OK
Thanks
I will try
Walter
To @LHQuam , @restorer-john , @MarkT and others
You may all recall some time ago when I complained about implementing residuals, a feature I had never used and was reluctant to spend time on.
Well…
I’ve been testing external dac support with the new idle code and Windows repetitive playback isn’t really a thing but there are workarounds. The results were driving me nuts. Keep in mind that the Windows sound buffer size is about 10ms.
I’m getting these awful spectral charts but the sine wave looks perfect. Here the left channel is loopback and the right channel is an external dac.
So, I finally thought, I wonder what the residual shows since the time domain looks perfect. Here’s the residual plot
Note the every-10ms peak which matches exactly with my 100Hz spectral mess. Proving it’s not really a sine wave even though it looks like one and clearly there’s a sound system repeat issue.
So, just a thank you for forcing me to implement this.
Mark
ps. here’s a blowup showing the sine wave error and residual display
2 Likes
@markz This is great with your update now I can measure HiZ 
Could there be a wish list it would be:
1: Current shunt measure for higher loudspeaker measure
2: Marker(s) on the measured curves with readout / diff
3: Better multi curve read in/out, to make compare curve easy
Other wise this software rocks
Thanks.
2 and 3 have been on my list for a while. The “Plot on Top” helps a lot but it would be nice to apply math to each curve. Manual (or peak or bandwidth) markers are very high on my list.
I’ll think about the impedance thing. Unfortunately, the Z math is replicated in a number of areas and changing the impedance math implies more work than it should.
I’ve got a ‘thing’ that I think someone else might have asked about before. I use QA40xPlot for calibrating reel to reel tape recorders. Simplified, everything is adjused one side of 1Khz then post 1Khz. Then there’s biasing which also affects post 10K. So, the funky stuff is all at the upper end. Are there any plans to implement an X-Linear scale to stretch out that end of the frequency spectrum, I predominantly use the Frequency Response graphing more than anything. Love X-Plot, had another look at the original quant software yesterday and it’s just so confusing.
A reviewer asked about this but I just replied it’s not supported since they had no use case. I’ll look and see what would be required to support linear but it’s harder than you think because of all the markers and mouse cursor and …
By the way, there’s a new video (literally yesterday) about running frequency response that you might find useful. See Frequency Response Test · MZachmann/QA40xPlot Wiki
Mark
Ok it was pretty easy to stick a Linear frequency option into the Frequency Response test (only) for testing. I’ll ship it in the next release as a trial (1.4.2) so see what you think of it after trying it out.
1 Like
Mark, you’re an absolute genius. I’ll give it a run through at the trial stage. Don’t know if you’ve had anyone else mention this but on occaision (regular) after changing the Unit of measurement in the frequency generator on the FR page, two odd things happen. If you switch between mV and V for example, the displayed value sticks so if you have a level of 745mV for example, then you change to Volts the display reads 745V (gets a bit scary) I’ll repeat later so I can detail the exact steps where it happens. What’s more strange is that sometimes the gen ouputs just stop generating a signal or go into a distorted square wave. Also had an odd looking triangular sort of affair. If you switch over to the scope page, all is good. Back to the FR page and it all goes weird signal again. I’ll get some screenshots straight from the Rigol scope if they’re helpful. Apologies for not explaining all that well, I’ll try and be a bit more coherant this evening. Thanks again!
Thanks.
Note that the behavior you describe is correct. It used to be the values changed when you changed units to keep the result the same but that is no longer the case. It was impossible to track that effectively when loading configurations (since you load both the unit of measure and the string entry at the same time) so for now they are not synched. Changing units of measure leaves the entered number alone and absolutely changes the value. So be careful!
If you want to track exactly what’s going on note that you can save (or snapshot) a spectrum test and load that result in oscilloscope and vice versa. As long as there’s time data QA40xPlot will load the result into Spectrum and Scope.
Also, the new Idle code has pointed out an issue on startup of oscillators that I’m cleaning up for the next release. As much as this has been painful the results that are being produced are already far better on the edge conditions than the old non-idle code and they rate to improve over the next month or two.
Mark
1 Like
Hi
On plot sw I have this frame
On
github I see this where there is the Impedance test I haven’t SOmething wrong by me?
OOPs!
I found the solution
Walter
1 Like
@thumb re: 1: Current shunt measure for higher loudspeaker measure
We’ve talked about this before but I don’t know what ‘higher loudspeaker measure’ means exactly. Perhaps a little more detail as to why you want this?
Here’s my thoughts. I hate adding complexity if there’s no more value but I’m willing to be convinced. My thinking is that…
-
You are not measuring the current through the shunt resistor. We don’t have a current meter. You are measuring the voltage across the shunt resistor and dividing by the resistance, like the current method.
-
The main difference here is that the std method uses a ground reference instead of two negative legs of differential input. Does that add much? I don’t think so.
-
The circuit you show is complex with 2 differential inputs and to work well requires a clean circuit board with the resistor and connections. A 0.1ohm resistor in series is difficult to get correct and doesn’t have much voltage to sense increasing the noise percentage.
So I did a comparison with a 0.5mH inductor while using my existing jig by connecting the second leg of the right channel to the bottom of my reference resistor and also by swapping the 10 ohm reference with a 0.1 ohm reference (though with interconnects I set it to 0.3 ohms).
Here’s the result->
At lower frequencies:
Full frequency range:
What you see here is that I didn’t get the 0.1ohm resistance right (I had to claim it was 0.3 ohms to get the inductance close and it appears I was slightly off). You also can see that the 0.1ohm shunt method is noticeably more noisy at the higher frequencies (impedances). The Shunt with 10 ohms and the std ratio with 10 ohms are almost identical at all frequencies and I suppose if I had the 0.1 ohm exactly right that would be the same also - or at least close enough.
When you take into account that speaker impedance changes with the current through it and the mounting method - so any measurements of impedance are approximate anyway - I kind of shrug and don’t really want to add more complexity.
Mark
Maybe try this
Without the diode/resistor the QA4xx has 100K/200K input Z and overload protect.
The sens res. should be in the return lead to 0V . to keep it away for the “hot” end .
Why to do it this way is to put some more power into the system/measurement , to see how a speaker/system behave in the box with port load etc.
That is the way Praxis and LEAP is doing it and they set the standart for everything loudspeaker
and I see a 1 Ohm / 5W is what Praxis use. LEAP use a 0.1 Ohm / 5W but they had there own made card, and then a power amp too.
I use your way now and it works fine for quick ,low volt/current, checking a speaker.
To be clear, QA40xPlot absolutely supports using an amplifier, it’s just not the default picture we provide to people. You don’t have to make any changes if the amp doesn’t float since it’s all right-channel input referenced. If the amp does float then just connect the previously-shorted input legs to the nominal ground.
Anyway, I vote we take this offline from here. This thread is so long already. Thank you.
You made a update ! To do ratio/individual nice and it works great, here is a plot with 10ohm DUT
with 0.000434 ohm diff that is awesome, thanks for your work Mark it is great !!
1 Like
When testing with just unbalanced L out to L in and R out to R in why do I get these weird delay values with 1.4.3 version? 1.2.10 shows correct 0ns values.
Also, 1.4.3 shows right channel red trace at a higher level but in 1.2.10 they are completely on top of each other.
And finally: Run/Once/Stop buttons feel very laggy with 1.4.3 while with 1.2.10 they react right away.
There are a bunch of differences with the new IDLE code.
-
STOP takes a few seconds because the generator is running constantly so I wait for it to completely stop before letting you restart it. That can take a few seconds depending on context.
-
The delay in the prior version was constantly calculated based on checking for zero. That didn’t really work in a number of conditions - particularly with chirps that start very slowly. Now it uses a fixed delay (set on the Settings page) that seems to work perfectly with the internal generators. If you’re using an external DAC there is a setting for the external dac delay.
-
In the old code I was removing DC offsets algorithmically which really wasn’t correct when dealing with DC drift caused by the QA40x or the test waveform so now I leave the DC offsets alone unless it’s an external DAC. My two channels are noticeably different DC offset levels.
Mark
Note that this stuff has changed a few times in the last couple of weeks and I expect additional tuning of edge conditions (like the DC offset stuff and speeding up stop) for a month or two. I’m trying to be very conservative to ensure it works before doing minor risky cleanups.
- Maybe make the buttons react right away and then lock them (and visually show that) until the generator stops. Now I was pushing the buttons and wondering what is going on. (I wasn’t in my regular work area and was using a different computer, so I didn’t know what was happening..)
- But why does it show -0.341s and 0.314s instead of 0?
If I had to guess the time delay shown is caused by offset mismatch. But that’s a guess. The time delay algorithm is a bit opaque (it’s based on ffts) but I’ll look into it.
I’ll see what I can do about the Stop button. You’re right it would be better if grayed for a bit.
M
I am having difficulty running QZ40xPlot on my Macbook Pro M4 with Parallels Desktop Windows 11. It does work but the USB connection fails frequently after running one or two tests. Once it fails, your software goes into an endless loop with error message that it cannot connect to the hardware and it will not let me exit the software without invoking Windows Task Manager. The Quant Asylum software runs flawlessly on this Win 11 virtual machine on Mac silicon. Has anyone else reported the problems I am having?
This is the first I’ve heard of this. The QuantAsylum factory software is based on a different framework which is generally more friendly to other operating systems. Have you tried using the REST interface instead of the direct USB connection? Just turn it off in QA40xPlot Settings and ensure the factory software is running to provide the REST interface.
Note I haven’t used/tested the REST interface in quite a while since people generally use the simpler/faster USB interface in Windows.