Vox Probe 15

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

Andrew Davie

**VoxProbe 15: one more AtariVox recording, please**

Attached is **VoxProbe 15** (`vox_probe15.bin`, 4K, build 1, sha1 d0fb98b0). It's a test ROM that makes an AtariVox play a fixed list of sounds, so the SpeakJet emulation can be checked against the real chip.

**Why this one.** The emulation is scored against four recordings of every sound, two from each of two AtariVoxes. Checking that score showed that one of them, the oldest recording from my AtariVox+, has noticeably more background noise than the other three, and it skews every score. This probe records every sound again so that take can be replaced. It also carries three tests the emulation still gets wrong:

- when a Pitch command really takes effect partway through a phrase (the chip seems to apply it one to five sounds late);
- how a vowel starts after a short pause;
- how loud TT and TH are next to a vowel.

**What you need**
- An Atari 2600 or 7800 and a way to run a ROM (Harmony, PlusCart or similar)
- An AtariVox or AtariVox+ in the **right** controller port
- A joystick in the **left** port (or use RESET instead of the fire button)
- A way to record the AtariVox's audio output

**Recording**
1. If you can, connect the AtariVox's audio output straight to the recorder's line input. Otherwise, a microphone in a quiet room in front of the speaker will do.
2. Record WAV at 44.1 or 48 kHz, 16 bit or better. Turn off noise reduction, automatic gain, compression and normalising. Set the level so the loudest sounds peak well under full scale.
3. Start recording, then load and run the ROM. The title screen reads **VOXPROBE 15 B1** and the screen turns **blue**.
4. Wait a couple of seconds, then press **FIRE** (or RESET). The screen turns **red** while the sequence is sent.
5. When it turns **green**, keep recording for **another 10 seconds**: the AtariVox still has speech in its buffer.

The whole run takes about **5 minutes**. Please send the WAV as it is, with no editing or trimming, and say which AtariVox it came from (original or AtariVox+). A second take is very welcome.

**What it plays**
- A reference tone, then all 72 allophones one at a time
- All the sound effects
- A run of AW sounds with pitch changes, short pauses and slowed vowels
- TT and TH next to vowels
- The reference tone again, then the 12 DTMF tones last

Thanks. Every recording so far has turned into a measurable fix in the emulation. The current release (0.5.3) has all 72 allophones and every sound effect measured against two real AtariVoxes.


SpiceWare


Andrew Davie

TY for the recordings.


**AtariVox speed depends on how it's recorded (USB power suspected)**

While checking the latest recordings (VoxProbe 15) I found that the same AtariVox doesn't always run at the same speed.

The test is simple: at the default Pitch (88), every voiced sound repeats at a fixed period, set by the chip's clock. Measured over steady vowels, to a thousandth of a millisecond:

unit                    recording                          period     vs console
AtariVox+ (unit A)      on the 2600, every session         11.437 ms  -
AtariVox+ (unit A)      VoxProbe 7, Man Goes Down (older)  11.414     0.2 % fast
AtariVox+ (unit A)      driven from Stella (VoxProbe 15)   11.386     0.45 % fast
original AtariVox (B)   on the 2600 (VoxProbe 15)          11.478
original AtariVox (B)   driven from Stella (VoxProbe 15)   11.474     same

On the 2600, the console sessions agree with each other to within 0.1 %, across recordings weeks apart. Unit A's take made through Stella runs 0.45 % fast. Its 2600 take, recorded minutes earlier, matches the console figure exactly. Unit B is not affected: its Stella take runs at console speed. The Stella takes of both units do carry 1.5 to 3 dB more above 8 kHz, which is the recording path, not the chip.

**What I suspect:** the AtariVox+'s supply. When it's driven from Stella it hangs off a USB serial adaptor, and an earlier session's notes say unit A was sometimes powered over USB. Unit B never was, and unit B doesn't drift. If the SpeakJet runs from its own internal oscillator, a different supply voltage would move its clock, and with it the pitch, every sound's length and the timing. The two older unit-A recordings at 0.2 % fast could be the same effect, smaller. This is a suspicion only: the recordings can't tell a fast chip from a recorder clock that runs slow.

**Why it matters here:** the emulation is fitted to recordings, so a recording made at the wrong speed pulls the fit away from what a 2600 owner hears. Scored against the other recordings, the Stella take stood out clearly (about 1.5 times as far from them as they are from each other), so it isn't used. No Stella take is in the fitting set. The two older unit-A takes are being checked the same way.

**From now on:** recordings for the emulation are made on a real 2600, with the AtariVox powered the normal way through the joystick port, not over USB.

**If anyone wants to help settle it:** record an AtariVox+ driven from Stella (or any PC) twice: once powered over USB, once from the console or a bench supply. Have a steady reference tone, such as a phone playing 1 kHz, on the recorder's other channel. If the tone reads true and the speech period moves with the supply, it's the chip. If the tone moves too, it's the recorder. Either way it would be good to know how far a USB-powered AtariVox+ drifts from a console-powered one. Anyone using one with an emulator is hearing that version.


SpiceWare

I reviewed the discussions and Vox Probe 10 appears to be the first time I captured using Stella and I wrote:

QuoteThe capture device is plugged into a monitor while the AtariVox Interface is plugged into a HUB, so there might be some issues that would be better if they were both plugged into the same place.


Starting with Vox Probe 11:

QuoteThis time for the Stella recordings the AtariVox Interface and the Audio Capture device are both plugged into the monitor.  My thinking is if they're powered by the same source it might reduce noise in the recording.

I left the setup the same after that so only the initial recording of the AtariVox+ had it powered by the USB Hub, for everything since it's been powered by the monitor.

Possibly both on the monitor is at the limit of what the monitor can supply.  I'll do a new capture of Vox Probe 15 with the AtariVox Interfaced moved back to the USB HUB so we can see if that makes a difference.

It will be after work before I can do it as the HUB is plugged into a USB Switch because it's used to share my keyboard & trackball between my work PC and Mac.


Andrew Davie

Quote from: SpiceWare on 17 Nov 2026, 03:22 AMPossibly both on the monitor is at the limit of what the monitor can supply.  I'll do a new capture of Vox Probe 15 with the AtariVox Interfaced moved back to the USB HUB so we can see if that makes a difference.

TY but let's hold off on that.  I'm not really sure what the best approach is yet.