I have two mark-1 neither of which seems to hear me when i am more than about a meter away or unless i am literally yelling.
Messing with the soundmixer has not helped at all and i just can’t use mycroft this way. Am i really the only one? I would happily solder in a new mic but don’t know much about them and would rather avoid having an external one.
@webshadow - Though our wakeword software is undergoing continuous development, this isn’t typical.
My Mark I units usually respond as long as I am within 10 feet or so, but I 100% believe that there is a problem.
I can think of several possible causes including the volume threshold, electrical interference in the power supply, improperly soldered components, improperly seated wiring harnesses, defective mics, etc.
Lets set up a time with @Darren-Mycroft and a support engineer and figure out what is wrong. We’d like to know so we can fix it remotely in other units ( if it is software ) or identify the problem and resolve it ( if it is hardware ).
Curious - are you using them as a user or are you developing for the platform, cuz our user experience is “Developer Pretty” ( i.e. Ugly ) right now.
I am something in between i guess, been a Linuxnerd for quite some time
but not a coder. Since i already own a Google Home there is absolutely no need
for Mycroft to become polished in a rush. I knew what i was getting into but without
MC hearing me i could not continue …
I would gladly give you guys ssh-access or something like that … Unfortunately remoteshells
are hell, even with a fresh image there is severe lag and i have to reconnect constantly.
All i could see were some connections dropped by “ufw” but disabling it didn’t help.
The usual network based functions work alright like speech recognition or pulling weatherdata.
As far as i remember using a cable does not improve things. If you know about this problem and have a fix or idea i’d like to try them or connect a screen and put my linuxskills to use once again …
@webshadow - @forslund has been working on a new image that might resolve some of these problems.
Ake - any word on a release date?
It’s in testing right now. Should be released end of this week/Monday if nothing unexpected happens.
I’m really looking forward to the update or do i have to reimage my mycroft?
I’d like to give helpful feedback and don’t mind starting fresh.
First release will be image only so you’ll have to reimage the device. We’re looking at automating this, there are quite a number of tiny updates to make the road forward smoother and therefore a lot of tiny things that can go wrong in an automated procedure.
I just want to chime in that I experience the same issue with my Mark I. I need to be within about 4 feet and speaking loud and clear for it to respond to me. Even 10 feet wouldn’t be that much of an improvement, but I understand that this is first prototype product. I enjoy tinkering and modifying things so I have no problem testing any improvements and modifications out. I would even consider swapping out the actual mic if it would bring a large improvement.
I’m another user with this same microphone issue. My Mycroft Mark 1 just can’t seem to hear anything.
I admit he has been sitting on a shelf neglected for many months as nothing seemed to work on arrival. So over the holidays I found a little time to dig in and discovered that the updates had never been working. It seems it was stuck on 0.8.17 which had the update bug. I ssh’d in and did a manual upgrade via apt, but that seemed to totally screw things up.
Rather than waste time trying to work out what mess the upgrade had created, I copied the SD card image (mark-1_production_20171019_build-10) to a new 16GB microSD and booted from that. Still seemed to be deaf, so I left him alone to install any futher updates. That of course lead to him sitting there for hours saying “UPDATING”, apparently hung on something - but I couldn’t see anything in the logs. I used his control knob to initiate a reboot and he again said “UPDATING” for a while, but finally came back to normal - but still defa.
Now getting back to the microphone issue, I can ssh in and run
mycroft-cli-client and the microphone level shown jumps around a bit near the 600 mark, but the movement does not in any way correspond to audible noise. I terminated the various mycroft and pulseaudio processes, and then use
arecord -d 5 test-mic.wav to capture the microphone input. Playing it back with
aplay gave nothing, and copying the file to my PC and viewing it in Audacity shows just a low level noise. I uploaded a small WAV file and was successful in playing that back with aplay (as expected, because I can head Mycroft talk). I tried a second recording with me yelling at Mycroft and tapping on the plastic housing, but still nothing.
Looking inside and comparing thing to the schematic, I see a bunch of missing components (microphone bias?) but they also appear to be missing from the BOM. I guess I can get out the oscilloscope and check if any signal is getting to the preamp and then into the WM8731 codec. Beyond that, I’d assume a software issue of some sort. Any thoughts before I spend more time on this?
Good question @Kean, and excellent troubleshooting information.
Couple of things to confirm first;
- Are you able to tell us which version of
mycroft-core your Mark 1 is now running?
apt-list | grep mycroft should do the trick?
- Have you run a
sudo apt-get upgrade && sudo apt-get update since burning the new image to Micro SD card?
If you’ve gone to all the trouble you have here, and that’s still not working then I think we need to look at a hardware replacement for you - if you email email@example.com and reference this conversation (URL) then we’ll be able to arrange that for you.
Kathy, thanks for the quick response.
pi@mark_1:~ $ apt list | grep mycroft
WARNING: apt does not have a stable CLI interface yet. Use with caution in scripts.
mycroft-core/unknown,now 0.9.13 armhf [installed]
mycroft-mark-1/unknown,now 0.9.13 armhf [installed]
mycroft-picroft/unknown 0.9.13 armhf
mycroft-wifi-setup/unknown,now 0.2.3 armhf [installed]
I thought I’d leave it to do the manual update thing after my earlier experience, but I’ve just done an
sudo apt update followed by
sudo apt upgrade. There was a bunch of stuff upgradeable, including things like
libavcodec* which could be relevant I imagine. Unfortunately during the unpacking of the updates the file system became read-only… So I think another re-image may be necessary, maybe on a new microsd. I’ve got plenty at my disposal, although this was a brand new one. I’ll have to try again later as it is way past bedtime now.
Seeing as I am based in Australia, that you’ve publish hardware docs, and that I’m an electronics and firmware engineer… I am not going to ask for a harware replacement. Instead I do kind of want to have a crack at fixing it myself.
I mainly wanted to see if anyone else had tracked down potential causes of this issue as based on the above I’m apparently not the only one experiencing it.
Ah, a RO filesystem is usually a sign of a failed Micro SD card.
Much respect for your fix-it attitude, kudos.
I’ll be at linux.conf.au in Sydney if you’re going in a couple weeks and want to catch up.
Yes, agreed re the RO FS. I re-ran the upgrade (after a dpkg configure to cleanup after the failure) and it completed, but I may swap out the SD later and start again.
Mycroft still can’t hear me talk, but I can now see a response to me tapping on the case - i.e. I can see jumps from ~620 to ~720 in the mic level while running
mycroft-cli-client. Mind you, just putting my hand near the enclosure without touching it causes the level to go up to ~700 as well.
I was hoping to get along to LCA 2018, but I have way too much work on right now.
Have you run the audio tests?
Hi Kathy - Sorry, what audio tests are those?
So, if you SSH to your Mark 1 and run
sudo mycroft-client-cli then you will be able to see the microphone levels at the bottom right. What do they show?
Not sure which of the two values corresponds to microphone level, but both numbers jump up a bit (max extra 100 units) when I tap loudly on the case. But when talking (or yelling) to him they hardly change other than the existing random noise they typically show.
I did a test on my device and the mic levels I’m seeing are much higher.
5000-7000 when tapping the case and ~1500-2000 when speaking two meters from the device. Might very well be a hardware issue but you might want to run
alsamixer, Press F6 select snd_rpi_proto Press F4 to select input devices and check that the Capture is set to 100. (my value is at 0)
Thanks @forslund - your typical values are good to know.
I’m sitting right next to the unit and there is 10-20 units of noise, and no appreciable difference when talking.
alsamixer as described and the Capture value was 100. I confirmed that the settings were taking effect as I could enable the 20dB gain (Mic Boost) and immediately the mic levels shown increased dramatically (although by a factor of 10, not 100 as expected for 20dB).
So yeah, it certainly points to a hardware issue. I’ll do some inspection and testing soon and report back.
I’ll also note here for any others that come along later that I had to set
TERM=putty (from default of ansi) to get the display and function keys to work in
alsamixer, and that also fixed the slightly scrambled display I had in