What an odd problem. It looks like some system noise is getting through to the video chip. I'm not a fan of "recap everything" approach but I would start with replacing all the secondary side electrolytics in the PSU. Do you have a scope? Checking the ripple level on power rails out of PSU, and then in near the video output could help narrow down what is causing it.

Do you have any mods in that machine? CPU replaced with something beefy, like DX4 maybe? I could see something like that pull current spikes that the mobo can't properly deal with. Also what monitor is that, what is it using for sync? Combined signal or HV like standard VGA? I belive the RGB outputs, as well as S (combined sync) use electrolytic caps on the output, right before going to connector. On one of my towers all 4 of these were bad, bloated and degraded. Here's what I noted down:

Nominal is 470u/6V3 105C, diameter 10mm, 12mm tall with 5mm lead spacing. Measured on the desoldered caps:
266uF/0,77R
272uF/0,73R
158uF/1,13R
124uF/1,15R

I've replaced them with same size Panansonic low-ESR FR series, 25V 105C 6000h rated. Hope that helps for the video output but that is not going to help you if using HV sync outputs. These are TTL and directly coupled I think. It could be the driving chip that has issues but if vertial has some issues too then I would start with power supply checks first.

The mods I have are a blue scsi, without it it has the same issue. And a self made picoPSU based on the design from Cyo The Vile, which is really just making the PCB fit with the Fresh. And before the normal internal PSU had the same issue as well.

I have a scope, ripple from what I measure is on the 5V line from the PSU 18~23mV and and 35~40mV on the 12V line, measured at the connector. I also checked the ripple of the 5V lines of the DAC, which is ~4mV.

As for the Monitor, it's a NEC Multisync 3D using both HV sync. I also used a LG M1717A, which has the same issues just not the sync but just the line, I believe it also uses HV Sync. Not sure my RGB15 to VGA converter even has S attached. But I have looked at the H and V Sync lines, and they look quite clean to my eyes.

I also have mapped out the analogue circuit from the DAC:

What's not depicted is that the 5V for the analogue circuit comes from a small 7805, which has its voltage coming from a 7808 attached to the 12V line of the PSU

Edit:
Should note, the HSync/wobbly screen issues only really show its head on some pictures, mainly the FMOS gray screen. Lines are more or less always visible if you look hard enough but how visible they are seems to depend on the colour on screen.

Ah, so the PSU was replaced. I like the originals, I repair them if needed. Either way your ripple is not excessive, it should not cause any issues.

My own cables are also RGBHV and I didn't connect S at all. I'm using OSSC with my towers (since I don't have 24k capable monitor) and my Fresh already has VGA connector on it. That's a nice schematic, I was not motivated enough to make one. As I've said I belive the S is using pretty much the same driving circuit as RGB channels, at least the capacitor and filter are the same.

This might be some odd behaviour thats "normal", that is it didn't show on the original Towns monitors but is visible with more modern ones. But for now lets assume something is broken. Double check the H and V signals on the cable with a scope. Not so much for ripple but for nice, clean edges - it should be TTL so a clean square wave. If the slopes seem tilted rather then almost perfectly vertical, especially in the logic shift region (0.8V - 2V) it might cause the monitor to detect the sync with wrong timing. The problem might be most sync pulses might look good, you want to try and somehow capture the pulses that make the lines shift. Might be very tricky. Either a sloped edge or a pulse with different length is what I would be expecting. Or perhaps some tiny glitches before the sync pulse that trip the monitor into starting the retrace early?

It's not just a noise now that I look at it. It seems to come and go, and shift the image by a certain amount that is quite constant across the screen. It's very stable too, moves only a bit along the edge. If it was pure noise then each frame should have a different pattern. Try loading the H line with a capacitor to ground, about 10nF? It will make the signal slopes worse but it will also help filter out any glitches. See if that changes anything. If 10nF completly breaks it, try a smaller cap, down to 1nF maybe.

The sync lines, I don't find any signals that are notably different from these. Beyond a bit of overshooting on the

I've also added a video, the lines aren't really visible in the video (still quite visible in real life) but it seems the horizontal sync issues might be something specific to the green line? Using the background colors of TownsOS I can change the colors how I want without any issues but once Green is above a certain threshold it gives the sync issues.
https://imgur.com/a/qNZkba8

Don't have 10nF or lower capacitors at hand, but was planning to check if the electronics store is open as I need to be near it anyway and need some other components. So I'll add it on the list.

relo999 wrote on 2025-06-22, 22:27:

I've also added a video, the lines aren't really visible in the video (still quite visible in real life) but it seems the horizontal sync issues might be something specific to the green line? Using the background colors of TownsOS I can change the colors how I want without any issues but once Green is above a certain threshold it gives the sync issues.

Your monitor does have Sync on Green. If the colors are not blanked properly all the way through retrace by your Fresh, and the monitor is somehow picking it up, it could explain the issue. And why it's perhaps "normal", I would expect typical Towns monitor to only accept HV or S, not sync on green. In fact the reason why this problem generates such uniform shifts could be due to monitor ignoring green "sync" as long as it's too far off the expected sync pulse. But once it gets into range, it triggers somehow.

See if that sync mode can be over-ridden in the monitor menu. Consider disconnecting the G line in the cable for testing and see if you can reproduce the issue with only R and B. If not it's quite likely the culprit.

Might be 2 separate issues issues. Can't disable sync on green on the monitor but disconnecting the green line solves the sync issue at any green value and reconnecting it brings it back. So that looks like the monitor sees sync on green.

However the horizontal lines are still present with green disconnected.

There is is a full service manual for that monitor linked on this page: https://crtdatabase.com/crts/nec/nec-jc-1404hma
You could mod the monitor, it's easy and reversible: Sync on green is separated from green video by an amplifier chain that starts with Q850. The signal arrives at the base of Q850 via capacitor C809 (10u/16V) from internal G line. Remove that cap and that should disable sync on green but leave everything else.

Restore the cable connection for G if you haven't done so already. After the mod try capturing the photo/video of the issue that remains and maybe we can take it from there.

EDIT: Or perhaps a better way would be to ground the base of Q828. This is what the monitor does internally if it detects other sources of sync but since it's all based on analog delays it's far from perfect if the signal isn't properly blanked.
So the easiest way about it would ba a short (with a wire soldered on the other side of the PCB for example) across R847. Or the CE leads of Q840 - but this might be SMD part, so R847 is probably an easier target.
If you need sync on green for other purposes and such a mod would get in a way there is also an option to make it switchable, the switch could be added on the back of the monitor perhaps. But in order to avoid routing the signals through wires across the monitor it would take a few extra parts. Nothing very complicated, basically a copy of what Q840 does except with the switch on the base of it.

Thanks! I'll look into that when I open that monitor up, need to fix some geometry issues anyway.

As for the other issue, it's hard to take pictures of that my camera picks up. The video shows it the best I think and it was easier to record on the 1717.
https://imgur.com/a/TUiUNT8
It happens regardless of color, but it's less noticeable on very bright or dark colors. Midtones show it the best.

The lines are also in sync with the sync issue on the NEC. It's probably best seen in the video here:
https://imgur.com/a/0TroQbg
But even without green attached those lines still exist (without the sync issue)

Just making sure I'm seeing it right - you mean these darker bars that seem to move vertically across the whole screen? That does look like some ripple, or perhaps even 50Hz mains hum on the screen.

So first thing to rule out would be the cable. Two different monitors, but how connected? If you are using the same cable or adapter, it's suspect. Could be not enough GND lines or unshielded RGB wires for example.
Second, and this is also a problem with original Towns PSUs. There is no earth connection on the plug, it's usually 2-prong plug. Japanese use a metal screw on the back of their devices, as well as in the wall socket, to connect extra wire if earth connection is needed. I worked around that (when repairing my towers) by replacing the old and somewhat damaged 2-wire mains cable with double-insulated 3-wire, with US style 3-prong plug. This way I have earth connection but can't plug it into EU outlet by accident, only to my 230V->100V transfomers.
TL;DR: Add earth wire to your Fresh, see if it helps any. The monitor should have it's own earth connection via the power plug.

This particular problem might turn hard to solve, if at all possible. If the noise originates in the DAC itself there is little you can do about it, even mobo mods are rarely helping. This might again be a problem of having too modern monitor, that can resolve more color shades than the originals could. But before that conclusion there are some things to try.

Sadly grounding didn't help any, also it's kinda hard where I live it only became a legal requirement for newly build buildings in the mid 90s here and where I live only the bathroom and kitchen are grounded. I tried different cables one that I know is shielded well and using the converter from my X68k but no luck, same issue.

Might be good to know, changed the caps of the analogue circuit (again) to the Nichicons VZ's the store sold me which I don't think are low-ESR but probably better than the Jamicoms I had in there for testing.
And rather than showing the sync issue at green value "9" at it's earliest it's the green value "11" from the background colour picket.

I also doubt the monitor is to modern, seeing as at least the NEC is from '86 and the Fresh from '94. The NEC does have a colour amount selection toggle, but that's only for TTL.

Can't be '86. I know the page says so but it also says it's a successor to NEC Multisync II, which was released in 1990. The manual mentions PS/2 and 8514/A, both of which were released in '87. And it's a 1024x768 CRT, with high refresh rates, that resolution alone would make it king before 1990. It has to be early '90 because at the end of that decade monitors started going digital (OSD, buttons instead of pots).

Anyway, it's pity the grounding didn't help but I wasn't couting on it much. BTW I have the same issue, the building I live in is old and has only live and neutral. Grouding bolts are tied to neutral - which is earthed, in the basement somewhere, but since it's not a separate wire in the socket all the noise that goes via N also affects PE. So it's good from electric safety point of view, but shit when it comes to noise suppresion.

Take a look at the schematic you posted. There's 2 driving transistors there, both look like emitter followers. BTW those two caps marked "?" at the bottom, connected to all 3 middle transistors, those should also be connected to some power rail. Otherwise it doesn't really make much sense. It should be something like top right corner, a pi filter with cap, choke and another cap or two. Probe (with a scope) both these points that power the transistors for ripple. At the end of the RGB lines, in the monitor, there should be 75 ohms resistors to match the line impedance, so when loaded like that these lines swing 0.7Vp-p. And at minimum there's 64 shades (6 bits) of resolution, so thats 0.7/64~=11mV or so. So any ripple close to that 11mV could perhaps be seen on the monitor as bars of different intensity.

EDIT: When measuring ripple pay attention where you connect your GND lead. Pick a spot near these transistors or you will get wildly different results.

You might be onto something
https://imgur.com/a/FQXbmyJ

The power rail has around ~10mV ripple and seems to have some noise from a video or H-sync signal. Used the ground from the video out as that's the easiest place nearby to attach the alligator clip.

Also I've updated the schematic, it's connected to the same power rail as the other transistors via a via.

Well, 11mV is not great but at the same time the emiter follower configuration should be passing very little of it to the output. This is right on the level of care/dont' care. You can try replacing those two 100uF caps in the pi filter with low-ESR variants, that should improve the ripple but probably won't resolve the problem.

Noise from H/V pulses should not affect the output. The pulses are outside the visible area so unless the monitor has some sort of automatic black level circuitry (and it shouldn't) it won't affect it. I wonder how well Fujitsu routed GND to the output connector, they had issues with it in some models. There are even factory rework wires to correct some it it. Try measuring ripple between the GND of the video connector, and GND directly at the output of the PSU. In theory should be zero but you will see some noise there. I just wonder how bad it is. Note the video connector metal outer shell is also tied to GND but usually not at the same place as the signals - so it can't be used for noise measurement. You need to use one of the GND pins in the connector itself preferably. In fact you can also measure the ripple between the pin GND and the outer shield but it should be minimal, just about the scope own noise level. If not it could be a problem, depending on how the cable adapter was made.

Might be worth investigating the DAC again. It too should have some filter for at least the analog section of it, perhaps it could also use low-ESR caps. Even if the noise is smaller there, all of it is passed to the output. You mentioned 4mV which I would not expect to be so visible on the screen, but who knows.

Will need to order parts, which will take a few days hopefully less.

But grounding between the grounding of the pins and the grounding of the connector housing (screw post), it's quite big I think. If I zoom in enough to get a relatively stable wave form it's 162mV, If I zoom out to attempt to get a frequency (it's all over the place) It's between 80mV and 250mV, and seems to average around ~130mV

Test your cables and any adapters again. Make sure there outer shield of cable and connector is not connected to the signal grounds in the cable/adapter. If it is, all that ripple is going to be injected into the signal ground, which is the same as having noise on the signal itself. If there is such a connection, it'll have to be changed to remove it. In ideal case each of the RGBHV(S) should have it's own separate ground wire all across the cable. Some cheaper cable combine some of these together, that's not optimal but not always bad if there is more than one GND wire. For example single GND wire for RGB channels, and another for HV(S). And a separate shield, not connected to any GND pins but only to connector shield. That's usually good enough for up to VGA resolution, though preferably the RGB should be mini-coax each, with separate grounds (using the coax shield) to the correct GND pin on the connector.

Assuming the cables are good - consider disconnecting the outer shield on the Towns side from the connector. This will still provide enough RF noise suppression from the monitor side, and possibly avoid any of that ripple being injected into the signal. Depending on how the montitor has the grounds routed, internal cable length, etc, it might be a factor. Here's some crude ASCII art to illustrate:

FM Towns ---> monitor
R signal ----- R signal
R GND ---- R GND
G signal ----- G signal
G GND ---- G GND
B signal ----- B signal
B GND ---- B GND
H signal ---- H signal
V signal ---- V signal
H,V GND ---- H,V GND (better if separate but combined is quite OK too)
connector shield -xx- cable shield ---- connector shield (disconnect on Towns side)

That's mostly what I have on the NEC using the attached wire, where they have all wires separate and shielded. Except all my converters have no shielding, not the connector shield connected and all ground wires put other.
The 1717 I use the same converters for but uses wires with RGB connected with the connector shield (but unattached the FM thanks to the converter not having that connection), and H and V grounds separate.

The converters are all around 7cm in length. Would it be good to try making a converter that has the grounds separate as much as possible and if I can with shielding?

Probably won't help, but are you properly grounding the computer by using the wire screw jack terminal?

Yep, didn't help.
But I found a way to solve the issue, as it IS a grounding issue by the looks of it. I ran a extension cord from the bathroom to my workbench and used that to power the system and the bar issue is gone. Wiring the ground from the Screw terminal from to the ground on the extension cable did nothing while powering it from my normal plug. If I used the original PSU it'd probably work I imagine. But it looks like the PSU I use to power the Pico PSU need to be grounded.

Sync issue seems to have gotten worse though with grounding the PSU, back to "9" and seems to be a bit more violent .

relo999 wrote on 2025-06-25, 11:57:

The converters are all around 7cm in length. Would it be good to try making a converter that has the grounds separate as much as possible and if I can with shielding?

It might help. It really depends on what is causing the dark bar issue in the first place, and hard to be sure without trying. The shield for the cable, as such, is optional, it only prevents the cable from radiating the H/V pulses as noise to nearby radios and TVs. The cable will not pick any significant amount of noise from outside, shield or not. But separate grounds can help, because at higher frequencies (like video signals) the returning electric currents tend to pick not the shortest route, but closest to the signal path. So having all GND points tied together in a cable or plug will defeat any balancing done on the PCB inside Towns unit, and the monitor.

As an example, here's the cable I used to make my own Towns DB15 to VGA connection: https://www.tme.eu/en/details/tas-tsk1073/vid … tasker/tsk1073/
It says 50, not 75 ohms, of wave impendance but that is not a problem over a few meters. I use these 3 coax cores for RGB, the inner wire for signal and the outer shield for GND for each. Then I use a single wire (white) for H+V ground, brown for H-sync, red for V-sync. There is no extra shield over the whole cable, so the connector shields are not connected on either side. This will radiate some noise from the V/H lines as I mentioned but it's not a problem for me, and the picture quality is great. Much better then a shielded but cheap VGA cable that had RGB routed using single wires (not coax) and all the grounds were tied together. I had visible ghosting on the screen with such cable.

I now obtained a good quality VGA cable, with coax for RGB and also extra shield for the whole cable. I can make another cable now, but frankly the one I'm using is great so I have zero motivation to be soldering another.