This approach aims to provide a more consistent and reliable solution by directly managing the pull-up configuration for the affected signals.
CPU TIMING PROBLEMS
In Exxos's comprehensive exploration of DMA issues, a notable discovery emerged concerning the effectiveness of transitioning to an HC variant CPU for addressing DMA-related challenges. Exxos underscored that motherboards equipped with Motorola CPUs consistently faced DMA issues, in contrast to those featuring the SGS type, which generally encountered fewer problems. Furthermore, it was disclosed that both the SGS and, more noticeably, the HC type processor exhibited significantly less noise compared to their Motorola counterpart. The marginal timing difference of just 2ns with the SGS type often proved adequate in resolving DMA issues.
The nuanced variations in timing between these two CPU types played a pivotal role in influencing the incidence of DMA issues. The introduction of the HC variant, despite its operation at the same 8MHz, demonstrated slightly faster reaction times sometimes up to 22ns in processing bus grant signals to and from the GST. This improvement facilitated more effective management of DMA signals, preventing synchronization issues that could potentially lead to problems. Essentially, the enhanced reaction times contributed to mitigating timing challenges associated with the DMA chip, resulting in an overall improvement in system stability.
Exxos has also noted that incorporating a CPU accelerator, even at 8MHz speeds, can occasionally disrupt a functioning system. This issue arises from logic delays that impact DMA timings. To address this challenge, Exxos is currently developing an addon board for the STE booster series to mitigate the problem.
Tested and verified CPUs are available in the store for convenient replacement.
POWER SUPPLY PROBLEMS
If your ST/E machine is experiencing sporadic crashes, subpar video quality, random issues with hard drive or floppy drive access, or audio disturbances, there's a good chance that your power supply unit (PSU) is deteriorating. Given the age of these PSUs, it's not surprising that they require an update. Some PSUs may exhibit seemingly adequate regulation, but when accessing the floppy drive, the screen brightness dims significantly, indicating a lack of "on-demand" power in many cases. Attempts to enhance peak power by adding additional capacitance on the 5V rail may not yield satisfactory results, as evidenced by minimal impact on screen dimming. STE machines may malfunction when installing perfectly known good extra memory.
Certain widely used power supplies, such as the Mitsumi SR 98, tend to suffer the most from failing capacitors. In contrast, others like the Atari DVE type appear more resilient over time. It's crucial to replace capacitors with high-quality, low equivalent series resistance (ESR) types. Mere capacitor replacement with the cheapest brands may prove ineffective, as some inexpensive capacitors, even from well-known brands, can be of lower quality than those originally installed in a 30-year-old power supply. Blindly replacing capacitors without ensuring they are rated for switch mode power supply usage can be counterproductive.
It's a well-established fact that Atari machines are prone to various noise-related issues, and a failing power supply exacerbates these problems. A stable 5 VDC output with a typical noise variation of 0.2 volts is essential. If the power supply exhibits fluctuations between 4V and 5 V, as depicted in the image above, Exxos discovered that the DMA circuit is the first to malfunction when the voltage is consistently too low. Failing power supplies can also lead to random crashes during floppy drive access, as the motor demands a relatively high current, causing the power supply voltage to drop intermittently and the machine to crash during drive access.
For more information about failing power supplies see Exxos's PSU article HERE.
It's crucial to recognize that while a CPU may possess the technical capacity to drive a bus despite its relatively modest output strength, this capability is most effective on smaller boards where the CPU is closely situated to connected devices like ROM or RAM. However, as the board size increases, the inherent inductance and capacitance variations in PCB traces escalate, diminishing the efficacy of this approach. On larger boards, these challenges make it increasingly difficult for the CPU to drive signals reliably, leading to potential issues such as signal oscillation, degraded signal quality, and the risk of latch-up. The intricate interplay of inductance and capacitance on larger boards necessitates meticulous consideration during the design phase. Implementing strategies like proper impedance matching, termination techniques, and a thoughtful PCB layout becomes essential to mitigate signal integrity concerns. In more complex systems, the use of simulation tools and analyses becomes imperative to proactively identify and address potential signal integrity challenges before manufacturing.
Expanding on this, the CPU's notable pulldown current capability allows it to effectively pull the signal to a low state, ensuring a robust grounding. However, the weaker pullup current becomes a limiting factor when attempting to raise the signal to a high state, especially in the absence of pull-up resistors. On smaller boards with shorter distances and minimized inherent capacitance and inductance, the CPU's inherent characteristics might be sufficient to drive signals effectively. However, as the board size increases, the interplay of these factors becomes more prominent, necessitating careful consideration of the weaker pullup current in the context of larger boards.
Practically, this information underscores the importance of a meticulous circuit design, taking into account both the board size and the CPU's pull characteristics. Addressing signal integrity issues through thoughtful circuit design becomes imperative, ensuring reliable transitions between high and low states, particularly in light of the challenges posed by weaker pullup currents on larger boards.
In further detail, pull-up resistors play a dual role:
Preventing Bus Floating: Pull-up resistors are crucial in preventing
the bus from floating, maintaining a defined voltage level when the bus
is inactive. This prevents undefined states and ensures signal stability.
Aiding Logic High Drive: Importantly, pull-up resistors significantly
contribute to the logic high drive capabilities of the bus. In scenarios
where the CPU's pullup current is relatively weak, the pull-up resistor
assists by providing additional current to effectively raise the signal
to a logic high state.
On larger PCBs, where challenges in inductance, capacitance, and signal propagation are more pronounced, the inclusion of pull-up resistors becomes even more critical. They not only stabilize the bus by preventing floating but also ensure that signals can be reliably driven to logic high levels, addressing the limitations imposed by the weaker pullup current of the CPU.
Occasionally, the occurrence of a gradual change in signal voltage, known as a slow slew rate, can give rise to problems, necessitating the implementation of a stronger pull-ups. The impact of this phenomenon varies depending on the specific logic family in use. Moreover, the non-monotonic nature of the signal suggests the possibility of reflections, which can be particularly problematic, especially if they occur within the range of logic thresholds. This issue becomes more pronounced with modern, faster logic, as such systems tend to be more sensitive to these irregularities.
In summary, dispelling the misunderstanding and recognizing the dual role of pull-up resistors both in preventing floating and aiding in logic high drive capabilities is essential, particularly in the design of circuits for larger PCBs with extended signal paths.
It's crucial to recognize that while changing the CPU to an HC type often resolves issues, it's essential to understand that, at a minimum, Atari fixes must be implemented to bring the machine up to Atari specification. Simply replacing the processor in one of the early versions, such as beta boards, which already experience weaknesses like the 10K bus pullups, may still lead to problems. In fact, it could potentially introduce additional issues until all the necessary fixes have been applied to address the specific quirks and shortcomings of the particular board.
Exxos has additionally observed that the stability of the board can be influenced by the presence of metal shielding. Notably, issues with audio may emerge if the metal shielding is not securely fastened to the outer edges of the motherboard. It appears that the shielding serves as a short path to the power supply ground rather than relying on the actual PCB traces. Removing the shielding entirely could potentially lead to malfunctions in the machine, while a complete and properly intact shielding may contribute to more reliable behavior.
In summary, the information discussed above addresses issues and provides resolutions that have been verified by Exxos and the community over the past decade. It's important to note that these fixes don't claim to comprehensively solve every problem with the machine. Indeed, there are always additional faults awaiting diagnosis. Therefore, individual users might encounter issues beyond the scope of this article, and these problems may not currently have known solutions.Overall they are almost endless ways a circuit can malfunction even with perfectly operating chips. This is described in the article HERE.
While the article aims to address common issues, it's essential to recognize that there are numerous ways a circuit can fail. It cannot be assumed that implementing the modifications outlined in this article will universally fix every specific user's problem. Some machines may have additional issues, such as broken components, faulty power supplies, board corrosion, bad sockets, or even broken PCB traces. The article specifically outlines common DMA related issues and the corresponding fixes that have undergone testing and verification by the community.