What about the AY-3-8910 at U1?

What about the capacitors and the audio amp IC?
What you are describing sounds like one of the pots on the soundboard is bad. (has a dead spot)

Try cleaning them. And if you get a bit better sound. Or more sounds replace them.

Hate to say it but you tried a shotgun repair and replaced a lot of components on the sound board that I rarely see fail. The soundboard does all the sounds in the game, the vocalizer PCB does all the voices. I would agree with replacing the AY-3-8910, as that is the component I see fail on these boards a LOT! If you don't have an easy way to get one, replace ALL ELECTROLYTIC CAPS on THE ENTIRE BOARD FIRST. I have fixed a dozen of them by just replacing all the caps. Think of replacing sound board caps as routine maintenance with Bally machines.

I don't suggest shotgun repairs but if you are going to go that route, at least replace the components that MOST often fail, FIRST.

I completely agree with that approach. Electronic chips will rarely or ever fail if correctly powered. All chemical capacitors MUST be changed first. Then I would replace power resistors if any, and all carbon based resistors with metal oxide resistors.

Yves

My bet is it's the sound ROM, I think U4. If some of the sounds generated are correct and clear, then its not any of the other stuff mentioned.

Xenon sound is a little bit different from all other Bally sound systems. Xenon uses multiple ROM for recording each of the words. I seem to recall that there should be 7 ROM with a different speech in each of them. You can verify that they work by pushing the test button.

Then Xenon, has a regular sound system producing sinewaves and other artificial sounds. It uses a different amplifier, different volume control and some unit had two speakers (one for the voice and one for the sounds). I suspect that in the case of this thread, the sound amplifier or the volume control or the speaker of the sound section is toasted.

Yves

Mine has two speakers, but they are connected in parallel, not separate.

I assume you have set self-test positions and/or dip switches to generate electronic sounds, not simulated chime noises...

https://pinside.com/pinball/forum/topic/xenon-vocalizer-eprom-repair-vids-guide

Sounds like you're getting some proper FM synthesis sounds from the AY-3-8910 but not all.
Have you checked with a logic probe if all the outputs of the 4049 (U5) that address the AY-3-8910 are ok?

The AY-3-8910 are used on lots of old arcade video games. That's a source if there's a market for old game boards in your area.

Otherwise, if you're in no rush, AliExpress might be your friend - here's the cheapest ones with pictures:
http://www.aliexpress.com/item/AY-3-8910-DIP40/32596325031.html
http://www.aliexpress.com/item/AY-3-8910-AY-3-8910A-dual-in-line-IC-DIP/32635016340.html
GI was the original brand, Microchip bought them later, though the 2008 date code on the Microchip shown looks a little dubious.
Do your own search there, you'll probably get different shipping prices to me.

Sorry for the long post but here goes.. Getting technical, this is how I read the schematics:

Looking at the datasheet of the AY-3-8910, it has two general purpose 8-bit I/O ports, and 3 independent analog output channels.

One of the AY-3-8910 general purpose 8-bit I/O ports is not used. The second 8-bit I/O port is configured as an 8-bit input port.
The MPU board connects to this 8-bit input port (actually only 5-bits are active) and uses it to issue commands (sounds to play) to the sound board. Consider this like a mailbox that the MPU board posts sound commands to the sound board.

The 6808 CPU on the sound board (via the 6821 PIA) will check the AY-3-8910 mailbox on a regular basis to see if any commands (sounds) need to be processed.

If a command comes in to play a vocal, the sound board CPU will then point to a pre-programmed location in the Vocalizer ROMs relevant to that vocal and read the data in successive ROM locations and produce a single digital bit stream on the output of the 6821 PB7 pin and the rate is controlled by the 6821 PB6 pin. These 6821 output signals feed the MC3417 speech decoder on the vocalizer board that return an analog signal back to the sound board at the Speech Control Pot and onto the final output amplifier.

Now onto the AY-3-8910. If the MPU board issues a command to playback a synthesized tone, the sound board CPU will read pre-programmed config data for this sound from the U4 EPROM and send them to the AY-3-8910 to playback on either/any/all of the 3 analog output channels of the AY-3-8910.
These 3 analog outputs are all directly connected together externally on the PCB to allow mixing and production of complex sounds.

So... from the short video clip you posted, you are getting communications between the MPU board through the AY-3-8910 then the 6821 then onto the 6802/6808 CPU. It could still be possible you have an open/stuck bit somewhere in the chain.
You also seem to be getting some of the right synthesized sounds but not all.

I would remove the AY-3-8910, and carefully bend the 3 output pins (pins 3, 4 and 38) up just enough so you can re-install the chip with those pins disconnected. Don't install the chip all the way in the socket because you could stress those 3 slightly bent out pins, just insert it enough to get all the unbent pins to touch the socket.
Start the machine. You should still get all vocal sound playback, but none of the synthesized tones.
Connect one of the 3 now floating outputs of the AY-3-8910 to TP6 (the pre-amp input) on the sound board with wire-clips. You should get some tones, maybe not the frequency you expect but it shouldn't be distorted noise like you're sometimes getting. Do the same with the other two output pins of the AY-3-8910
I would expect one of these output channel pins to be faulty.

The 6821 can also apply some controls to the mixing on the pre-amp connected to the AY-3-8910 output, one seems to be a volume reduction controlled by 6821 CA2 output, the other (6821 CB2 output) seems to control some sort of filter. I wouldn't expect the volume reduction to be an issue, the filter possibly if the caps are way out of spec.

Hopefully this is thereabouts on how it functions and makes some sense to help you narrow it down.

Yepper; I dealt with this problem on Williams sound speech boards.
The problem is your MC3417L.
There are 2 versions of these. (Unfortunately).
It took me about the same amount of time to figure out what the heck was going on.
Speech section from a Black Knight pin would not reproduce speech.
I put in Firepower speech EPROMs and voila............ speech like there's no tomorrow.
Speech IC4 on the speech board for FirePower is unused, but populated for Black Knight.
So, I had a bunch of MC3417L chips and started swapping these until all of a sudden Black Knight started talking.
What causes it, I don't know. Must be something internal of the 3417.
But I bought some extra of these chips and filtered out the ones that only work for speech in Black Knight.
I recommend you do he same, buy a bunch of these (EBay) and find those who make all your sounds work with ALL eprom locations populated.
The sound itself has nothing to do with it; it is the amount of (speech) EPROMs used in a game.