Stock Board - Voice issues

[Flaps 30]

Having got the Emartee voice module working with Open9x I have found a number of problems (or my stupidity) is lurking around.

Setting up a function switch to play a track when a switch is activated causes the track to be constantly repeated. That continues to cycle ( I have counted six repeats) for some time after the said switch is switched back to 'off'

Related to the above. How do you get a file to be played just once when a switch is activated? Is it possible (as per 9x) to get a track to play say every 15 or 30 seconds if a track is activated?

Other things noted that I have posted about on another thread concerning the Emartee (V1.1 Version sound card. The delays reading out a voltage using Open9x isn't helped by a couple of things.

(1) When reading out a voltage. The decimal point is one sound block, unlike the 9x version where the point and the number after it is one block instead of two with Open9x, hence one part of the additional delay which is considerable when compared to the 9x voice system.

(2) Most of the audio files with Open9x have a section of silence of around 0.2 - 0.35 seconds after the word plus the click. The two that come to mind that have had the said issues have been 0002 (two) and the one that a huge click combined with a long period of silence was file number 0256 (gear up) Both have improved the problem with the delays and clicks. I have edited a few more to prove the case along with the file that said 'point'. Not too sure why the click goes away with the deletion of the silence after the word. But it does work. The editing software used was Audacity.

Personally I prefer the voice used in 9x (Julie in Text Aloud) so I will probably use that after spending some time renumbering the files to suit. But that is a personal preference.

[thomas9x]

Setting up a function switch to play a track when a switch is activated causes the track to be constantly repeated. That continues to cycle ( I have counted six repeats) for some time after the said switch is switched back to 'off'
Related to the above. How do you get a file to be played just once when a switch is activated?

You need to use the momentary Function switches. These are the ones with the little "m" at the end of the function name. Such as GERm, ONm, etc. With these, you will only hear the voice one time.

Is it possible (as per 9x) to get a track to play say every 15 or 30 seconds if a track is activated?

Absolutely! This will require a custom switch. An example of speaking the timer value every 30 secs is posted here:
http://www.rc-cam.com/forum/index.php?/ ... entry26931

(1) When reading out a voltage. The decimal point is one sound block, unlike the 9x version where the point and the number after it is one block instead of two with Open9x, hence one part of the additional delay which is considerable when compared to the 9x voice system.

The underlying problem is that you have the "new" voice module with the unfortunate long delay. I totally agree that adding the decimal point to number's voice file will mask this issue, but this does not fix all the other situations that are goofed up by the new module's extra delay.

A more thorough solution is to use the "old" module. The voice chip batch code list I mentioned in your other discussion can help in sorting out a good module. Or you can wait for MikeB and Oliver to finish their DiY voice module project. Not sure which solution will be the fastest or easiest, so flip a coin.

(2) Most of the audio files with Open9x have a section of silence of around 0.2 - 0.35 seconds after the word plus the click.

I agree that the silence periods at the head/tail of a voice phrase need to be minimal in the voice files. Especially those that are related to speaking a numeric event.

... the one that a huge click combined with a long period of silence was file number 0256 (gear up)

Interesting. I don't hear a click on that file. Is your speaker directly driven by the module, or do you use an LM386 audio amp?
FWIW, speech alerts like Gear Up are standalone (don't require other phrases to complete the alert). So a few hundred milliseconds silence in their files are not going to be noticed. But if removing the silence cured your clicks, then that is awesome.

- Thomas

[Flaps 30]

Thanks for that Thomas. You are a star!

As for file 0256 - Yes my speaker (8 Ohm one inch diameter) is connected directly to the module. Yes the click did go away. No doubt there are many other sound files that could benefit from a trim. Ohh! Yes.. The one that says 'Volts' also had a click that went away with a trim at the end to remove the unwanted silence.

Yes the delay issue could be improved considerably by having files with the point and the number as one file. I do recall that the readout goes to two decimal places. For me that is a bit overboard. I would prefer the improved readout over the extra digit readout.

[thomas9x]

Hmm. Later today I will re-install a V1.1 module (version with the delay) in my 9x to see if I hear the clicks. I'll post the results after the test is performed.

- Thomas

[mhotar]

Download voices-english-stock-r1361.zip voice files available in companion, they are without clicks and silence.
r1361 means version of the tts_en part in open9x, so it is last version of the voice files.

[mhotar]

Yes the delay issue could be improved considerably by having files with the point and the number as one file. I do recall that the readout goes to two decimal places. For me that is a bit overboard. I would prefer the improved readout over the extra digit readout.

It seems to be worth to try, at least for the stock board.

[Flaps 30]

Is it possible (as per 9x) to get a track to play say every 15 or 30 seconds if a track is activated?

Absolutely! This will require a custom switch. An example of speaking the timer value every 30 secs is posted here:
http://www.rc-cam.com/forum/index.php?/ ... entry26931

Looked at that and lost the will to live to be honest. That looked pretty complex for something that should be so simple. Well it is with 9x as it is just one part of what is available as standard (see safety switches) along with playing two files for one switch. Sorry for the comparison, but I do feel that 9x has the edge over Open9x when it comes to voice operations. Maybe that will change in the future.

[thomas9x]

Sorry for the comparison, but I do feel that 9x has the edge over Open9x when it comes to voice operations. Maybe that will change in the future.

It might be best to summarized exactly what you would like and present it to Bertrand. He takes an interest in improving Open9x's user interface.

... along with playing two files for one switch.

Agreed. That is something I would like too.

- Thomas

[Flaps 30]

Download voices-english-stock-r1361.zip voice files available in companion, they are without clicks and silence.
r1361 means version of the tts_en part in open9x, so it is last version of the voice files.

Thanks for that. Yes it does all you say. No clicks audible, along with much shorter (with regards to silences) files. That begs the question as to why the 'english-english-stock.zip' file is there, which is the one I used initially. Of course to me the CSV file is essential to doing any additions/alterations.

Sorry for the comparison, but I do feel that 9x has the edge over Open9x when it comes to voice operations. Maybe that will change in the future.

It might be best to summarized exactly what you would like and present it to Bertrand. He takes an interest in improving Open9x's user interface.

Hmm... Difficult one that. The big boss scares me. Keeping it short. It is having the ability to repeat a voice file that has been selected with a switch or whatever every 15 or so seconds. To do that by using another switch or/and a timer, to me isn't easy as I saw from your link.

The 'both' part is icing on the cake and it saves a software switch and a lot of fiddling.

*edited for clarification

[mhotar]

That begs the question as to why the 'english-english-stock.zip' file is there, which is the one I used. Of course to me the CSV file is essential to doing any additions/alterations.

You are right, 'english-english-stock.zip' is automatically generated file by tts.py script. The second file is handmade one. It needs to be renamed.

[mhotar]

Are you sure the reduction to one decimal point number is not a problem? I do not have FrSky telemetry, so I ask?
I am not also sure the sound number(delay)point_number will sound naturally. I have read somewhere in the forum Jhsa added a silence between point_number to have similar gaps with his module.

[jhsa]

yes I did. to sound more natural.. bit it saves one file on the cue waiting to be played.. maximum number of files on the cue is 16 if I'm not mistaken, so if you get rid of some by putting them together you might have enough space to squeeze one more voice event in the cue if necessary

[Flaps 30]

Are you sure the reduction to one decimal point number is not a problem? I do not have FrSky telemetry, so I ask?

From the standpoint of voice readout of the flight pack. I don't see a problem. It works fine with 9x. That doesn't mean that the display/accuracy of say the vario voltage should be limited to one decimal place.

I am not also sure the sound number(delay)point_number will sound naturally.

Same as above. The readout with 9x is reasonable (not good) with the Emartee V1.1 module. For the reasons already mentioned regarding the decimal point being another block, it is far worse with Open9x and with the additional number being read out, it is all very slow. Bit like pulling teeth out, regarding the total duration. Anything that speeds the whole thing up, and reduces the spaces between the blocks/words is a plus.

[MikeB]

Er9x does NOT limit you to 1 decimal place, just has files for "point zero" to "point nine" to save on the delay. Another digit for two decimal places may be put on the end.

Mike.

[Flaps 30]

Found another little bug.

On switch on the volume level of voice files is at full. You can alter the volume by changing the setting in the radio setup. When adjusted the volume follows the settings that you have set.

Added to my 'It would be nice list' Volume setting when radio inactivity alarm should go to full when inactivity alarm goes off. Repeat voice file (within specified times) and the 'both' is in there.

[MikeB]

If you are having trouble with the "welcome" message, this may be why the volume is wrong.

Mike.

[Flaps 30]

My TX doesn't play the 'welcome' message. I haven't loaded the splash screen, so I presume for that reason it doesn't play the message.

Companion wouldn't allow me to have Audio, Voice and Haptic + splash screens with the stock board. So something had to go, and that was the splash screens.

[mhotar]

I commited change to open9x svn (r1517) to use point_number files and reduced decimal point numbers to only one.
Can someone test it?

160_168voices_en_o9x_r1517.zip

(114.31 KiB) Downloaded 311 times

[thomas9x]

My TX doesn't play the 'welcome' message. I haven't loaded the splash screen, so I presume for that reason it doesn't play the message.

In Open9x, No splash = no welcome message.

Found another little bug. On switch on the volume level of voice files is at full. You can alter the volume by changing the setting in the radio setup.

If this is with Open9X then ensure you are using the latest release. That bug was fixed a couple revisions ago.

Hmm. Later today I will re-install a V1.1 module (version with the delay) in my 9x to see if I hear the clicks. I'll post the results after the test is performed.

I see that flaps30 solved the click issue by using the updated voice file distribution. So no reason for me to do the test I promised to do.

- Thomas

[Flaps 30]

I commited change to open9x svn (r1517) to use point_number files and reduced decimal point numbers to only one.
Can someone test it?

160_168voices_en_o9x_r1517.zip

I would love to try this if I could find the firmware to load up.

In Open9x, No splash = no welcome message.

The question is. Does the volume setting rely upon the splash screen being there to set the initial volume settings? My setup always ignores the volume setting and set the level to full.
Regarding the release that I am using. Companion loads up version 1464 as being the latest for the stock board

Hmm. Later today I will re-install a V1.1 module (version with the delay) in my 9x to see if I hear the clicks. I'll post the results after the test is performed.

I see that flaps30 solved the click issue by using the updated voice file distribution. So no reason for me to do the test I promised to do.

Yes. The voices-english-stock-r1361.zip pack fixes the clicks and gives reduced delays between the words.

[thomas9x]

The question is. Does the volume setting rely upon the splash screen being there to set the initial volume settings? My setup always ignores the volume setting and set the level to full.
Regarding the release that I am using. Companion loads up version 1464 as being the latest for the stock board

The volume setting does not need the splash screen.

I performed a volume test on the 'new" module. On a cold boot, where power has been off for more than one minute, the speaker volume is at maximum. But if power is cycled the volume is correctly set to the menu entry. This issue does not occur with the old module.

The new module requires a few hundred more milliseconds to initialize. So at this point I would blame that characteristic for causing the problem. My gut feeling is that if the V_STARTUP voice state was increased to about 2 seconds the missing volume setting will be restored. But it is just a hunch. I don't have the compiler tools to experiment with the code, so I'm not much help beyond my peanut gallery comments.

BTW, as mentioned before I am collecting sound chip batch codes to help keep track of the old (good) vs. new module (with delay problem). If anyone wants to contribute to the list then that would be great. The list is being published here: http://www.rc-cam.com/forum/index.php?/ ... entry26935

- Thomas

[Flaps 30]

Thanks to everyone for the help. It is appreciated,

If I quickly blip the power switch. I can get the volume to work at the level that is set in the setup. Like others. I am very unhappy with the new module and what they have done with it, but we have to live with it and do whatever is needed to make it work. When I open the TX again. I will get the chip details for you.

Sadly I am no good at all with writing/modifying software. I come from the era when TTL ruled the roost.

[jhsa]

I'm waiting for the outcome of Oliver's sound module.. will build it and if it is better than the new emartee module, and I believe it will be, will definitely dump the other one into one of my son's toys..

[mhotar]

I am also looking forward to Oliver's/Mike's sound module. I received two emartee modules last week

[MikeB]

See viewtopic.php?f=5&t=948&start=1770#p27788.

Mike.

[mhotar]

Good job Mike. Thanks a lot.

[MikeB]

Just commited er9x r777 with voice driver changes.

Mike.

[bertrand35]

Thanks Mike, I just ported them (I kept the 16 sounds list). I see that you did like my code on GVARS Have a look to the commits I did later, I fixed some bugs there!
Bertrand.

[MikeB]

Yes to the GVARS, I'm also using them to get the trims and rotary encoder availabe as mixer inputs, still some work to do before I "turn them on".

Mike.

[bertrand35]

Yes I have seen, but for the moment I don't like too much, I prefer having the ThrT, AilT, RudT, EleT and the RE available in the list while the Global Vars are used to replace weight / expo / offset / differential ... : only fields. My opinion today, but perhaps I will change my mind later, depending on the use cases !

Bertrand.