Hey all..
Starting using my 403 in actual work today as opposed to just doing loop back tests learning the software. I’m having loads of issues. This post is regarding the oscillator outputs pulsing on and off. I am developing an audio compressor and need a steady output and my 403 is always pulsing on and off. Tried everything but still does it. The rate depends on the sample rate and the FFT settings. Tried several computers and reinstalled software still same issue. I can’t use this is the oscillator always does this…
Thanks so much for your help and advise..
Cheers
Skip..
Hi. During FFT acquisition, the behavior you describe is normal. If you want to have continuous output from the DAC, you must press the “IDLE” button. The output is continuous with the button pressed, provided the analyzer (FFT acquisition) is not running..
Ok but how we get thd and other data? That stops as well. I need to have that data as I adjust distortion nulls and so forth…..
Thanks a million
I think you will have to use an external oscillator (not the QA403’s DAC) because there is no way to alter the operation of the QA403’s DAC when acquiring an FFT. I at least do this when using the 384KHz sampling rate, a frequency at which the DAC of the QA403 does not work
Well that’s a bummer…
I will not be able to use this unit for my work…
I must be able to have a constant osculatory and get audio data. I would have never imagined this would happen… I sure hope this can be fixed as it renders this product useless for a load of applications…
Maby in the future..
Matt, what do you think??
Well that’s a possibility…
Would you kindly be able to provide a download link for that driver??
Cheers..
I already edited my post with links
Thanks a bunch..
I sure hope this can be rectified.. makes the otherwise nice unit somewhat useless for soooo many applications…
I’ll look at REW. I have my doubts though..
Cheers and thanks a bunch….
Skip
Hey all..
I checked out REW and it will not work either. Unfortunately the out level is fixed as well as the attenuation. You must go change a file every time you want different output and in levels standards. I am soooo bummed that this is the only way this system will work. I wonder why the heck would it be this way? A pulsing oscillator is not good for any sort of measurement I can think of. It creates all kinds of problems in analytics of the circuit. I sent an email to support and if this can’t be rectified I will need to dump this unit. Not sure how but it’s a real shame as the unit has potential.
Thanks here for all your advice. Really appreciate it…
Cheers
Skip
You can change the length of the shaped burst the QA403 uses by increasing the FFT size.
That is nearly 22 seconds for a 1024k FFT @48kHz.
Yeah thanks…
Might help a tad. When doing a distortion null on consoles vca’s and compressors you typically send a signal at different levels, adjust threshold for specific amounts of compression then keep adjusting the null control to minimize thd. This process some takes several minutes per channel. My absolute surprise is why the heck would this pulsing output even be part of the design. I really can’t figure this strange phenomenon. I wonder if the dsp in the box is so limited that you need to pulse it between fft analysis? Just doesn’t seem right.. I really hope Matt can change this. The other thing is the pulse really will change the analysis in a lot of circuits. Not a good thing IMHO.
Thanks for the advise
Cheers
Skip
I replied in a thread I was linking for you…
Hi @skipburrows, the design currently works on packets so that there is always a fixed phase relationship between output and input.
For making measurements with FFTs in general, your amplitude needs to be constant for the measurement to make sense.
For the compressor measurements you need to make, what is the attack time of the compressor? If it’s 100 mS, then you can increase the latency compensation to a number greater than 0.1 seconds. This will add some extra tone duration at the front of the burst, allowing your compressor attack to do its thing and stabilize, and then the FFT will be made after stabilization. And you can just start running, and tune whatever needs to be tuned based on the THD updates you are seeing on the screen. The thread @restorer-john linked to should give some more detail
Once the application is moved to Avalonia, there will be a button on the front that lets you set whether the entire system is running in continuous mode (the mode you want, like a normal soundcard with no guarantee between in/out phase) and bursted mode (like the QA40x is today, with bit-exact timing aka fixed phase relationship between input and output). So, what you want will come, but probably not in 2025. Or if it does, it will be missing certain things (like visualizers) and then those will be rolled in during 2026.
But I really think you can make the measurements you want today no problem. You just need to bump up the latency comp based on the attack time.
PS. You can measure the compressor attack time you are seeing using some of the plug-ins.
Hey Matt, thanks for the reply..
As I do design and in field calibration that absolutely requires a steady state output. I service SSL consoles on a regular basis. So on a VCA systems you must send a calibrated level at varying amounts and do different null adjustments to minimize THD distortions. You mentioned a different development package. Perhaps then this product will become a good replacement for folks like me lugging my Audio precision all over the world. However until then, I’m afraid I won’t be able to use this product. Please don’t take any offense to this. As they say it is what it is. My only complaint is I wish this was made extremely clear in the product description. If it were I would not have purchased it. Not sure how I can use it now but perhaps in the future….
I do appreciate your involvement in the discussion…
Let me know if you need any beta help… I’d be happy to lend a hand.
Cheers
Skip
1 Like
Hi @skipburrows,
I’m not clear why you think you can’t “send the calibrated levels at varying amounts” from the QA403? You have an attack time, and after the attack time has elapsed, you have a steady-state signal–the compressor isn’t doing any more modifications to the signal at that point. All you need to know is the attack time, which can be readily measured.
Below is a plot from a compressor output from the other thread. We can see here that the compressor is actively doing its job for the first 20-30 mS, but after this time, the compressor doesn’t care if the applied signal is a 40 mS burst or has been playing for an hour–the output is the same. From this plot, we can estimate this particular compressor plot shows a 10mS attack time and 3:1 compression ratio.
But the important point below is the FFT (and thus the THD, etc) is happening on the region inside the box. And as you can see, in that region the amplitude is constant.
Can you grab a plot like this from the SSL VCA when the front-panel setup is in the config they recommend for calibration? Looking at the cal procedure for the SSL 82S6XL010D, they want the front panel set as follows:
"It is also assumed that unless otherwise specified, all switches are released and all potentiometers are at unity, minimum or indent position as appropriate. The required accuracy for each adjustment will be specified along with the target value.
Those settings should yield a 1mS attack time in cal mode. Which means after a mS or two, the signal is at steady state. And thus, the burst (versus continuous) shouldn’t be an issue.
Hi Matt.. attached is a sample for part of the calibration of a Daking FETIII.. this would be extremely difficult with the current 403 design
I do appreciate your help in explaining some other details of the 403 use…
Cheers
Hi @skipburrows, thanks for the procedure! I think you can readily test this. I’ll be a bit verbose in places below, but let me know if there are steps for which you need more detail.
Analyzer Setup
- File->New to put the QA40x software in a known state
- Set to dBu mode (right click dBV button in AXIS control group, and select dBu in Y Axis). And also click the +6.02 dB output gain button because we’re going to run in balanced mode
- Set to 16k FFT
- Right click on RMS in MEASUREMENTS control group and change Measurement Stop Frequency to 22 kHz
- Disable right channel (click on RIGHT button)
- Add RMS measurement (click on RMS button)
- Add THD measurement (click on THD button)
- Set to + 18 dBV full scale input
- Set to +4 dBu 1 kHz output (right click on GEN1)
- Enable GEN1
- Press IDLE button
At this point, take a DVM and measure across L+ and L- and confirm it is +4 dBu/1.227Vrms. The QA403 is now generating a fixed 1 kHz balanced signal at +4 dBu.
Basic Setup
Check meter at “GR” to be “0”. If not, adjust “Meter Zero”
There is no connection to the compressor at this point. So, just confirm 0 reading with meter switch at GR setting
Apply 1kHz @ +4dBu at the input. Check that the meter reads “0”
at “Input” Setting. If not, adjust “V gain”.
Apply balanced QA403 output signal to compressor input and verify 0 reading on meter with compressor meter switch at INPUT setting.. Adjust if needed “VGAIN” as needed. There is no bursting from the QA403 at this point.
Check “Output” for 0 on meter. If not, adjust “Pinch off”. The actual
measured output will be about +4.3dBu.
While still in continuous mode, switch meter select switch to OUTPUT and again confirm 0 reading on meter.
Set compressor attack to minimum
Now, turn off IDLE mode. Run a single cycle via on the QA403 (ctrl+space) and confirm RMS measurement is reported as ~4.3 dBu.
Switch to time domain and inspect the plot. I’ve posted a plot of the time domain measurement in loopback below so you can get the flavor of things.. The goal here is to ensure any attack artifacts are limited to the first 40 mS. We can increases this 40 mS limit via Latency Compensation, but I suspect with Attack at minimum all artifacts will be finished by 40 mS.
Check distortion. Should be .08% or better using noise+thd, 20–
22kHz filter method. If not, adjust “Distortion Null”. If you are not
getting these results, your measurement setup may be incorrect.
The FET’s really go bad.
Press space bar to run continuously. Display will update at 16K/48K = 340 mS = 3X per second.
If distortion is not better than -62 dB (0.08%), then adjust Distortion Null on the Compressor to minimize. Remember the display updates 3X per second, so adjust, and then wait for the display to catch up and stabilize.
Compression Adjustment
Set “Threshold” to 0, Ratio to 2:1, “Attack” fastest. Release =
fastest, “Make-up Gain” = 0
We’ve already set Attack to minimum. But set the other settings as required
Step A:** Apply 1kHz @ +4dBu at the input. Adjust “Trim Threshold”
until the measured output is -1.5dBu
We’re still running in bursted mode and displaying measured compressor output level. Adjust Trim Threshold to achieve -1.5 dBu reported.
Share the time domain plot at this point
Step B: Increase the generator level by exactly 10dB, to +14dBu.
Adjust the “Ratio Trim” so the measured output is -6.5dBu
Still running in bursted mode, increase generator output to +14 dB and adjust Ratio Trim to achieve -6.5 dBu
Share the time domain plot at this point
Repeat Steps A & B. If you do this several times, you can achieve
an accuracy of 0.01 dB. All units are calibrated to this standard,
as checked on an Audio Precision.
Repeat as needed. Only the first time domain plots are needed.
We might need to tweak. But let’s try this for a start. It’s very important to share the 3 time domain plots. This is where we’ll need to determine if additional latency compensation will be needed (or not).
Hopefully we can get this sorted so you don’t need to take the AP box on the road with you!
Hey Matt…
I really thank you for all your explanations to try and get the 403 to work…
As I currently have a FETII that had some bad film Dc blocking caps on my bench I decided to try and set the compressor up following your detailed description.. here’s the rub…
I tried and tried but once I start adjusting things when compression is active things get weird. I believe there are 2 things blowing up your suggestions. 1st the attack time at various ratios has a unique KNEE to the slope of compression. Then the release is actually a dual constant release control. Meaning it releases the first portion of the amount of compression applied at a quick rate then the last portion releases slower. Also the speed of the attack is also dependent on how quickly the signal crosses the set threshold. So Inthe end I gave it the old college try with the 403. Then I went to my Ap and looked how close I was. The basic in and out levels were on. The VU meter was on. However the level settings once compression is applied were a good bit off. The THD was also surprisingly off. Perhaps I didn’t quite get thing’s correct. However I did it twice and came up the the same differences between the 403 setup and the audio precision…
I’m just guessing on the compression characteristics as being the cause of how the 403 is interpreting the data and therefore my adjustments.
Look forward to your thoughts…
Cheers.
Hi Skip,
Got it, thanks! Can you please provide the time domain signals I requested? That will answer a lot of questions.
I’m looking at the manual located at the link below (let me know if not the correct product).
In the compressor cal setup, they ask you to set thresh to 0, ratio to 2:1, attack to fastest, release to fastest, and makeup to 0.
They provide a plot that shows the expected behavior. The applied signal will cross a thresh (set to 0 dBu) and then the attack will delay the specified time (250 uS is the minimum) and then the gain will be pulled back to 4 dBu in/3 dB compression ratio=1.33 dBu output for the picture below.
In the settings for prescribed for cal, with a 4 dBu input, the gain would be pulled back to 4 dBu in/2 dB compression ratio=2 dB dBu output. The release doesn’t come into play until the input falls below the thresh.
1st the attack time at various ratios has a unique KNEE to the slope of compression.
This doesn’t seem to match their description in the manual. Once the attack time has passed, the compression is applied. It’s either compressing or it’s not.
Then the release is actually a dual constant release control.
But the release doesn’t come into play until the the end of the burst, and by then, the measurement has already finished. So, the release setting should have no bearing on the measurement. It’s just good to have it minimum like they advise.
Also the speed of the attack is also dependent on how quickly the signal crosses the set threshold.
This doesn’t seem to match their description in the manual.
Can you please show the time-domain signal so we can see what is going on? Thanks!