infiniteneslives wrote:Neither the 8910 nor the 2149 is self sufficient, but they're each lacking different things.
AY-3-8910: You must add a clock divider since it doesn't have one built in. However it's output doesn't require an amplifier.
YM2149: Has built in clock divider, but has a weak output so you have to add an amplifier.
Both require the same address decoder.
Yep, but this new YM chip seems to use conventional I/O logic instead of that INTL specific control pins the AY-3-8910 has.
The amp on the original cart is super simple and easy to dupe. Just one inverter port of a 74HCU04. Maybe if he uses a 74HCU00 he could use one of the NAND ports with the inputs tied, as a inverter ? That would be cheap enough, no ?
I've made a design with just a single 74HCU04, two diodes and some resistors, but I haven't tested it yet. The 74'04 does both decoding and amplification/buffering.
I think the logic is wrong? He's got R/W -> /WR, that's fine, but he's also got NAND2(/ROMSEL,A14)->/CE, which I think should decode on writes to $4000-$7FFF, not $C000-$FFFF.
Fortunately, he's got plenty of free NAND gates to re-invert /ROMSEL.
Also, as (I think) everyone else said, you don't want the divide-by-2 on the YMZ parts.
I guess I'm "lucky"... One of the two carts I bought already contained a Sunsoft 5B (Dodge Danpei, the first one). But I will try "my method" on the other cart.
I find it weird that the 5B has a datecode of week 6 1990 (datecodes are in WWYY w=week y=year format), when the first and only game that ever benefited from it was released in early 1992. Did it take almost 2 years to develop?
Here's my plan, please feel free to laugh at my drawing skills.
lidnariq wrote:I think the logic is wrong? He's got R/W -> /WR, that's fine, but he's also got NAND2(/ROMSEL,A14)->/CE, which I think should decode on writes to $4000-$7FFF, not $C000-$FFFF.
Fortunately, he's got plenty of free NAND gates to re-invert /ROMSEL.
Also, as (I think) everyone else said, you don't want the divide-by-2 on the YMZ parts.
yxkalle lidnariq
How about this fix (attached)?
---
Why I don't want the divide-by-2 on the YMZ parts? I don't understand.
The YMZ parts have the divide-by-2 already included inside.
If you compare the datasheet for the AY-3-8910 for the datasheet of the YMZ284, the former says "the frequency of each square wave [...] is obtained by [dividing] the input clock by 16 , then [dividing] by the programmed 12-bit Tone Period value." The YMZ284 datasheet instead says "F_t = F_master / 32·TP".
I think this revised plan looks ok.
Yamaha's clones of AY-3-8910 contain built-in source clock divider by 2. It is optional in YM2149, controlled with a pin; but seems to be always enabled for YMZ284.
yxkalle wrote:
Here's my plan, please feel free to laugh at my drawing skills.
20131007_233347.jpg
What is the point of inverting a line twice ? Just connect it straight and save 4 soldering spots. The best way to go is plan the circuit in a way you have to solder the less pins/wires possible.
aodinets wrote:
yxkalle lidnariq
How about this fix (attached)?
---
Why I don't want the divide-by-2 on the YMZ parts? I don't understand.
Yes, that's exactly like how I would do it myself using 74xx00's.
l_oliveira wrote:What is the point of inverting a line twice ? Just connect it straight and save 4 soldering spots. The best way to go is plan the circuit in a way you have to solder the less pins/wires possible.
There's probably no point of doing so, I just wanted to be sure that the voltages where good before reaching the diode (low being close to 0V and high being close to +5V). I'm 99% sure it would work just fine without doing so, but I like the number 100 better.
yxkalle wrote:I find it weird that the 5B has a datecode of week 6 1990 (datecodes are in WWYY w=week y=year format), when the first and only game that ever benefited from it was released in early 1992. Did it take almost 2 years to develop?
2 years is a reasonable development time. Also it's extremely common for game projects to be cancelled before completion, so this might not have even been the first game they were planning to use it on.
l_oliveira wrote:What is the point of inverting a line twice ? Just connect it straight and save 4 soldering spots. The best way to go is plan the circuit in a way you have to solder the less pins/wires possible.
Diode logic sketches me out when it comes to loading signal lines. I'd prefer to put the two inverters there just to avoid loading CPU A14.
Actually, I'd probably prefer to use a 74hc00 and figure out some other amplifier topology, but I don't have any of the YM audio parts to play with.
l_oliveira wrote:What is the point of inverting a line twice ? Just connect it straight and save 4 soldering spots. The best way to go is plan the circuit in a way you have to solder the less pins/wires possible.
Diode logic sketches me out when it comes to loading signal lines. I'd prefer to put the two inverters there just to avoid loading CPU A14.
Actually, I'd probably prefer to use a 74hc00 and figure out some other amplifier topology, but I don't have any of the YM audio parts to play with.
I would be very conservative about using "mickey mouse logic" (but hey in the 60's the computers were built like this, no ? DTL perhaps ?) and now that you exposed the concern about putting electrical load on the address lines, I kind of agree with your point of view.
I still believe that using what Kevtris used on his design was the most elegant solution. But hey I was already using 74LS139s and 138s for address decoding logic way before I ever met Kevtris. lol