[WIP] RCA Sync to 14-pin Sync (mostly)

Forum DIY
21

Thanks for all of this extra background info. I really only kind of, somewhat know what I’m doing. My assembly skills are far superior than my EE skills.

I’ve got a second Castle VCO that I’ve done the resistor swap on and will be able to compare them side by side. I’ve got a lot of moves to make on this case and honestly, the power supply is hot garbage. I should replace it but that’s a whole other can of worms.

Either way, after a year+ of building this system it’s finally starting to come together.

IMG_6940
IMG_69401389×1851 939 KB

22

Hi guys. There has not been much progress on this project, so I thought I would share what I am up to. (I am also @Analogmonster but I am moving away from using that username).

I have had a batch of LMH1980 VSSOP to DIP adapters assembled at JLCPCB.com for various experiments. I have a full Cadet system and some Castle modules, but I also have Syntonie VU007B, so I am starting to get a mix of bus sync and RCA sync modules.

I want to stop using Cadet 1&2 and have VU007B as my sync gen modules using HD component video.

Using an LMH1980, I can get C, H&V sync. I could get blanking maybe too with a bit more research into one-shot timers, but I think it is unnecessary complexity. There are only 2 modules I know that use blanking on bus sync - Cadet 2 and Syntonie VU007 RGB2Component. If you have VU007B you don’t need either of those modules anyway. All the other modules I am aware of only use H&V sync.

1 Castle module uses sync, the clock VCO. This gets sync from the Eurorack power bus on the gate and CV pins. The LMH1980 signals are inverted then scaled to 1V and buffered by the opamp before being sent to the Eurorack power bus. The LMH1980 signals are inverted twice (essentially a CMOS buffer) before being sent to the sync bus as 5V.

With an LMH1980, inverter and video opamp circuit on breadboard I have got functional H and V sync. I had to use a CD4049 for now as that is all I had in stock, but the inverters will be swapped for an HC14 when I next do a Mouser order. Tested using VU007B with Cadet 4 (Ramps), Cadet 9 (VCO) and Castle 010 (Clock VCO), but so far only as composite. Need some 1/2 watt 75 ohm ressistors to mod some circuits before I go HD. Not sure how the modules that were designed for composite will handle the Tri-level sync. Have some VU009 oscillators with bus sync that are in my to build pile that I will test soon too, most parts are in stock just need to build them.

Haven’t implemented the frame sync yet, but Lars posted how to do that on here somewhere, just need an inverter and an AND gate.

I’ve looked through all the legacy products on the LZX site that require sync. It is unclear whether Diver and Fortress use RCA sync, but I believe so. They’re the same period as Escher Sketch which definitely uses RCA sync. Video Ramps uses bus sync, but it seems to be a non DIY version of Cadet 4 anyway so although I can’t test it I will assume it will be compatible.

Placed a new parts order at JLCPCB of LMH1980s today, takes a while for them to actually receive the parts. Will get them to do the PCBA for the LMH1980 only, the rest will be solderable by hand. I’m going to lay out the PCB in the next few days when I have the spare time. If it is fully functional, I will have a few for sale as I only need 1.

WIll have front panel H sync, V sync and frame sync on jacks. Rear will have RCA sync in and through, sync bus out and Eurorack power in. There will be jumpers to select whether to put sync on the gate/cv pins, and a jumper to terminate the RCA sync if needed. I might have another item or 2 on the front panel, depends on spacing in the layout.

Thinking about having the opamps run off -+5V as well has have dual DIP8/SOIC8 footprint for maximum opamp compatiblity. LM6172 is no longer marked end of life but is painfully expensive - more than double what I paid a year ago. TL972 has a max supply voltage range of only 15V, but has acceptable slew and bandwidth and is literally 18 times cheaper! Even factoring in regulators, that is significantly cheaper. Sadly, through hole TL972 is marked end of life but the SOIC8 is easy enough to solder without specialist tools like a microscope.

Will update when I make some progress

4 Likes
23

Nice. I abandoned my sync format conversion efforts for various reasons, but it’s nice to hear you’ve been working on it.

Do you have any tips for having JLCPCB do PCBA for the LMH1980? That’s something I’ve been meaning to tackle for another project, and it’ll be my first time having anyone do PCBA, so I’m a bit apprehensive.

24

JLCPCB parts are done in a few different ways. They have a basic parts library of things they have in stock. They are already loaded on to the machines so there is no additional setup costs. Any part that is not a basic part, as well as the part cost there is a setup cost of $3 - if you have a reel of 3 or 300 parts, it will be $3 per non basic part.

They have extended parts, which are often not in stock but you can order parts through JLCPCB. The rates are often better than what we have available eg Digikey and Mouser. There will be a unit cost, but some parts also have a minimum order quantity.

If a part does not show up in their own database, then you can use their global sourcing parts tool. This is what I often have to use. Its a search engine that connects with a number of distribtors (around 20). Here there will be the same as before, unit prices and minimum order quantitiies. There is also a column labelled estimated attrition. A part might be hard to pick and place, so they sometimes expect you to order extra to plan for parts to fall off the picking head.

You pick who to order from, you pay JLCPCB, they try to order for you. The price for global sourcing was only a quote. There is a chance the price will change and you will have to pay the difference.

Parts get delivered, depending on the lead time. Once JLCPCB have your parts order, they store them indefinitely. Unless its a really expensive part I tend to order more than I need for a single design if I expect to need more in the future.

One final thing I have only partilally used is their consign parts service. If there is a part you want that doesn’t show up in any of their databases and it fits some specific rules they may be willing to add it to their database. I believe JLCPCB and LCSC are the same company. I wanted a specific tiny RGB LED. Found it at LCSC, provided links so they could check eg datasheet and they approved the part, it is compatible with their pick and place machines. Now if I want to assemble a design with that part, I should be able to treat it the same as the global sourcing way.

25

They’re a pain! I tinned the pads and used hot air in a pool of flux then hoped for the best. Luckily it worked.

I’ve noticed the DIP package has more than doubled with what stock is left. I think the SOIC version is going to be in stock for a bit longer.

26

Yeah I think this is advanced soldering. I can do it because I have suitable tools, but damn it takes practice to work with chips like these. Would prefer to have this an easy enough soldering job for most people.

I’ve been testing Cadet 9, Cadet 4 and Castle 010 with my breadboarded circuit and VU009 as sync source. Some interesting results.

Cadet 9 has been solid, takes half a second to catch up to the different sync sources as I switch between the 12 different sync options of the VU009. Sync is probably so good because anything in the cadet system that uses sync has schmitt triggers, whcih will help with edge restoration and any voltage drops.

Castle 010 seems a little more unstable, but I think its an artifact of using a PLL It locks in perfectly at some frequencies though so I think this is how it should be. Castle 010 does not have sync inputs buffered by schmitt triggers. Instead, sync is buffered by a transistor and would handle 5V no problem. I took the sync singals straight from the inverter, sending 5V sync signals through the power header and it worked just fine. Think I shall keep it this way.

Cadet 4 is more of an interesting one. It was designed with SD interlaced signals in mind. A constant current source charges a capacitor linearly, creating a ramp. A sync pulse resets the ramp to zero. Through a bit of maths, a ramp down and a double ramp are created. A trimmer allows you to adjust the ramps so the double ramp is centralised.

I alread knew that if i switched between PAL and NTSC, the ramp position moved. Makes sense, it’s analog and the sync pulses have different timings. Now I have VU009B, Even more sync timings mean different ramps according to format. I always get a gradient, but they can be shifted or there is no double ramp.

Wanted to see what others are doing. Don’t own them but read the manuals. Syntonie Rampes is analog. Appears to fix the progressive issue but does point out that there will be some shifting between picture formats. LZX angles does not imply it has this issue, but I am wondering if they resolve the issue by analysing the sync source and digitally generating the ramps with perfect timing. Gen3 modules seem to use FPGAs alot so I can see an FPGA or MCU design being an elegant solution to the issue.

If I want to work with my external cameras, it must be 480i, that’s the camera’s only format and I don’t have a frame buffer/TBC. If I want to do anything else, I want it to be 1080p. As a future project I might have a go at modifiying Cadet 4, either with an expander or it may end up as a redesign. A way to change a few parameters depending on sync format so that the ramps have a consistent position

27

I tried that too, it worked fine; no problems at all. I added an opamp buffer to my outputs just to be safe but honestly my output doesn’t seem as sharp with the buffer but it still looks good. You’re probably a-ok keeping things as they are but I’m nowhere close to an EE so grain of salt and all that.

28

Oh, interesting thanks. What op amp did you use for the buffer?

29

20 characters of LM6172

30

That’s right! To be more precise, Rampes could be described as a digitally controlled analog ramp generator. It is based on Cadet 4, and there is a multiplexer switching the resistor that sets the current which charges the capacitor/sets the ramp timing. The multiplexer is controlled by an FPGA that detects the system resolution/framerate, and switches the resistor accordingly.
So there is a bit of imprecision because of component tolerances and some temperature dependency, meaning that ramps are not perfectly aligned across all formats, and also requires some trimming. Then I thought it was good compromise between having a “user settable” ramp generator, where the amplitude and offset of the ramps would have front panel controls so it can be set precisely by the user, and a solution that involves a DAC, that would have cost a little bit more and would have been a lot more development time for me.

That’s right again! As far as I know, all Gen3 ramp generators (DSG3, Angles, Scrolls) are based around a FPGA (for format detection and digital ramp generation) and a DAC (for digital to analog conversion). And if I’m not mistaken, Visual Cortex ramp generator was done in a similar way.
This allows for really precise timing, could even be said to be “pixel perfect” since the FPGA surely runs off of a pixel clock derived of the system resolution/framerate.

For a DIY ramp generator, the “front panel tunable” ramp generator as mentioned earlier, would probably be the most straightforward solution. That was kind of the idea behind VU009 (sawtooth oscillator), being able to set it to one line or one frame cycle so it can be used as a pseudo ramp generator. Then the resulting ramp is not super linear, and there is also some jitter when set to vertical sync, which makes it a bit hard to precisely set to do a one frame cycle.

2 Likes