Page 1 of 1
Posted: 11 Apr 2019, 08:37
I'm very pleased to find this nice and clever computer design. And well documented too.
In this regard I found in the schematic you're using germanium diodes for logic (at least for the AND gate, in decoding ROMs the diode model is not specified). This looks odd, and in fact, the videos and photos show diodes in the style of the common 1N4148, that, BTW, is perfectly ok for this use.
Please, confirm the diode model used in the kit.
Also I was a little worried about the delay of this diode-resistor logic, but I found it was no more than 20ns, more or less in the same order of the logic chips.
Posted: 11 Apr 2019, 11:23
In the design all signaling diodes are the same. We recommend 1N60P or BAT42 compatible diodes, especially when using 74LS logic. Dave Jones has used the more standard 1N4148 diodes for his 74HCT-based build, and that combination seems to work for him.
The consideration is voltage drop versus the noise margin in the logical low region (≤0.8V). Voltage drop for diodes depend on the current (lower is better in our case) and temperature (higher is better). [Edited: Speed is mostly determined by the pull-up resistors. L-to-H transitions are slow. But we get a little helping hand from the TTL levels here: ≥2.0V is sufficient to become H.]
P.S: I don't believe that 1N60P is a true germanium diode, despite what some datasheets suggest.
Posted: 15 Apr 2019, 10:25
So they are Schottky barrier diodes. This makes more sense, because new germanium diodes are difficult to find (not in Digikey, nor Mouser...)
Just for comment: You can save one diode in the AND gate. Just remove one diode and place the resistor in ts place instead. This two component AND gate will have a longer H-to-L delay in its resistor input, but it will probably work in the Gigatron...
Posted: 29 Apr 2019, 14:32
I ran some simulations of the diode AND gates and I was surprised to find how fast these gates are.
This is the circuit I simulated. The 45ohm resistors takes into account the output resistance of the preceding logic (from an HCT138 datasheet), and the 10pF capacitor is an estimate of the gate loading (two 5pF, HCT, inputs).
And this is the simulated waveform, along with that of an 74HC8 gate:
It's amazing to see how the diode gate has almost no delay (the HC08 delay is 13ns). Yes, there is a long transient for the L-to-H edge, but the voltage is already high enough to represent a logic high. This fast response is due to the parasitic capacitance of the diodes.
The simulation of a 5-input diode AND gate gives a worse transient, but the digital delay is still good:
I was also considering a simplified diode gate in order to save a few diodes from the part list. In this gate the pull-up resistor replaces one diode:
But the simulation shows this is really a bad idea. it shows glitches and delays up to 52ns long.
Maybe these simulations explain, at least to some extent, "How to beat the datasheets"