Surprised you've done that much routing before finishing placement to see how much things can be pushed together to reduce overall size (I'm assuming there's quite a bit more to fit onto that layout). Seems to be quite a lot of unused space near the BNCs that the bottom-left corner stuff could be pushed into.
As mentioned earlier,people may want to display it, so there should be room to lay the regulators and tall caps flat to allow a sensible distance to a clear cover.
You certainly also want a few more mounting hole locations for a PCB that size, and some near the BNCs for support in case the panel-mount isn't used.
Using a discrete crystal or resonator in place of the coil for your master oscillator still keeps your circuit a valid all discrete design.
It will also prevent users from accidentally destroying their TV h-sync yoke driving transistor on their TVs if the inductor driven oscillator doesn't drift the H-sync by 1/4 of a KHz. Modern digital sampling screen's ADC use tightly locked crystal PLLs, they wont be as forgiving with out of spec HSync speed compared to an all analog CRT like the Commodore 1084s monitor you are using.
The master oscillator with the current crappy inductor is already stable enough for any analogue TV or monitor and will be amply so with a better inductor. I can't immediately see how any TV could burn out its yoke due to sync frequency drift. In the absence of a video signal the raster scan oscillators are without any signal to lock onto at all and are are free to rail out or drift around within their operating frequency limits. Back in the day hobby circuits published in electronic mags for producing video displays often used a pair or 555 timers injection-locked to each other to provide the H and V sync! I have two such "vintage" articles sitting on my desk right now and I can't recall ever reading a warning to monitor your TV for smoke in addition to a stable image whilst tweaking those 555 frequency-setting trimpots!
Resonators typically have crappy tolerances and limited scope for frequency tuning and I'm not sure if they're readily available in 500 kHz. I'm going to buy one of those composite video to HDMI converters to asses how sensitive it is to the field and frame frequency stability/accuracy. Until it is proven that the basic LC oscillator is inadequate it stays. If it does prove to be inadequate or marginal I will substitute the LC oscillator with a 2MHz crystal oscillator followed by an additional pair of toggle-flip-flops to divide down to 500 kHz - 2 MHz is the nearest readily available crystal frequency, making that the only real highter-stability master oscillator option.
I put the soldering iron away this afternoon and made a start at documenting this discrete-transistor Pong MkII. This is mostly because the circuits were part scribbled on multiple sheets of paper and part in my head. By alternating between soldering and documenting the design I have less chance of getting lost.
Here are complete details of the horizontal timing and video generator board. A master LC oscillator of 500 kHz clocks a 5-bit binary ripple counter made up of cascaded toggle flip-flops.
The necessary video signals are simply decoded by (N)AND gates. Some trickery was required here however as this method of decoding a ripple counter is prone to glitches in the decoded outputs as the clock state ripples through the binary counter due to the propagation delays of the flip-flops. Glitches on the decoded outputs are not acceptable here as the video signals are being generated live as each horizontal line is being scribed on the display screen.
I solved this problem by designing the flip-flops of the ripple counter to be quite fast (using MPSH10 transistors) and conversely designing and (N)AND gates for decoding to be only adequately fast for the application, rather than excessively so, such that the decoders are too slow, by a comfortable margin, to respond to and pass through the erroneous binary states of the counter during the ripple-through of the clock.
The DTL (diode-transistor-logic) N(AND) gates are made "slow" by using comparatively-lowly BC550C transistors with a relatively weak base pull-up. The AND gates that provide the horizontal video signals for generating the scoreboard display digit segments are comprised of a BC550C NAND followed by an MPSH10 DTL inverter. The through-put propagation delay of this combo is a clean and stable ~120nS and the MPSH10 inverter provides (in addition to the required polarity inversion) a squared-up output with fast rise and fall times.
The MSB (HR16) of the horizontal ripple counter is at the line frequency (15,625 kHz) and it is used as the clock source for the vertical line counter. The horizontal video component of the paddles and the court net are approximately 600nS wide and 300nS wide respectively. As this is substantially less than the 2uS master clock period, these video signals are produced by triggering high-speed monostables.
All of the signal outputs of the horizontal timing unit are digital, bar "H_SCAN", an analogue line-scan ramp signal. H_SCAN is provided for the Ball Horizontal Movement circuit.
The board itself. I currently have a rather crappy ferrite-cored "choke" with a rather poo temperature coefficient soldered into the tank circuit of the master oscillator, as this is what immediately at hand. I'll will be sourcing a better core material to wind a better substitute.
There were some computer monitors that would blow the horizontal output transistor or other parts if they were fed an incompatible sync signal, but most would tolerate it at least for a while.
With the worst possible +/- 2khz error on the 500khz resonator I listed, GK's H sync generator would be from 15.562-15.625-15.6875Khz. The error doesn't even make it to NTSC's 15.734k. For B&W, the listed resonator is a safe cheap 0.7$ reference which wont destroy any monitors at all and would probably drop into his current circuit with ease. He could even cheapen it up with a 2 transistor oscillator instead of the current 6 if he likes. Just accidentally touching a point with your finger on the inductor tuned oscillator will go way further out of bounds.
With the worst possible +/- 2khz error on the 500khz resonator I listed, GK's H sync generator would be from 15.562-15.625-15.6875Khz. The error doesn't even make it to NTSC's 15.734k. For B&W, the listed resonator is a safe cheap 0.7$ reference which wont destroy any monitors at all and would probably drop into his current circuit with ease. He could even cheapen it up with a 2 transistor oscillator instead of the current 6 if he likes. Just accidentally touching a point with your finger on the inductor tuned oscillator will go way further out of bounds.
I didn't claim that an equivalent resonator oscillator would be less stable or accurate than an LC one. Re the design involving those 6 transistors - the LC oscillator requires a frequency trim and you can't get 1nF trimmer capacitors. I biggest I could get at the local Jaycar last week afternoon was 8.5-100pF. I put that in parallel with 150pF and that still gives me adequate range to trim for the inductor tolerance. To get down to 500 kHz with ~205 pF a 470uH inductor is required. This is a poor LC ratio for a typical basic oscillator design as the inductance is large and the capacitance is small. For these values XL=XC=~1500 ohms at 500 kHz, so in an idea case restive loading of the tank circuit must be ~15k for an tank circuit Q of just ten, but in reality higher as the resonate Q is lowered further from the ideal, in this case primarily by the inductors non-zero series resistance.
The Darlington-connected input transistors present a high impedance load to the tank circuit maintaing Q. The high input impedance also helps preserve Q by permitting positive feedback to be lightly coupled (via 47k) to the tank circuit with minimal loop-gain attenuation. The Darlington transistor pair connected in a long-tail pair double operates as a limiting amplifier, giving a squared-up and stable amplitude of oscillation, further aiding stability. The first high-speed toggle flip-flop of the horizontal ripple counter requires a clock source of adequate amplitude and rise/fall dv/dt. This is provided by the second long-tail-pair limiting amplifier, bringing the total transistor count to six.
The whole purpose of this design exercise is to construct and demonstrate something more impressive than a LED-blinkie with rudimentary components. For a high frequency clock source they don't get much more rudimentary than a trimcap and a non-specialized, non-custom inductor.
Would an RC osc be significantly less stable ? Maybe a phase-shift osc?
The trimmer value issue shouldn't be a major problem as you can shunt it with fixed caps to get within the required trim range.
Damn, you've been busy GK! I have a feeling I won't be able to fit this project all on a single large board... It looks like it's going to be *a lot* of parts!
(Though laying out a PCB for this project should be *a lot* faster as my library is already mostly setup and I've learned a lot so far doing the Scope Pong board.)
Speaking of which, I'm hoping to have the first revision of Scope Pong done by week's end.
I am, admittedly, feeling a tad immodestly-full of myself this evening.....
Seriously now, am I just f%$@ing awesome or what?
Damn, you've been busy GK! I have a feeling I won't be able to fit this project all on a single large board
Damn, you've been busy GK! I have a feeling I won't be able to fit this project all on a single large boardVertical resistors & diodes ?
For hand soldering, 0805 isn't really any harder than 1206, though 1206 does allow for more tracks to be jumped over.
Also on a home made board without a solder mask, with even 1 trace under an 0805 could potentially solder short without one knowing it. 1206 is the safer choice and you can keep relatively thick traces.
Seriously now, am I just f%$@ing awesome or what?
Also on a home made board without a solder mask, with even 1 trace under an 0805 could potentially solder short without one knowing it. 1206 is the safer choice and you can keep relatively thick traces.True, but is anyone seriously going to bother homebrewing a PCB that complex when you can get proper PCBs so cheap?
And for a SMD board, there would be a lot of through-links, and allowing for homebrew would mean you couldn't put any vias under components.