A bigger problem

[jkrafft]

Hi there
While playing around with various pedalboards I noticed that my guitar sound sounded a bit muffled but I thought it was some eefects that was a bit dark sounding
I turned off all the effects and it was still muffled.
The sound was normal while the pi-stomp was bypassed.
I made a test an empty pedal board (so you have a cable running directly from the input to the output) and I made an A/B test.
Turns out, there a big difference in the sound.
There's a clear drop in the high starting as low as 3kHz.
I made little sample, starting bypassed, than engaging the pi-stomp and back and forth: https://macserver.fr.to/s/FcwjB2xRKttqWNx
Might by software related but I fear it might by hardware related.
Any clues ?

[Randall (Admin)]

In your recording, I do hear what you're talking about.  I've not measured the frequency response before, but that might be interesting.

So your guitar is plugged directly into the pi-Stomp, and amp right after?  If not, maybe try that to avoid any kind of possible loading issue.

If there is a high frequency cut, I'm near certain it would be a function of the soundcard.  The only thing between the input jack and the soundcard is the unity buffer with a 1MOhm input impedance and the relay.  There are no shunting capacitors so other than parasitic capacitance which is probably way less than 1nF, it should be full bandwidth (like 100kHz) all the way to the A-to-D chip.  There is one series 0.1uF cap going into the buffer and a series 10uF going into the A-to-D chip, but those would roll off lows (calculated and spice simulated to be around 1.6Hz).

I have tried a HiFiBerry DAC+ADC card, and noted that it did sound "better" and had a lower noise floor.  I didn't note anything specifically about high end or tone suck.  But it's bigger and costs $50 as opposed to $18 for the AudioInjector Zero.

The software solution might be to add a graphic or parametric EQ plugin somewhere in your virtual pedalboard to regain some of that brightness/presence.  I know that's not a warm-fuzzy solution, but I've noticed that very few multi-effects pedals have true-bypass, and I believe one reason is because they know that we'll realize that there is some sacrifice in dynamics and tone going from A to D and back again.  My Line6 X3 Live has major tone suck with no effects engaged.  But they know that you're not likely to actually use it with all effects bypassed and you can always boost the global EQ to get back most of what is lost.

I will try to find some time to do a frequency response measurement to see if it can be shown empirically. 

[jkrafft]

Yep, should have given my signal path, so here it is:
Guitar->Mesa Triaxis->Two-notes C.A.B.->pi-stomp->audio interface (Audient ID14)

I've also done a little test, I'll post screenshot when I'll be back home but, using a white noise generator, I've compared the results and it seems a few things are happening:
- There's a low-pass filter at 20kHz (that's expected)
- there's a drop in the lows (around 125Hz)
- there's what would look like a -5dB hi-shelf EQ at around 25.kHz
That doesn't really suggest something electonically related.
Might be a some kind of bug (I mean undocumented funcctionnality) in the sound card implementation.

Right now, I'll compensate with EQ.
Long term fix, if part of the sound card limitations, might be to have an EQ in the jack signal path to avoid having that on MOD.

I'll send aditionnal info as soon as possible.
thx

[jkrafft]

Meant at around 2.5kHz
here's some screenshot.
For lack of better method for now.
First is bypassed, second is engaged with a blank pedalboard (lead going from input to output)

[Randall (Admin)]

Interesting.  Thanks for doing that and posting the results.

Yeah, the 20kHz rolloff is expected for most any audio DAC.  The low dips seem fairly minor, but the roll-off above 2.5k would seem to be noticeable.

It would be interesting to pin down whether software is a factor at all (MOD, jack, alsamixer, etc.).  You might want to check alsamixer ($ alsamixer -c 0) and determine the effect of the Playback Deemphasis and ADC Highpass controls.  From the pi-stomp setup script, I set the default to have Deemphasis enabled (which could definitely roll-off highs) and ADC Highpass disabled (which would roll off lows).

Unfortunately, I've never found good documentation for alsamixer that describes any of the settings to much extent.  It seems like possibly enabling "Output Mixer Line Bypass" and disabling "Output Mixer HiFi", would be the best test for assuring software isn't modifying the signal.  I'm fairly certain the A/D and D/A conversion is still done in that case (as opposed to an analog bypass of the signal which I don't believe is physically possible), but again, without documentation, I'm only guessing.

I will ask Flatmax (creator of AudioInjector) if he knows the expected freq-response or use of low-pass filters in the design. 

[jkrafft]

OK, it seems the "playback deemphasis" is the problem (cf screenshots).
It sounds much better now.
You can still here a very slight tone shift when engaging the pistomp, but it's now very subtle and as you pointed out somewhat expected.

[Randall (Admin)]

Nice, thanks.  It's probably worth considering disabling the deemphasis by default.  I enabled it because some plugins, like the guitarix amp sims, can produce some annoying high frequency artifacts and deemphasis seemed to help.  I didn't realize it was cutting high-mids though and to the extent that you revealed.

I'm glad that at least there is a switch for it.  I'll at least add something to the pi-Stomp wiki about it.

[jkrafft]

Yes, at least now we know what the issue is all about.
Long term update might be usefull to ajust that kind of settings directly on the pistomp.

Side note, "output mixer line bypass" seems to still do the A/D D/A conversion and is fairly transparent (I had a number of pedal with buffered that weren't that good, so I guess the audioinector is doing a pretty good job, at least on guitars).
That might even be usefull if one day you want to implement a software full bypass footswith assignable capability.