Sign in to follow this  
Followers 0
ThoughtPhreaker

TOPS/ACTS recording

3 posts in this topic

This might be the last time I get to hear a US West TOPS switch hassling me for money, so I thought I might record it. I didn't have a pickup coil with me at the time - still don't actually, I should probably find my way to one. But anyway, sorry about the automatic gain control. Next time I do this, I'm going to use something a little cleaner. All I had at the time was my Dialogic box, though.

 

In case you were wondering, this switch is indeed the sort of thing you can redbox, but it typically doesn't ask you for money retroactively. It's doing this (it actually never cut me off if you're wondering; I sat there for like twenty minutes. The tops_2.wav stuff is the last thing it said) because Qwest doesn't use TOPS for operator services anymore. It's not programmed to automatically cut you off and there's no person it can call to intervene, so, well, it just lets the call go on forever. And probably raised an alarm on the console. I've never heard it myself, but the TOPS manual says it can actually get pretty aggressive; it'll call you back to try and get you to pay if you let it. I was really disappointed when it didn't. If you listen to the way it says "past", you can hear this subtle looping sound on the end of the T syllable. This is a characteristic thing the Nortel EDRAM card does - the closest we'll get to proof here that the tandem is a DMS.

 

Funny enough, we actually do have the original files the switch is playing back; it's some form of 32k ADPCM. It's all in some sort of strange container format that nobody could ever figure out, though. If you'd like to try your luck with it though, this is the archive with all the stock EDRAM stuff. eacts0ae.bin44 has all the ACTS stuff in it: http://www71.zippyshare.com/v/1XzPMAeZ/file.html . I'll post a manual for the card at some point. The .bin44 extension implies that it's binary as per usual, but the 44 after indicates the logical record length of the file is, well, 44 bytes.

 

 

tops_1.wav

tops_2.wav

2

Share this post


Link to post
Share on other sites

ThoughtPhreaker, is there any chance I could get those TOPS ACTS and other audio files? I'd like a shot at recovering the audio.

 

The link here is liong dead.

 

Thanks,

 

Don

0

Share this post


Link to post
Share on other sites
On 4/14/2020 at 11:21 AM, df99 said:

ThoughtPhreaker, is there any chance I could get those TOPS ACTS and other audio files? I'd like a shot at recovering the audio.

 

The link here is liong dead.

 

Thanks,

 

Don

I finally got a copy of the audio files (EDRM0001.SLM Part 1 and EDRM0001.SLM Part 2 from a friend on the NPSTN board. I think they were probably from Thoughtphreaker.

 

My friend was interested in recovering authentic ACTS audio from the TOPS system. There were a few documents with the zipped files that indicated that the files were for loading audio data to the non-volatile memory of an NT1X75 recording announcement board for a Nortel DMS 100 switch. 

 

After perusing the documents, I extracted a single file from the zipped archives, eacts0ae.bin44. This is a binary file designed to contain multiple sounds and information to load into the proper non-volatile address space on the NT1X75 board.

 

The file contains multiple audio phrases. The file is organized into 44 byte records (or "lines", if the file was ASCII text). There is what appears to be a 6 byte header at the start of each line. The first two bytes are always hex 47 and hex 32. That is followed by the number of valid data bytes in the record (36 bytes for a full record). That is followed by a hex 00, followed by two additional hex bytes that probably specify the 16-bit board memory address into which to store the record. There is also a 2-byte trailer at the end of each record. I believe the second to last byte is a checksum and the last byte is always hex 16.

The boundaries between each audio phrase can be discovered by looking for a record with a length less than 36 bytes. That is usually the end of the audio phrase. The next record starts a new phrase. Records with less than 36 bytes of data are padded with zeroes to make up 36 bytes in the data portion.

To recover the audio, I simply chopped the first 6 bytes and last two bytes off of each record and saved the file as a binary, using a hex editor with a column selection editing mode. This leaves other extraneous data present that plays as noise, but that data now is before/after each phrase, not interleaved with the phrase data samples. I used a hex editor with a column select mode to do this, omitting the first 6 bytes and last 2 bytes of the selection, then copying and pasting to a new hex file and saving as a new binary file.

I then ran the raw data through various telephony codecs using Audacity and a program called GoldWave, designed for converting and manipulating various telephony audio storage formats. I found a manual for the audio DMS100 card that states the files are ADPCM, 4 bit. I found some settings that decode correctly. The problem is clipping of the audio samples. Many Oki and Dialogic ADPCM decoders allude to the fact that the decoders produce clipped audio. I’ve tried four different decoders, including SOX, and the output is all clipped. Not sure of the fix for that. Running the audio output through a low-pass filter helps the legibility some. The problem is that peaks of the recovered audio are missing, causing severe distortion.

Intelligible results were also obtained playing the audio back as straight uncompressed 8-bit A-law PCM at a sample rate of 8KHz. However, the audio also clips severely when importing into that raw format in Audacity.

 

Still working on it. It could also be some custom ADPCM CODEC.


Attached is an mp3 with the entire file played back as A-lew, 8-bit PCM at 8 KHz in Audacity. I'm also including the converted binary file with the extraneous record data removed, as well as the original .bin44 file. Import as raw a-law at 8 KHz in Audacity. Similar results on output audio were obtained decoding as ADPCM in Oki format at 8 KHz sampling using SOX and the VOX/DEVOX program, compiled from source code.

I could probably separate out each phrase from the surrounding junk without much difficulty.

 

Regards,

 

Don F.

eacts0ae.bin44test4.bin

eacts0ae_4.mp3

eacts0ae.bin44

TOPS_DMS100_ACTS_ROM.pdf

0

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!


Register a new account

Sign in

Already have an account? Sign in here.


Sign In Now
Sign in to follow this  
Followers 0