The Dangers of 3.3V Flash in Retro Consoles

EDIT:

Measurements and proof here

EDIT 2:

After witnessing several complete misinterpretations of my article I decided there was 1 word in the article that needed to change:

  • Avoid -> Use sparingly
    • This was my original intent, avoid as in “avoid getting drunk too often”.

Let’s be clear that this article is NOT fear mongering, nor should the article be twisted in that way. In the article’s introduction I clearly state that I am an Everdrive user myself. There is no hidden agenda, my only intent was to inform people of a design flaw. Also, the explosion picture above is meant to represent how you should dispose of your NES and Neogeo Multicarts (not Flashcarts), after all, these are the only which I recommend to not use and to “burn”.

TL;DR

Stop using your Aliexpress multicarts, keep using your Everdrives – but know the design flaw.

Original article follows

The use of 3.3V flash in retrogaming can lead to serious problems. Following a lengthy podcast discussion on the subject I felt it was important to elaborate on the matter by:

  • Giving technical explanations of the problem
  • Addressing common rebuttals
  • Listing devices which exhibit this problem

Let me be clear that using 3.3V Flash in a 5V system in of itself is not inherently bad; it can be done properly using level translators such as what has been done with the SD2SNES. What is bad, and what I am speaking out against, is directly connecting 3.3V Flash to 5V buses without proper interfaces such as level translators.

Let me also be clear that not all multicarts, Everdrives, and what not, are poorly designed. I own a few Everdrives myself and I enjoy them. Actually I take back the comment on multicarts, every single one I examined for this write-up was horrendous.

Technical Explanation

“But db!!! This flash is 3.3V and it’s LESS than my console’s 5V so there’s no harm – I can math!”

No, when the console outputs 5V into a 3.3V input the extra voltage must go somewhere; 1st law of thermodynamics. It is converted to heat through the unintended conduction of clamping diodes, which can be harmful to integrated circuits. These diodes are there for protection against electrostatic discharge (ESD), which are very short and infrequent bursts of energy. They are not designed for continuous conduction and, therefore, continuous heat dissipation. Let’s take a look into why this occurs.

Driving a 5V signal into a 3.3V input

To understand what happens when a 5V signal is applied to a 3.3V input, we must first understand what a 3.3V input looks like. Typically, a digital input has 2 clamping diodes (D1 and D2 below) on its input to protect against small electrostatic discharges (ESD). When a logic high of 5V is applied to the circuit below, D1 starts conducting and essentially short circuits the additional voltage to the 3.3V supply. In certain flashcarts, including several Everdrive designs there is a small resistor (R1 – 100 ohms) to limit the short circuit current to approximately 12.5 mA, under ideal conditions, between the 5V supply and 3.3V supply – this protection is inadequate since typical CMOS current ratings are ± 5.2mA.

On other common devices, such as multicarts, there is no R1 and a logic high of 5V driven directly into the 3.3V flash results in a short circuit between the 5V and 3.3V supplies. In either case, this causes unnecessary and potentially damaging stress on both ends:

  • On the console output since it is not designed to supply nearly 12.5mA, under ideal conditions, (or more on multicarts) per pin
  • On the 3.3V Flash input since the clamping diodes D1 and D2 are not designed to dissipate large amounts of heat
Voltages in yellow – Currents in blue – (e-3 = m = milli)

Everdrive Ratings

Most ealier Everdrives use a 3.3V Flash from the M29W family. The datasheet of these devices clearly indicates the maximum input voltage (Vih, see below) to be Vcc + 0.3V. In a 3.3V powered chip, this maximum voltage is 3.6V – yet, when using most Everdrives were are subjecting this Flash chip to approximately 3.75V on its digital inputs and significantly exceeding the current limit of the pin.

 

3.75V is in fact lower than the Absolute Maximum Rating Voltage Input of Vcc + 0.6V. But, designs should never exceed the DC Characteristics of a component. Furthermore, as can be read from the paragraph in the M29W datasheet below, “These are stress ratings only and operation of the device at these or any other conditions above those indicated in the Operatings sections of this specification is not implied”. Or, to summarise, don’t count on the part working properly above the specifications of Table 11.

Now, to Krikzz’s credit, I will at least say that his circuits are not nearly as bad as multicarts due to the addition of R1 (100 ohm) to limit the short circuit current. But, it doesn’t make the circuit any less out of specification. Moreover, I highly suspect the addition of R1 comes from a misinterpretation of Altera app note AN258 which states that Altera Cyclone devices can be interfaced to 5V logic provided that PCI clamping diodes are activated and that the current limiting resistor is properly sized to respect to current output capabilities of the 5V device. However, a careful reading of Cyclone datasheets will reveal that PCI clamping diodes are not available during configuration and thus the FPGA is not protected during the first few seconds of power-up in Everdrives such as the Mega Everdrive. It is also important to note that Altera devices have PCI clamping diodes while most, if not all, other 3.3V Flash devices have no such fancy clamping diodes and therefore this app note most certainly does not apply to anything other than an Altera Cyclone FPGA.

All in all though, I do get the sense, from the revisioning of Everdrives, that Krikzz is on the right track in improving his designs with proper level translation.

What Damage Can Be Caused?

Prolonged use of components outside of their specified tolerances inevitably leads to failure. On the console side, the stress is excessive current output on digital outputs when driving a logic high. On the cartridge side, the stress is excessive heat dissipation due to conduction of the clamping diodes. I have already heard from several friends that their NES consoles have died most likely due to their admittedly heavy use of cheap multicarts. These are particularly bad. I would avoid these like the plague. I suspect poorly designed Everdrives will require more time before we start seeing failures.

Common Rebuttals

I thought it would be important to list common rebuttals to my 3.3V Flash criticisms and give explanations and or analogies as to why they are misguided.

“Well, my device has been working properly for 2 years now.”

This is like saying: “I’ve been smoking for 2 years and I don’t have cancer. Therefore, smoking does not cause cancer.” If you believe this statement, you deserve to damage your retro console.

“5V Flash doesn’t come in large enough densities for modern applications.”

While this is true, all it really means is the designer needs to thoroughly address the need to properly translate the voltages between the 3.3V side and the 5V side. This does lead to an increased component count and increased cost, but, this is the proper way to do it. The SD2SNES cart is a prime example of how to solve this problem correctly.

“Only older Everdrives have this problem.”

Not true, see the final section below for a detailed examination of the Everdrive family.

“You’re just jealous!”

This is hilarious! I don’t even make Everdrives!

“You’re a straight white male!”

Yes, and I’m also French which means I’m an asshole as well!

Devices

This is obviously not an exhaustive device list, nor is it an exhaustive Everdrive list as well. I did list a few “good” Everdrives below. If the Everdrive your are worrying about is not listed (N64, GBA) it is because those systems run on 3.3V and therefore there is no inherent voltage translation problem.

Flashkit MD

Verdict = Avoid

M29W type Flash connected to the Genesis 5V bus using 100 ohm series resistors. This is not a good interface. Inspect your homebrew carts since this board is currently being sold to unsuspecting homebrewers along with a burner for physical releases.

Everdrive MD

Verdict = Use sparingly

The databus is connected through several 100 ohm resistor networks (no label, but in the bottom right corner) instead of using a third level converter. Also, the lack of component designators on earlier Everdrive really irks me.

Mega Everdrive

Verdict =Use sparingly

M29W type Flash (back side) and Altera EP2C5 connected to the Genesis 5V bus using 100 ohm series resistors. This is not a good interface. Moreover, this particular designs shows the misinterpretation of Altera app note 258. Another really bad design choice on this board is the fact that two FPGA outputs directly drive the cartridge audio inputs without any sort of amplifier. This is particularly bad since most Model 1 Genesis’s have 75 ohm inputs on the audio lines. This input impedance is much too low for the FPGA to drive directly and the pin current limits are surely exceeded.

 

Mega Everdrive x7

Verdict =Use sparingly

This sligthy newer version shows improvement, there are two 16 bit level converters below the EP2C5 FGPA, but I still say avoid this one because from the looks of it the databus is connected through several resistor networks (labelled RA3 to RA7) instead of using a third level converter. It is not quite clear to me after searching through the internet if this Mega Everdrive v3 is the same PCB as the Mega Everdrive x7. There does appear to be more circuitry near the cartridge audio inputs (bottom left) but it is unclear whether or not U2 is used as an amplifier for the Model 1’s low input impedance.

Mega Everdrive x5

Verdict = Good

Finally, a Genesis Everdrive with proper level translation. Good job! Actually, there is no flash on here, ROMs are loaded into a PSRAM which explains why these models of Everdrive are so much faster than others. But wait, there’s no ground plane? I just realised now that some Everdrives I’ve looked at has no ground pour…

Master Everdrive

Verdict =Use sparingly

M29W type Flash connected to the Master System 5V bus using 100 ohm series resistors. This is not a good interface. No ground pour, no reference designators…

Gamegear Everdrive

Verdict =Use sparingly

M29W type Flash connected to the Gamegear 5V bus using 100 ohm series resistors. This is not a good interface. No ground pour, no reference designators…

Turbo Everdrive v1

Verdict =Use sparingly

M29W type Flash connected to the Turbografx-16 5V bus using 100 ohm series resistors. This is not a good interface. This PCB has ground pour in a few spots.

Turbo Everdrive v2

Verdict = Good

Hey look at that, new version and now we’ve got proper level translators on there. Good job! But hold on, we’ve lost the ground pour which was present on the version 1…

Everdrive N8

Verdict = Good

Added by popular demand. The Everdrive N8, by all looks and appearances, seems to be properly interfaced to the NES 5V bus. There are sufficient level converters to handle all signals. I do see a 100 resistors network but I will give the benefit of the doubt and say that those are used for the audio input to the console. Good job!

Super Everdrive

Verdict =Use sparingly

The databus is connected through several 100 ohm resistor networks (no label, but in the bottom right corner) instead of using a third level converter. Nice ground pour on this one.

SD2SNES

Verdict = Pure Gold!

The SD2SNES is not part of the Everdrive family and is a prime example of how to properly design a flash cart. Excellent work! Every signal is properly translated between the two different voltages. I should also note that the SD2SNES actually uses a ground plane as opposed to most Everdrive designs.

NES 150 in 1 (Famicom version)

Verdict = never use again!

Now this design is complete garbage. Whoever made this design should be ashamed of themselves. The Flash used here is a MXIC 29GL256E which are clearly rated for 3.6V maximum. The most frightening thing about this board however are the two diodes above the MXIC Flash chip. I REALLY hope they don’t use those to lower the 5V rail down to power the Flash. This is the type of PCB which has reportedly caused some NES consoles to die within a few months of use. Presumably the NES version is the same shitty design. Stay away!

NES 400 in 1 (Famicom version)

Verdict = burn it now!

Same story as above but with some striking differences. I don’t see any voltage regulators or coupling capacitors on this board! WTF! Are they powering this 3.3V Flash directly off of 5V! Burn this immediately if you own one.

Neo Geo 161 in 1

Verdict = Avoid

Most Neo Geo multicarts are plagued by voltage problems – looks at all those 3V Flash chips and not a single level translator in sight! There are even Youtube videos of people attempting to fix the voltage incompatibilities. This particular one pictured here has a convenient footprint for a voltage regulator yet is seems to use two series diodes instead. Presumably the designers noticed the cartridge wasn’t working so well with 3.3V Flash (MSP55LV100S) and decided to up the voltage a bit. What a shitty design. In fact, the little information I did manage to find about the MSP55LV100S suggests that it’s a 3V part. Neo Geo hardware is expensive, don’t subject it to this kind of torture!

FYI – never put bare PCBs on the carpet.

 

Looks like some models have the regulator installed. Still a shitty design.

 

René

Electronics engineer and retrogaming fanatic!

120 thoughts on “The Dangers of 3.3V Flash in Retro Consoles

  • July 5, 2017 at 15:21
    Permalink

    Salut René et merci pour ton article très intéressant.
    Est-il possible de corriger le problème de voltage que tu soulignes sur les Everdrive ?
    Merci and keep up the good work.

    Reply
    • July 5, 2017 at 15:36
      Permalink

      Pas vraiment non, ça serait très difficile d’ajouter des level translators là où il en manque.

      Reply
      • July 5, 2017 at 15:51
        Permalink

        Merci pour ta réponse.
        En fait, je revenais te dire que tu pouvais ne pas publier ma question car je venais d’avoir la réponse dans le Live Stream #5, mais tu as été trop rapide 🙂

        Reply
    • July 6, 2017 at 13:06
      Permalink

      The Neo SD has level translators coming out of its ears! I am sure the Neo SD is fine – much much better than a 161 in 1 etc.

      Reply
  • July 5, 2017 at 16:50
    Permalink

    Du coup, n’est-ce pas là une opportunité de lancer sa propre série de cartes flash ? ^^

    Reply
  • July 5, 2017 at 16:56
    Permalink

    Can you add the new Neo Geo Flash Cart on here and give us your thoughts on that?

    Reply
  • July 5, 2017 at 16:57
    Permalink

    Hrmm, ever looked into a GDEMU or USBGDROM for the dreamcast? Just curious.

    Reply
    • July 5, 2017 at 22:21
      Permalink

      The Dreamcast is a 3V3 system, and the Holly chip also apparently has 5V-tolerant pins for the G1 / IDE bus, so there shouldn’t be an issue with the current crop of GD Emu boards AFAIK.

      (if we’re talking about the boards from Mnemo or Deunan.)

      I would have a search for the article on the StoneAgeGamer site about one of those boards before buying though.

      Reply
  • July 5, 2017 at 17:50
    Permalink

    As this article has been pointed out to me and questions asked, I’d just like to add my LynxSD cart is SRAM based and runs natively at 5v. The only level translation required is talking to the SD card, handled by a resistor divider.

    My NeoGeo Pocket cart is natively 3.3v and uses 3.3v flash, no translation required, all good.

    Given the popularity of the Everdrives I find the lack of attention to detail a bit scary given doing it properly will add bugger all to the overall build cost of the cart.

    Reply
    • July 5, 2017 at 18:33
      Permalink

      Exellent! Thanks for taking the time to come here and mention this.

      Reply
    • July 6, 2017 at 13:05
      Permalink

      Fancy seeing you here lol! I didn’t doubt for a minute you would have protected the good old Lynx 😉

      Reply
      • July 6, 2017 at 15:53
        Permalink

        @GadgetUK164

        Not you again! :p

        How’s it going?

        I think we really all need to meet up some day, and have a good natter about retro stuffs over a few beers. hehe

        We couldn’t exactly get more like-minded tbh.

        @SainT – how is the Jag flash cart coming along?

        I started my own design last year, but have shelved it for a while due to other projects (Mini N64 mobo, which is now booting).

        I’m wondering if you would be interested in working together on the Jag thing, so we can get a cart released sooner?

        For example – I see the SkunkBoard only used the lower 16 bits of the 32-bit bus, so I’m assuming that would work fine for 99% of games, as it’s only really the 68K that accesses the carts for those?

        I might have to pluck up the courage to hop on the Retro Roundtable soon, if that’s OK with RetroRGB Bob, and everyone else?

        I think we could achieve some awesome projects if we do a bit of teaming up. 😉

        Reply
  • Pingback: Are FLASH carts safe?! – RetroHQ

  • July 5, 2017 at 18:28
    Permalink

    Thanks for sharing such informative albeit concerning information. You mentioned not being able to confirm the PCB for the Mega Everdrive X7. I opened mine and took some pictures of both the front and the back which you can see here: http://imgur.com/a/cD2wB

    So you should be able to feel confirmed that the item you have pictured is indeed the X7 PCB.

    Reply
  • July 5, 2017 at 18:35
    Permalink

    What about Mega Everdrive x3?

    Reply
  • July 5, 2017 at 20:09
    Permalink

    Report from the field:

    I just got Super Russian Roulette, which is a custom cart with custom CPLD mapper (XC9572XL) and two small memory chips (SST39SF020A, 2MBit). Everything appears to be 5V logic to start with, so no problems there.

    Reply
    • July 5, 2017 at 20:09
      Permalink

      Also, it has a ground plane. Solidly-built cart.

      Reply
  • July 5, 2017 at 22:29
    Permalink

    Thank you for this article ! I hope it will make to get flashcarts working safer in long-term use.

    Can you please give me your opinion about this memory cartridge for Saturn ?
    Picture (front) : http://ppcenter.webou.net/satcart/files/20161114_rev33b_front.jpg
    Picture (rear) : http://ppcenter.webou.net/satcart/files/20161114_rev33b_back.jpg
    PCB sources : http://ppcenter.webou.net/satcart/files/20160714_usbcart-kicad_r33_b.7z

    Flash chips are directly connected to Saturn data and address buses, but those are SST39SF040, whose are working at 5V logic, so I guess it’s safe ?
    On the other hand, MAX3000A CPLD is directly connected to Saturn data and address buses, so I’m a bit concerned if that will work fine on the long term or not.

    I would like to feedback your comments to the developer of this cartridge, so hopefully this would help to get better homebrew cartridges for Saturn.

    Reply
    • July 5, 2017 at 22:33
      Permalink

      MAX3000A family of CPLDs is specifically spec’d for operation in 5V systems, so that sounds safe.

      Reply
      • July 6, 2017 at 03:04
        Permalink

        Thank you for the prompt reply !
        In fact, I’m the designer of this cartridge (sorry for hiding this information on my first message, but I wanted unbiased comment from you), so I’m relieved it passed your check.

        MAX3000A is a very convenient IC for retro things, and I plan to continue using it in future projects 🙂

        Reply
        • July 6, 2017 at 07:31
          Permalink

          Isn’t the MAX3000A family going end of life next year? Dr abrasive mentiones this on Twitter.

          Reply
          • July 6, 2017 at 07:35
            Permalink

            Yes it is, I’m very sad about that. I use that part on several designs.

  • July 5, 2017 at 23:33
    Permalink

    Is there a way that we (and by we, I mean a modder for us with no soldering skills) to modify these carts to solve the issue? I have several Everdrives and I don’t want to feel like I’ve wasted my money.

    Also, any experience with the Everdrive GBA, Everdrive 64, or the GDEmu in this situation? I don’t know if these consoles’ flash carts would have the same issue.

    Reply
    • July 6, 2017 at 10:40
      Permalink

      GBA and Nintendo 64 both have 3.3V busses, so connecting 3.3V parts to them is fully within spec.

      In the Dreamcast the drive interface uses 5V in revision VA0 and 3.3V in VA1 and later – this is why GDEmu does not support VA0 consoles.

      Reply
      • July 6, 2017 at 15:30
        Permalink

        Hey, Unseen. 😉

        I forgot that the GDEmu doesn’t work on the VA0, but I’m not sure exactly why this is, because the HOLLY chip itself apparently has 5V-tolerant IO pins for the G1 bus anyway, but is still compatible with 3V3 levels?

        I seem to recall that the VA0 might have another 5V device on the G1 bus though, so I’ll have to look into that. I think it was the BIOS or NVRAM Flash maybe?

        In theory though, the GDEmu itself should work fine on all DCs if it weren’t for those extra 5V devices on that bus? (since all HOLLY chips are the same, and I’m pretty sure all revisions have series resistors on the G1 bus as well)?

        Reply
        • July 6, 2017 at 15:45
          Permalink

          According to the SEGA Dreamcast block diagram, the G1 Bus has the ROM (BIOS), FLASH (NVRAM), and GD-ROM connected to it. The HOLLY chip core runs at 2.5/3.3V, its I/Os are 3.3V. The G1 Bus is 5V tolerant, whereas the G2 and Maple Busses are 3.3V tolerant only.

          Reply
          • July 6, 2017 at 23:31
            Permalink

            Yep, that’s the one.

            It’s also stated in the System Architecture manual about the 5V tolerance on that bus.

            But, I can’t recall if some revisions have a 5V Flash or Mask ROM on the G1 bus. I think the VA0 might have a 5V ROM actually?

            (which would cause problems with the GD Emu boards, since the current ones are only 3V3-based, to the best of my knowledge.)

            I’ll have to check the schematics again later.

  • July 6, 2017 at 00:01
    Permalink

    This is a bloody fantastic write-up, and articulates the concept extremely well. Thank you!

    Reply
  • July 6, 2017 at 00:24
    Permalink

    Is there a reasonable way to modify the problematic Everdrive carts to be safely used with their respective consoles?

    Reply
  • July 6, 2017 at 01:15
    Permalink

    Product suggestion: 3.3v step-down adapter. I think they are called Buck Converters?

    Reply
    • July 6, 2017 at 15:26
      Permalink

      The issue isn’t to do with the main voltage regulation, as the 3V3 stuff on the cart will already have a 3V3 reg for that.

      The thing is that with the designs that use only series resistors on a 5V console with devices on the cart that are not strictly “5V tolerant”, it could in theory be running the chips close to their absolute maximum ratings regarding the current through each input pin (and the current drawn from the bus on the console side).

      Admittedly, it’s probably quite a low risk, as it’s proven to work OK in tens of thousands of carts for a long time (as Krikzz said below), but it’s still not really an “ideal” to not be using proper voltage-translation buffer chips.

      Even dev boards like the popular Terasic DE1 only have 47-ohm resistors on each of the GPIO pins (which go to two 40-pin headers), and I’ve never had a problem with damage to the FPGA when inputting 5-Volt signal levels.

      BUT – the big difference there is that the DE1 has extra clamp diodes on every single GPIO pin (between each resistor and FPGA IO pin), and that means the voltage can be correctly clamped, and the external diodes will be rated for the extra current flow (rather than relying on the ESD clamps diodes that are internal to the FPGA.)

      As Rene also pointed out, even though most Cyclone FPGAs (and I think Xilinx and others too) have the option to enabled “PCI Clamp diodes” on specific IO pins, the problem is that those diodes won’t be enabled during FPGA configuration (which can take say 40-200ms from SPI Flash, depending on the exact FPGA, and the size of the bitstream).

      So, enabling that option on the FPGA IO pins still isn’t a reliable way to properly protect them.

      I’ve had multiple FPGA pins fail recently due to voltage spikes creating a short on the internal ESD clamp diodes, and that’s even while using 100-Ohm (or higher) resistors on each important IO pin.

      That was mainly while plugging in the JTAG cable from the PC, so obviously a lot less likely to affect most Flash carts, as you don’t normally need to plug any external cables into them during normal use.

      I have had FPGA pins get zapped for no good reason as well though, and that was on the Cyclone II / III.

      I can’t vouch for Xilinx nor other manufacturers, but I do think the Cyclone chips seem to be fairly sensitive to the issue (unless proper buffering / voltage-translation is used).

      Reply
    • July 6, 2017 at 09:08
      Permalink

      Thanks for your feedback. I will however disagree with your first point that because you don’t see any failures it makes using the components outside of tolerance acceptable. Even if your testing shows it “can” work under those circumstances the manufacturer has clearly stated that these absolute maximums must be respected.

      Reply
      • July 6, 2017 at 12:17
        Permalink

        Well, stable work of thousands carts during many years isn’t proof that all ok? I think it is good enough proof. Otherwise we would see lot of noise in internet around this problem. On other side, there is no any proofs of opposite point of view. Side effects which i really seen is extra emi and power consumption, some noise on the bus while signal edge changes, but not critical for such slow system like retro consoles and can be filtered with fpga.
        I not trying to prove that design without voltage translators is a good design, but at least carts do not damage itself or consoles.

        Reply
        • July 6, 2017 at 13:36
          Permalink

          Yes, I do agree – if it was huge problem I am sure we would see lots of failed systems and carts from Everdrives, but we haven’t seen that. I think its important for articles like this to point out why things should be done a better way, but the chance of failure seems to be a lot less than people might expect.

          The main benefit of articles like this is to influence better designs I guess. I like to compare this subject to overclocking – it’s very similar in that we know it can in theory kill the CPU, and its not good practice, it might shorten the life on a CPU (at least that’s what the physics and theory tells us), and yet here we are 20 years after people started overclocking and I think we’ve more or less proven that whilst it technically isn’t a good idea, it just works despite that.

          Reply
  • July 6, 2017 at 08:38
    Permalink

    Wow, thanks for posting such a great writeup! I knew some of these carts were sketchy, but didn’t realize just how bad they really can be.
    Are buyers so cost-conscious that a few dollars for level translators matters? Or did the designers just not take the time to add them?

    Reply
    • July 6, 2017 at 09:06
      Permalink

      Level translators are not expensive, there is really no reason to not use them.

      Reply
  • July 6, 2017 at 10:10
    Permalink

    Thanks for the the info on flash carts! There is something I didn’t quite understand fully and would hope to get a bit of elaboration. Sorry if this is stupid question but I’d like to understand and to assess if I have already caused some damage to my Mega Drive.

    In the “What damage can be caused” paragraph you write “On the console side, the stress is excessive current output on digital outputs when driving a logic high.” If I’m not entirely wrong, the only digital outputs there are on the MD console are the controllers. Does this mean if I’m not experiencing problems on their part I should still be fine? Again, sorry for kind of stupid question.

    Cheers for the education and bringing the info forth!

    Reply
    • July 6, 2017 at 10:29
      Permalink

      I’m talking about the cartridge signals here, there are 16 data lines, 24 address lines and many control signals – all digital I/O.

      Reply
      • July 6, 2017 at 10:45
        Permalink

        Ah, I see. I suppose I have yet caused any major damage to the console itself since all the games run as well as I remember them running before starting to use the flash cart.

        Thanks for the clarification on the digital I/O stream!

        Reply
  • July 6, 2017 at 10:59
    Permalink

    Very naiss writeup ! You should mention the possible failures in the consoles (fried pins and whatnot).

    Reply
  • July 6, 2017 at 11:39
    Permalink

    Don’t you know why after more than 5 years of active usage the consoles don’t break down?
    Or should the consequences be visible after 10 or 15 years of usage?

    Reply
  • July 6, 2017 at 13:01
    Permalink

    Great article =D Really pleased someone got around to doing such an in depth review of this and in particular I am pleased you went through so many different carts there. I’ve shared on Google+

    Reply
  • July 6, 2017 at 13:35
    Permalink

    In the end, current is not as high as 12.5 mA (with the 100 ohm resistors) because as the current increase, Vo of the console will tend to Vol (and get farther from the 5V). But I agree that this stresses the console and should be avoided. And in fact my flashcart designs always use transceivers 😉

    I have also seen some extremely bizarre chinese designs that use simple diodes connected to the flash chip power lines. No resistors, no 3.3V regulator and of course no transceivers. Pure crap, but I tested one of these on my Megadrive and was very surprised because there was no apparent heating on the flash chip!

    Reply
    • July 6, 2017 at 14:29
      Permalink

      You’re right but I didn’t want to make the analysis too complicated for non-engineers; 12.5 mA is the easy number to calculate.

      Reply
    • July 6, 2017 at 15:42
      Permalink

      A very common thing on some older “backup” units was to hook up one or two diodes in series with the 5V supply, to drop the VCC to the TTL buffer chips to around 3.8-4.4 Volts. lol

      That would actually work “OK”, but it was definitely out-of-spec for the voltage on the 3V3 device side.

      I think even some newer voltage-translation chips recommend that method, but some of those work OK like that, as they are based on internal FETs to properly buffer the signals, and apparently shouldn’t violate the 3V3 levels on the output (and so the current flow should be within spec too).

      The best method is just to use some 16244 / 16245 buffers that are powered via 3V3, or even better is to use buffers which have dual supplies.

      The issue with using a simple buffer powered by 3V3 that will interface to a “5V” console is that the voltage thresholds for either side might also be less-than-ideal.

      A dual-supply, or more modern voltage-translation chip is the best way to deal with this tbh.

      Personally, I think that if a design used a 3V3 FPGA / CPLD with series resistors, that would be fine as long as it also had external clamp diodes (like the DE1 and other devs boards use for GPIO).

      The external clamp diodes can then be rated for the extra current flow, and also be chosen to clamp the max voltage to proper 3V3 levels.

      Then again, the proper buffer chips don’t really cost much at all, and are often easier to manually solder to the board, rather than having multiple resistors (or resistor arrays) to solder.

      (Obviously if the boards are being assembled by a factory, then using buffer chips isn’t much of an advantage, but it’s still good practice to use them regardless.)

      Reply
  • July 6, 2017 at 19:59
    Permalink

    Wrt some of the Everdrives I think the effects are diminished by observing for example that V out high for a 68K (data/address/control signals) is Vcc-0.75 so it is not 5V but really more like 4.25 (http://bitsavers.informatik.uni-stuttgart.de/pdf/motorola/68000/68000_16-Bit_Microprocessor_Apr83.pdf section 8.1 DC characteristic).
    This alone would reduce the current by more than half as instead of going 5V->3.75 [I supposed you rate the D1 clamping diode of the flash at ~0.4] we are actually really doing 4.25->3.75 so there’s like 0.5V in the drop across the resistor instead of the 1.25V calculated which puts the current to around 5mA across the 100ohm resistor.
    During read operation (post flash) the data bus is driven by the flash chip so there’s no mismatch, the only concerns is for those carts that do have address/control lines still driving the flash (instead of going thru the FPGA) but as it shows the effects are likely less than half of what you put forth earlier.

    For a Z80 I can only find this
    http://datasheets.chipdb.org/SGS/SGS8400.pdf
    and the Vout high is 2.4 volts so I don’t think for those systems is even that bad, maybe the original Zilog was higher that that, not sure, at worse is like the 68K, what do you think?

    Again level shifters are the way to go but I just wanted to point out that the effects on the Everdrive line are NOT as bad.

    Reply
    • July 6, 2017 at 23:43
      Permalink

      Turns out Vcc-0.75 for the 68K is actually a min in presence of pull up resistor (note to self: read DC data correctly), for non pull up the min is 2.4 (like the Z80) and they do not say what the max is, so one would need to measure to figure out how close/far from 5V that actually is.

      In fb msg you also brought up a good point that at least for the MD there may be a custom sega asic in the path of many if not all signals, in such case that’s the chip that aslo need to be checked for V out high.

      Reply
  • July 6, 2017 at 20:36
    Permalink

    Thank you for the write up, very helpful. Do you know by chance if this tends to be a problem for reproduction carts as well? The discussion seems to be limited to flash carts, but I’m not sure how much they may differ technically speaking.

    Reply
  • July 7, 2017 at 07:04
    Permalink

    Nice write-up. Thanks for posting!

    I wonder how the old trusty PowerPak flashcarts fare in this regard.

    Also the new Famicom releases from Columbus Circle, like the 8-bit Music Power carts, seems to have the same issue. They straight up won’t work on some consoles, and from what I remember reading it’s because they don’t downregulate the 5V at all. So even if they do work on a given console it can’t be good.

    Reply
  • July 7, 2017 at 14:05
    Permalink

    If you want an example of a really awful homebrew cartridge, try 8-bit Music Power for the Famicom : http://imgur.com/a/F2WMH/all

    No level translators, no resistor networks, only a voltage regulator for flash chips with a max operating range of 3.6v. Killed at least one Famicom.

    Reply
  • July 7, 2017 at 15:17
    Permalink

    Excellent investigation, Rene. Hopefully this will serve as a wake-up call to Krikzz, to get the EverDrives in line. It looks like he’s been slowly improving them, but there’s still work to be done. Proper grounding and memory voltage should be high on his list.

    Are the 161-in-1s fine after adding the missing regulator? I did this to mine following GadgetUK’s tutorials. I guess that I won’t really be using it any longer after the Darksoft Neo is released anyways however.

    Reply
    • July 10, 2017 at 20:21
      Permalink

      Hi,

      I just replied on GadgetUK164’s vid about the Neo Geo 161-in-1 carts, and it’s possible that changing from the ~3V6 diode dropper to a 3V3 reg might actually cause more current to flow into the IO pins on the Flash chips?

      So, rather than a voltage drop of roughly 1.4 Volts between the supply rail and the 5 Volts on each IO pins, there would be 1.7 Volts instead, so the internal protection diodes of the chips would be dropping more voltage (and hence more current will flow from the console).

      As others have said on here though, a lot of TTL chips weren’t good at driving the outputs fully to the VCC rail, so the actual output voltage may be a fair bit lower than 5V.

      (and then you have the trace impedance and other things, but that normally only relates to the edge rates, and faster clock signals, rather than the “static” DC current when the IO signals are idle.)

      Interesting stuff to actually look at on a ‘scope.

      I need a new scope. lol

      Reply
    • July 7, 2017 at 21:46
      Permalink

      This is 12.5mA per signal which is bad, the measurements I made of commercial cartridges is the TOTAL current for the entire cartridge.

      Reply
  • July 8, 2017 at 13:11
    Permalink

    I didn’t see it answered in here, but is there any way to fix on existing flash carts?

    Reply
  • July 8, 2017 at 20:31
    Permalink

    You go through much of the Everdrive line (and a few others) but what about SD carts for other systems – like Atari 2600 & 7800 or Colecovision. (I’d mention Vectrex, but I’m not sure there is anything available for that right now…talk about a rare console worth protecting from s poorly designed flash cart.) I mean, if I blow up my 2600 with a bad flash cart it would be disappointing, but I’m sure I could replace it economically. Still, I’d rather not blow it up….

    Reply
  • July 9, 2017 at 13:07
    Permalink

    Checking the photo of this cartridge seems to me there is a voltage regulator of 3.3 volts U7, if it is feeding the chips, would not everything be correct? I took a picture of my here and saw that it is a 6206a, and it is a 3.3v regulator

    Reply
  • July 10, 2017 at 07:41
    Permalink

    BUT some level-shift devices just can not solve this issue.

    we did experiments again and again by comparing Mega Everdrive x7(which has level-shifter on it) and Mega Everdrive V1(which just use emi resistors to the socket of console) with UMKT, the highest compacity rom on MD/Genesis platform. and the result was amazing and unbelievable: the whole system include 3.3volt flash that using resistors works better than the one which has level-shifter.

    Mega Everdrive x7 will make lots of uncomfortable noise while running UMKT, however Mega Everdrive V1 plays perfectly!!

    and till now, krikzz hasn’t patch this bug on Mega Everdrive x7~

    so, some design rules are just written rules, i believe the fact which can be tested and measured.

    Reply
  • July 10, 2017 at 16:55
    Permalink

    Krikzz just made some tests with everdrive MD v3 (a quite old everdrive model) and did not found much more power consumption than with normal carts, at least VERY far from the supposed 12,5 mA per pin that you mention in your article to claim that everdrives can damage console internals and cause overheat on the flashcart side.
    https://m.youtube.com/watch?feature=youtu.be&v=mM3BA1qbWpw

    Did you actually made measurements yourself (regarding current flow going through the console woth or without flashcart, voltage applied to flash inputs, etc) or was it just theoretical calculations and observations of the carts design?

    Reply
    • July 10, 2017 at 17:12
      Permalink

      Most 3V3 ICs are designed to send high logic level within the values acceptable by the bus. What should be seen is whether the 3V3 integrated circuit is tolerant to 5V communication on the bus, and many 3v3 ICs are designed to work in conjunction with 5V bus without the need for extra circuits, in doubt just pick up the Code of CIs used in the cartridges, consult the datasheets and verify that the bus supports high logical levels of 5V.

      Reply
    • July 10, 2017 at 21:11
      Permalink

      Krikzz did not do what I would consider to be a good test there – he should have run the game for 30 seconds and then run the same game in the Everdrive. I’m no electrical engineer, but I really suspect that running game code might require a bit more electrical power through the cart instead of a static screen that says that there is no card present – I mean, there is no activity on the cart. But that’s just me. I’d be interested to hear Rene’s thoughts.

      Reply
      • July 11, 2017 at 03:19
        Permalink

        That would not have changed anything or much, because:

        1) cartridge bus activity is not really related or proportional to what is happeninv on screen. CPU is what is driving the address lines and reading data from Flashcart and it still runs the same if this is a static or animated screen, i.e read new instructions/data from ROM area every 4 cycles at most. Actually, the least complex instruction loop (NOP) is what would stress the cartridge bus the most, not the opposite.

        2) current consumption is more directly impacted by the number of pins set high on average than number of cartridge access. To test maximal current consumption you would have a test ROM that continuously read FFFFh from address 3FFFFFh (which is max ROM address) but that wouldn’t be a realistic test case either since that’s definitively not what normal games are doing.

        Reply
      • July 13, 2017 at 13:16
        Permalink

        I like that video, it would have been informational to also measure the duration of the peak (how many nsecs, it looks like 50% of the entire “square” pulse). ESDs are supposed to be able to drop 0.5kV up to 200mSec to rptoect the chip.

        When /OE is high wouldn’t the flash chip databus be in HiZ? (you measured D12)
        As such what’s sinking the current? Are we sure it is just the ESD clamping diode and not something else? In HiZ the pins should be theoretically isolated but I am not sure how it happens in practice.

        I get that the extra 6mA are what worries you (0.6V per 100ohm), at 16 pins for the databus on a 68K with a random probability of being high/low leaves 8 pins on avg high when the internal RAM access happens each sinking 6mA = 48mA total.

        Reply
  • July 10, 2017 at 17:52
    Permalink

    I’m going to rebutt this article for a moment. I first read this the other day but I think the risk to hardware is grossly exaggerated. If anything this article is perpetuating fear mongering against unlicensed devices, whether the source is Far East bootlegs or botique flash carts…

    I can and do run clones and everdrives in my hardware, as you well know from my Music Machine dumps. These flash chips are not shorting the CMOS logic to ground; in fact there’s barely more than 1V of droop on the output if that, and I’m skeptical they are really pulling anywhere near 12mA per output. Yes, the voltage output will droop a bit, but it’s no worse than original cart hardware that exibits “bus conflicts” or other issues. During a “bus conflict” event, the on cart logic circuit attempts to write a different value to a register compared to the CPU. Bus conflicts can and do occur on bidirectional busses, and typically with CMOS the grounded output wins over the high logic output. Since the high outputs are weaker than low output, and it is the high output being pulled low, if CMOS or TTL logic chips (yes they are different families with different characteristics but I’m not going down that rabbit hole) are designed robustly enough so that during a bus contention event, they fight each other, pulling high outputs low enough to register a low signal on another input (LOW logic <= 0.8v according to specs), without damage to either logic chip, then I don't see how a 5V CMOS chip, which does not see damage when a high logic output is pulled down to below 0.8V during a common bus conflict event, is going to see any damage at all when a high logic is pulled down to 3.9V. The maximum current ratings is the maximum current the chip is capable to supply while maintaining a proper logic level within specs. Exceeding this threshold the chip won't burn up, but the signal level will experience brownout. These electronics are built by design to have relatively high output impedance characteristics dealing with signal logic, such that overloading the output will never cause enough internal heating to burn anything up. The chip may run say five degrees warmer above ambient temperature, but certainly it won't get scalding hot, and certainly not generate enough heat to achieve a thermal runaway scenario or reach breakdown temps at the silicon junction.

    Room temp 25C. Body temp 37C. Water boils at 100C. Solder melts around 200C. MOSFETs break down above 300C.

    There are exactly two ways to destroy MOSFET transistor junctions, one is to exceed the breakdown voltage, ie through an ESD event, which occur well above the nominal 5V or 3.3V operation. The other failure method is extreme heat. Enough to cause the junction to literally break down and release it's magic smoke. Sure, operating logic at well above rated voltage will cause thermal issues. Twice the voltage is equivalent to twice the current and four times the heat dissipation in a purely resistive load, but it is worse with semiconductors. Silicon diodes have a nominal voltage drop of .6-.8V, and any series of logic will have series voltage drops in addition to some resistance threshold. Generally the junctions also pass more current when changing state than when steady. So small increases in voltage will consume more current and power than small increases to a purely resistive load. But the 5V logic isn't being operated over 5V, it's just sinking excess current into the 3.3V device. And look at how insanely huge those DIP 5V chips are compared to the 3.3V SMT parts. That cockroach sized CPU or PPU or whatever vintage 5V processor could dissipate an order of magnitude more heat before it burns up than that tiny housefly sized logic chip on your unauthorized device. I would be far more concerned with the safety of your unauthorized device running at 3.3V and interfacing with a higher voltage logic bus than I ever would be with your console's 30-year-old 5V hardware.

    End Rant.

    Reply
  • July 10, 2017 at 21:22
    Permalink

    What about the súper UFO 8 Pro?
    Has somebody tested it?

    Reply
  • July 11, 2017 at 04:38
    Permalink

    >there is no activity on the cart.

    There is ALWAYS activity on the cartridge bus because the CPU continuously read instructions (even if it is an endless loop doing nothing much more). It just says there is no “SDCARD” installed but this is irrelevant since SDCARD is not accessed anyway while playing flashed game (since the game is obviously copied to Flash ROM). Here the console is running Everdrive OS directly from Flash ROM so the cart is drawing current like a normal cart and it IS actually a valid test.

    >I’m no electrical engineer, but I really suspect that running game code might require a bit more electrical power through the cart instead of a static screen
    A static screen is still running game code (i.e CPU is still active reading and executing instructions). The only difference is VDP activity but this is irrelevant to extra current generated from cartridge bus activity (unless VDP is doing DMA from cartridge bus but again, this happen for very short period while screen is inactive, not when screen is active).
    The ‘complexity’ of instructions being executed is also irrelevant to current consumption, what matter is the number of accesses made to cartridge bus on average (short/basic instructions would actually maximize cartridge bus activity, contrary to what you seem to believe) and the average state of address / data lines (VA0/VA23 & VD0-VD15) input / output from / to cartridge (which is pretty much random and game dependent).

    Reply
  • July 11, 2017 at 11:36
    Permalink

    https://youtu.be/kx3M9Z4RAok

    Sonic 1 on a real cart followed by an Everdrive, played for more than a minute each time. 25-30mA difference on a digital multimeter.

    Said Krikzz on his forum: “This difference mostly due the normal power consumption of own cartridge components.”

    René?

    Reply
      • July 11, 2017 at 23:45
        Permalink

        Thank you.

        I hope Krikzz will not mind me quoting him again:

        “As for mega ed v1: testing of this cart will not give clear results, because it use SDRAM and FPGA which has pretty high power consumption”

        Were you using that version of the Mega Everdrive?

        It’s interesting that your measurements for the Mega Drive itself are so different. I wonder if that’s because someone is measuring incorrectly, or because the Model 2 is much more efficient than the Model 1?

        Reply
        • July 12, 2017 at 00:37
          Permalink

          Scratch the last two sentences in the reply above. I didn’t notice the line in your table where you measured the MD Model 2 at 300mA. So, you and Krikzz are getting similar measurements there.

          Now, it’s just a matter of whether your 130mA reading for the Mega Everdrive is due to the normal power consumption of those components, or due to the diode clamp shorting-effect that you described in your article.

          Krikzz has posted videos of the Turbo Everdrive 1.0 showing about 30mA increased current consumption over a regular Hucard, and of the Master Everdrive showing about 30mA increase over an empty system during play, 40mA during flashing.
          https://youtu.be/e-oJSp1-KRo
          https://youtu.be/iU2izK_XCwQ
          https://youtu.be/d_kjD6bzcxc

          Reply
      • July 12, 2017 at 03:53
        Permalink

        So, what exactly your measurements shows, except like normal power consumption of cartridge components?
        MAX-3000 at 50Mhz
        Cyclone II at 133Mhz
        MT48 SDRAM with forced overabundant refresh cycles.
        Look at datasheet and you will see that it is pretty much normal power consumption for such hardware configuration

        Reply
  • July 11, 2017 at 18:01
    Permalink

    Can you give a verdict on EDGB?

    Reply
  • July 13, 2017 at 09:50
    Permalink

    Merci pour cet article qui a certainement dû prolonger les jours de ma Famicom! J’ai commencé ma collection avec ce fameux multicart 150-in-1; désormais je me contenterai des cartouches officielles. Pour le reste (Little Samson et compagnie), l’émulation suffira.

    Reply
  • July 14, 2017 at 11:48
    Permalink

    Wow. Awesome article, and fantastic website! I’ve taken photos of a Mega Everdrive X3 for you to inspect. Throw me an email and I’ll send them to you the full resolution images to add to your collection (hopefully you will give it the all-clear; what a terrible time for me to have bought everdrives…all so far are affected…)

    Reply
  • Pingback: Flash Carts Could Be Slowly Killing Your Retro Consoles – ABC Content Board

    • July 16, 2017 at 15:12
      Permalink

      Bootleg GBA carts?
      The GBA uses 3.3v logic, so there’s nothing to worry about. 3.3v chip, 3.3v console. Safe to use.

      The GBC carts on the other hand may be an issue as it’s a 3.3v cart on 5v hardware.
      Also, I’m not sure if it’s safe to use with a GBA.
      If you are concerned the GBC cart might kill your Game Boy, buy a GB Boy (and replace the crystal with a 4.19 MHz one while you’re at it) or a GB Boy Colour.
      Right now the GB Boy is $15 and the Colour is $32 on Aliexpress.

      Reply
  • July 17, 2017 at 03:32
    Permalink

    Ah, so maybe it’s a good thing I’ve skipped over these Flash Carts.

    Reply
  • July 17, 2017 at 15:29
    Permalink

    Hi, Nice work and explanations.

    I wonder if the Chinese PCB designers have learnt?
    I looked at two Multicart PCB’s, my electronics knowledge is limited so I might have missed something but they both feature 3.3v regulators. Vout from the regulator then goes through two capacitors before going into the Flash chip. I followed the traces with a multimeter and looked up the pinouts of the flash chip, regulator & NES cartridge port.

    So in these two cases the Flash chip is being supplied with 3.3v. So are these two (relativity) ok to use?

    Here are pics:

    https://ibb.co/gNaUja
    https://ibb.co/gp7vqF

    Thanks

    Reply
    • July 17, 2017 at 15:51
      Permalink

      No, these are not OK to use since there is no interface at all between the 5V side (console) and the 3.3V side (cartridge). They are as bad as the multicarts I have shown in this article. I would refrain from using these immediately.

      Reply
      • July 17, 2017 at 18:12
        Permalink

        Thanks for the comment.

        To clarify for the less electronic savvy amongst us, there are at least two issues?
        1. Supply voltage to the flash chip/and or other chips. Console generally outputs 5v, chip needs 3.3v. Solve using regulator?
        2. Data signals from console are 5v. This cannot be solved with a regulator. Resistors can/have been used, but a better method is a signal level converter?

        Thanks

        Reply
  • Pingback: Flash Carts poderiam estar matando devagarinho o seu console! - Blast Processing

  • July 18, 2017 at 09:56
    Permalink

    How about the Game Boy carts? (EMS, Everdrive GB, cheap Chinese bootlegs & multicarts, etc)

    Reply
  • July 18, 2017 at 10:38
    Permalink

    Think I might have a newer version of the board in the famicom 150 in 1.
    http://imgur.com/a/DHiFL
    Looks like they changed flash suppliers but its still a 3v chip, garbage pile for this one.

    There are so many flashcarts, multicarts, repos etc out there. It might be beneficial to start a database/repository with pictures of the PCBs of the known good and bad ones. Most people aren’t going to be able to look at the board and know if its going to kill their console. Whats most concerning at the multicarts like the above since the author mentioned they can kill the console in only a few months, and they are a decently popular item for gamers on a budget.

    Reply
  • July 18, 2017 at 13:34
    Permalink

    What about the Super UFO PRO 8 SD flash cart?? It’s a very common flashcart that we use because it’s way cheaper and plays most of the games.

    Reply
    • July 18, 2017 at 19:45
      Permalink

      The board scans I’ve seen show a MAXII CPLD at 3.3V directly on the SNES bus without even current limiting resistors; pretty bad if you ask me.

      Reply
  • Pingback: Mega Everdrive X7 Review - VideoGamePerfection.com

  • Pingback: SD2SNES Review – VideoGamePerfection.com

  • Pingback: Turbo Everdrive review – VideoGamePerfection.com

  • July 20, 2017 at 09:52
    Permalink

    I think you mean “sparingly” not “sparringly” unless you Canadians spell it differently 🙂

    Reply
    • July 20, 2017 at 11:12
      Permalink

      NO!!! Let’s spare over it!

      Thanks I’ll correct it.

      Reply
  • Pingback: Everdrive GG Review - VideoGamePerfection.com

  • Pingback: Flash Carts, Everdrives & Nintendo NES Bootlegs (The Future of Collecting!) – Cygnus Destroyer | Gamester 81

  • July 22, 2017 at 04:28
    Permalink

    Have you tested the NEOSD AES yet?

    Reply
    • July 24, 2017 at 11:35
      Permalink

      No but it clearly has proper level translators so it should be fine.

      Reply
  • August 2, 2017 at 16:27
    Permalink

    Quick question for you René, and I apologize if its a dumb one as much of these finer electronics details go over my head. I was thinking about the Everdrive GB (link to PCB below), which can be used in a:
    GB-DMG: 5v
    GBC: 5v
    GBA/P: 3.3v
    Is it possible to make such a flash cart that can be safely used in all systems? I would assume yes since real GB/GBC carts are obviously officially supported by GBA/P and I can’t imagine Nintendo would give the green light to something that would damage the console.

    http://i.imgur.com/SnOUr3Z.jpg

    Reply
  • August 17, 2017 at 13:17
    Permalink

    Hello, do you think the boards sold at Infiniteneslives are subject to this problem aswell?

    Reply
  • August 20, 2017 at 16:48
    Permalink

    Wow, I’m glad I read this before shoving the 400-in-1 into my new AV Famicom.

    Reply
  • August 23, 2017 at 16:31
    Permalink

    Aliexpress SD2SNES should be fine? It works like a charm…

    Reply
    • September 8, 2017 at 22:34
      Permalink

      No. Aliexpress sells cheap knock-off garbage.

      Reply
  • August 24, 2017 at 11:11
    Permalink

    Can you recommend any Neo Geo multicarts with proper levels? Or can they all damage your Neo Geo MVS board?

    Reply
  • August 25, 2017 at 17:47
    Permalink

    Is the Everdrive X3 verdict the same as for the Everdrive X5? It looks like they are mostly identical other than the X3 sharing ROM?

    Reply
  • September 7, 2017 at 15:23
    Permalink

    Would this just put more stress on the vregs in the console or more components than that?

    Reply
    • September 9, 2017 at 21:58
      Permalink

      3.3v carts overload the 5v CPU’s address/data lines (Same can be said with the PPU as well, in the case of the NES and maybe the Neo-Geo as well)

      Reply
      • September 10, 2017 at 02:04
        Permalink

        Would that just make the cpu generate more heat? If we put a heatsink on it would it mitigate potential damage? Might just stop using my 161 and 138 in ones.

        Reply
  • September 11, 2017 at 08:44
    Permalink

    Using series resistor networks is a hack that can be a fine solution.

    Assume that the IO from the NES goes through some output buffers (I think I saw 74HC393 on the NES Schematic?)
    Current draw from 5V source (and most older IO sources won’t even reach 5V) is limited to about (5V-(3.3V+0.6V))/100R = 11mA which is usually fine.
    Abs Max on the 74HC393 is 20mA.

    Source: EE with 20+ years experience in industry.

    Reply
    • September 15, 2017 at 10:32
      Permalink

      Great analysis of the console side of things, for 1 console. Now consider clamping diodes in the cartridge, consider that the 3.3V regulator is not a shunting regulator and the effects of sinking current into the rail.

      Reply
  • September 15, 2017 at 10:03
    Permalink

    I would like to know what prevent Krikzz to accept René’s advices and work on revised versions of his Everdrives to fix the pointed issue? Flashcards are amazing things to retrogamers, but collectors want to preserv their consoles… I mean, I thing it’s nothing difficult to add something to fix this problem, and add components to fix this problem will not even “tickle” to the Everdrives’ prices …

    Reply

Leave a Reply

Your email address will not be published. Required fields are marked *