Jeremy, thanks for looking, but 6x6 wouldn't be very useful.
I was in the garage yesterday and saw the Invotron 16x4 router that was donated. I believe it's got customised firmware running on it which could make it awkward to drive via its RS232/RS485 ports. There is also ******* all documentation. The interface to the switch elements looks very simple. There are 4x 4042 transparent latches so I reckon it's no more than a 4 bit data bus and 4 enables. If I do decide to use them one day I'd probably pull out the CPU cards (x2, one for the audio switch, one for the video switch) and plug my own controller into the half width DIN41612 connector.
Somewhere I've got an old 8051 class CPU on a board which was a testbed for some of my professional designs. A couple of decades ago I hacked out button scan routines in assembler and still have the code lurking on my PC - these things don't get thrown away. So perhaps one day.
For now I'll finish off my relay switch.
Took a closer look at the Invotron switchers this morning. There's a 1U rack with a 16x4 video router and another 1U rack with companion stereo audio router. Each rack has the router card, a PSU and a slot for a controller. Only the video unit has a controller. There's a 15 pin D connector on the back of each to slave the audio to the video.
I examined the slave interface and it's very simple. Just 4 bit data (1 of 16 inputs) and 4 enables (one for each output). It's at standard 5V TTL/CMOS levels. The data latches are 4042 which are transparent rather than edge triggered. I'm starting to wonder if it wouldn't be simpler to hook up some switches to this interface than continue with my homebrew relay switcher.
What's not so impressive is the switcher itself. Ordinary 4051 8:1 multiplexers. A pair for each output. I know from experience that the crosstalk on these isn't marvellous. Probably OK in this application but it's inherently limited by the "on" resistance and stray capacitance. When I've used this class of device in professional video switching I've always used a double break design, where each crosspoint has 2 switches in series, plus one in shunt which grounds the midpoint when the crosspoint is switched off. Even with relays I'm using this technique to get low crosstalk.
I suppose it's worthwhile running up the Invotron routers to see if they still work. At afirst glance connecting them to latching pushbuttons or rotary switches would need about 40 diodes to encode the switches to binary plus some kind of simple CMOS or TTL sequencer to run round the 4 outputs.
I've probably got some BCD encoded switches to use for test purposes.
I may be doing a disservice to the Invotron video switch design. The main switching is done using 4051 8:1 analogue mux chips but it appears that the video input buffers each have 4 separate outputs, one for each output of the switcher. So it's possible that when a 4051 doesn't select a given source a transistor is biassed off thereby giving a 2nd break in the signal path. I can probably prove this by making some measurements. The circuits won't be easy to trace as the input buffers are on small sub-modules with SM components.
If this is how it works it's slightly reminiscent of the old Cox video crosspoints. When S3 is off it's simply a pair of emitter followers. When S3 is turned on to block the video, both S1 and S2 are biassed off, thus providing 2 breaks in the signal path. S3 also grounds the signal path.
Ah - I remember those. Very elegant!
The circuit design was OK, the mechanical arrangements were not. The crosspoints were mounted on plug-in modules, typically 10 on a module. The individual pins on the modules frequently made bad contact with the sockets on the main board. Also suffered from dry joints on the plug-in module as the pins tried to twist as the modules were plugged/unplugged.
Later Cox stuff used thick film hybrids, one per crosspoint, based on the same circuit. These worked very well.
Many years later I did some video router designs for Quartz using Analogue devices AD8116 16x16 video switch chips. This was real trailing edge tech, just to provide Quartz with a cost competitive range of analogue routers for the last few years of analogue video in broadcasting. Even if I say so myself these were very good products. Most of the design "magic" was in the physical layout and PCB layer structure.
(28-01-2016, 07:45 AM)ppppenguin Wrote: [ -> ]Most of the design "magic" was in the physical layout and PCB layer structure.
Ain't that the truth
In audio power amplifiers, PCB layout can change the distortion by several orders of magnitude. Like, 0.003% vs 3%. And crosstalk is primarily a function of layout. Too many start out assuming that PCB tracks are zero ohms. And when they realise that's not true, they try to make them zero ohms! Doesn't matter how much you "beef up" the tracks; if the layout is wrong, the circuit won't perform correctly. Even ground planes aren't enough in most cases. Eventually people figure out how to "see" where the current flows and use correct noding, but a lot of commercial designs still get it wrong today. The difficulty with audio is the dynamic range - 0.003% is 90dB
This is one of favourite free resources:
http://www.ti.com/sc/docs/apps/msp/journ...aug_09.pdf
We still have some of those Quartz routers at work. When we eventually decide to bin them, I'll think of the museum - but as you've already said, control is the issue. Perhaps now is the time to learn about PICs or Arduinos
I decided on either 6 or 8 layers (it's along time ago) for the Quartz routers. The basic problem is to turn a 3D problem of input and output buses into a 2D one with all the IO on one edge of the PCB. So it wasn't the usual method of a couple of ground planes, a power plane and a couple of signal planes. There wa ground planing on every layer to shield critical bits from other critical bits. I specified a LOT of vias to link the ground planes. My PCB layut guy said he'd never seen so many in his life and it almost broke his CAD system.
There was a fair bit of 75R microstrip tracking. One unexpected problem was track resistance in the microstrips. The design brief had said no preset pots so all gain was determined by precision resistors. WIth a bit of "adjust on design" to give a final trim. Definitely no AOT. Because of differing track lengths we had noticeably different output channel gains. I think I compensated with different "adjust on design" resistors for some channels.
The knottiest gain problem was when Quartz and I were cosnsitently measuring the gain about 0.05dB different. This used up more or less all the design tolerance. A tricky investigation with both our network analysers showed that it was due to poor return loss on one port of their analyser. A 6dB pad sorted it. The analysers do an automated "thru" calibration so the extra loss didn't matter.
I did some PCB design for a small audio company and part of their brief was to have a 0V star configuration with every stages' 0V tracked back to the star point which was a 4mm "bolt hole" as part of the mechanical fixing. It proved a nightmare to implement with so many tracks going from east to west. That had to be a 4 layer board but it was great fun doing it (better than solitaire)! I also did a microprocessor card using a BGA. That had to be 6 layers. I still have the software but I'm sure that it would take me ages to get back into it.
I've had the Invotron video router up and running on the bench. My speculation about the double break in the video path was correct. It looks like each input to a 4051 switch is driven by an emitter follower. When an input isn't in use the video on that pin disappears and is replaced by something close to the +ve rail. I can see how this would work, using a single load resistor on the output of the 4051. However the "on" resistance of the 4051 is quite high so you'd need to use a current source rather than a resistor to get predictable gain. See the sketch.
Now need to see how the audio router interfaces to the video. There's a D15 "slave" socket on each. Need to find how this connects back to the processor card. Won't be using the processor but still need to hook up to its conenctor. This is a 48 pin (half width) DIN41612 socket. The spreadhseet shows where I've got so far.
I've buzzed out the controls to the audio router. No problems.
The video router seems to work OK, subject to some final checks. So now to investigate the audio router. This is stereo (2 separate routers in effect) and balanced IO. Since all the sources and destinations at the Museum are unbalanced I need to find out how the router does its balanced IO. Each input has thick film hybrid with a NE5532 dual opamp. So it's likely to be a true differential input. Hence I can connect unbalanced ground to input ground and -ve, input signal to +ve.
The output is harder. Again it's TF hybrid with a NE5532. I don't think you can do a balanced floating output without more components. Electronic balanced floating outputs are tricky things, with positive feedback to ensure that when you ground one leg the other leg doubles in amplitude. More likely it's balanced around ground but not floating. So you can't really ground one leg of the output. Easy enough to check this with a signal passing through. Just ground one leg and see what happens. Hence I'll be taking an unbalanced output between one leg and ground. This also halves the amplitude of the output so hope the gains presets have enough range to compensate.
The input conenctors are terminal blocks so how best to hook up? In the existing rig all audio is on phonos. So either I bodge phono sockets on to the terminal blocks or, possibly better, make up some phono to wire end leads. At present there are only 4 sources of audio and 4 destinations so it's not a big job.
Input allocation:
1: BBC1 offair
2: ITV off air
3: BBC2 off air
4: BBC1 testcard (video is open, audio to be music one day)
5: ITV testcard (video is open, audio to be music one day)
6: Test card F
7: DVD
8: COW
9: Aux 1
10: Aux 2
Output allocation:
1: BBC1 aurora
2: ITV Aurora
3: BBC2 modulator
4: Preview
An interface to interlocking pushbuttons or rotary switches is pretty simple. Looks like 6 CMOS chips or 3 chips and a bunch of diodes. I'm sure I've got plenty in the stores. The counter simply runs round all 4 outputs and pulses the enables for each output when the data is stable. The 4051/4052 analogue switch could equally be a 1:4 or 1:8 decoder but that forces me to have negative logic* at the encoder matrix. That might work out well, need to sketch it out properly.
*Negative logic: The signal is asserted at logic 0. Common decoder chips like the '138 have outputs like this. Traditionally the enable pins on memories etc have usually been negative logic. Not sure why a tradition grew of using -ve logic. Maybe because old fashioned TTL inputs default to logic 1 if left open circuit.