• Hardware considerations

    Julien4 hours ago 0 comments

    While I’m trying to figure out the specific vga timing I decided to put down the current progress.

    The plan was to add VCAs to the synth. However, without surprises a LM13700 can’t handle 1Mhz (not really designed for it). The only «ready » solution at those frequencies is the LT1228 from Linear. However it costs 8,88 euros. Needless to say, it’s not for me.

    Another solution is to build a vca from discrete components. Simulation are being made with what I called the Yusynth VCA although it is a standard VCA design seen a lot in Synth design. A major drawback is the offset of 500mV at the output. I’m wondering if there will any VCA in that project after all, this will give 5 VCOs instead of 4.

    This led to another constrain, the bandwith of the op-amps throughout the synth. The bandwith and the slew rate should be sufficient for the minimun 1MHZ frequency. It’s a nightmare to chose an opamp from parametric search because there’s so many. I reduced the selection by taking into consideration only parts that can be bought on aliexpress or ebay. Remained the LM6172 or the MAX4392 that works only with +5/-5V which may lead to redesigning the VCO. The synth will run with +12/-12V, I’ll stick with the LM6172 for now, besides is the same price as the MAXIM part.

    The pin 3 of the XR2206 controls the frequency. It is fixed at 3V and the design should not pull more than 3mA out of it. The problem is to not to exceed those specifications while modulation is applied to the input. The first solution is to clamp the signal when it needs to be in order to not fry the XR2206. However it requires more components and is still not ideal in terms of use. Another solution is to limit the range of the manual frequency control whereas the highest and lowest frequencies are only obtainable with modulation applied. I took this solution.

    The same problem applies to the « fine » frequency control when, again, modulation is applied. Therefore I removed the « fine » control.

    The PCB has been designed for cost and size.  Jacks and switches will be linked with wires.  A prototype area will be installed in the bottom empty space of the board


    Julien01/13/2018 at 21:19 0 comments

    The VCO are an essential piece of the project. They will produce the shapes and contours of the images. The VCOs will be all the same. They will have a large range from 1 or 2 minutes to 1Mhz (despite the fact that the waveforms are less neat around this frequency). 

     1Mhz is actually a small limit for the VGA scanline. Meaning that once locked the vco  will produce less vertical bar that it can produce horizontal bar (vertical sync). This is the result of a scanline twice faster that the one used in PAL or NTSC (about 15Khz). Although test will be run to run the clock generator lower.

    The VCO is using the XR2206, obsolete but still available from China. Are they counterfeit? Clone? Rejected part at the factory? I don't really care as long as they work. Some external corrections are needed to constrain the waveforms into what it is expected for the project.


    To get stable image we need to lock the vco to the monitor clocks (H sync or V sync) and the XR2206 hasn’t got a dedicated pin for it. I manage to get the VCO to reset but  it creates a «tick» in the waveform that should be removed for processing. This is done with a clamp circuit that reduce the negative peak. The two other op-amps are here to make the waveform into the 1VDC signal range.


    Approaching the MHZ range, the square output tends to become something else. I needed to create a kind of « reshaper » circuit to get back a proper square wave or at least something less ugly.


    Julien01/10/2018 at 13:28 0 comments

    For the display, two things are mandatory. Video output drivers and blanking.

    Each vga color wire is terminated by a 75R resistor,  hence the need for special op-amp able to drive such high load. I'm using the LM6181 op amp which seems to be now obsolete. I may switch for the MAX4392 or equivalent if I need to.

    The blanking is done by using the horizontal sync signal. Documentations show a particular signal timing for the blanking of colors but my tests gave the expected result on two different monitors. Maybe the monitors create the needed blanking.

    Considering the limited space in the case I'll probably use (an old tool box) a switch is present to send a monochrome signal, letting the color's VCA (if there's any) be used for others tasks. Monochrome is achieved by driving the RGB signals with the same signal (all three colors at full brightness give white).

    For the prototyping stage a cheap vga signal generator from ebay is used. It gives timing signals and color input. It'll be replaced by a STM32f0 microcontroller.


    Julien01/10/2018 at 12:50 0 comments

    I've been interested in video creation and modification for a long time, right after I've read about the Sandin project, the synkie project and LZX industries modules.

    All use the composite analog video protocol (PAL/NTSC). It's a complicated signal that combined the synchronisation signals needed for the monitor and the color informations into one signal !

    This signal contains digital (sync signals) and analog parts(colors).
    It requires large bandwith that most op-amps can't handle, at least 3Mhz.

    Being an old analog protocol, it tends to disappear. Analog tv is a thing of the past now and monitor with composite input will certainly vanish as well. I don't even need to mention the death
    of cathode ray monitor.

    Two reasons for the choice of VGA, the obsolescence of the PAL and the simpler protocol of the VGA. The VGA protocol is simpler, the monitor receives all the signal on different wires.

    I want to keep the all project simple and as cheap as possible (the most difficult part I guess).

    I decided to take the LZX Industries standard for signals (1Vpp DC)