At the end I prefer symmetrical connections making things easier - so TRS beats BNC for me.
Nice to have top of the line DACs and ADCs + input protection + calibrated gain switching + software. I have the QA400 and quite like the software.
However, loopback THD is not stellar compared with what various dedicated audio DACs can do, and neither is ADC THD with a dedicated external source. IVX gets much better performance with his Cosmos ADC which uses the same chip (and he has also very good E1DA 9038S3 and D USB stick style DACs).
So how do they manage to be easily 10 dB worse in HD? Layout? Quality of passive components?
On the other hand, those multitone and two-tone IMDs in the QA403 PDF look almost too good to be true, given the HD performance. Are they real or is this a binning thing?
However, loopback THD is not stellar compared with what various dedicated audio DACs can do, and neither is ADC THD with a dedicated external source. IVX gets much better performance with his Cosmos ADC which uses the same chip (and he has also very good E1DA 9038S3 and D USB stick style DACs).
So how do they manage to be easily 10 dB worse in HD? Layout? Quality of passive components?
On the other hand, those multitone and two-tone IMDs in the QA403 PDF look almost too good to be true, given the HD performance. Are they real or is this a binning thing?
IMO it's important to compare the complexity of input and output circuits, Zin/Zout etc.
RE connectors- Audio measurement equipment has standards, lots of them. . . HP used double banana connectors and then switched to BNC's. AP uses XLR's and BNCs and Bananas depending. . . Shibasoku uses Banana or XLR's + BNCs There is on single standard.
I prefer the BNC. The shielding is better, you have flexibility in swapping polarity etc. (All the QA analyzers are balanced in and out). Inevitably you end up with a pile of cables and adapters. I prefer to make cables that go from BNC to the intended connection, It reduces a lot of potential confusion. There are more ways to connect an unbalanced source to an XLR than you can imagine, and XLR to single ended destination is another confusion.
I prefer the BNC. The shielding is better, you have flexibility in swapping polarity etc. (All the QA analyzers are balanced in and out). Inevitably you end up with a pile of cables and adapters. I prefer to make cables that go from BNC to the intended connection, It reduces a lot of potential confusion. There are more ways to connect an unbalanced source to an XLR than you can imagine, and XLR to single ended destination is another confusion.
The Cosmos is cute but really limited. I have one but I'm not sure how best to get utility out of it. Really low distortion is nice but its not the be-all and end-all of audio. I don't think its a fair comparison. Opamps are good but driving an ADC is not a simple voltage issue. The input Z changes dynamically. Interfacing to the Cosmos needs to be done carefully to get its performance.
There are two routes for XLR: Adapter PCB or cables. For the adapter PCB you would just have to find the stacked BNC connectors the QA40x uses (or ones with the same height spacing) and buy a bunch of BNC-to-BNC adapters, measure the spacing accurately, ... . But you could have all XLR inputs and outputs.
I've gotten pretty good results by using the latest rev of the AutoRanger with the Cosmos. Perfect? No. But, it gives really good real time results - nice for tweaking stuff.
Here is a plot of a Viktor oscillator of 2013 vintage. So, it's not quite as good as the more recent ones. This was taken with the QA401 and QA480 (for just the notch filter) combination. The oscillator is battery powered.
This second plot is of the same oscillator, powered the same, same output level, into the AutoRanger and then directly into the Cosmos ADC. No notch. The Cosmos has the temperature controller board installed. These two plots are not intended to be direct comparisons - I just happened to have similar plots recorded for both test set-ups.
Looks pretty usable to me. If I need to dig deeper, I can always use a notch filter. At some point, though, I have to ask why. It's already getting somewhat academic at this point, at least for amateurs building and testing gear at home. (I know not everybody here falls into that category, but I do.)
Personally, I'm waffling about the QA403. On one hand, the hardware is clearly better in just about every way than the QA401. Maybe every way. But, on the other, the software might really require using Windows 10. Based on the trials and tribulations you guys all keep posting about with regard to various little glitches in Windows 10, I kinda like the idea of sticking with Windows 7 that I use with the QA401 and the AR/Cosmos/various application approaches. I also can use the AR/Cosmos with one of the Macs here, which is my much preferred solution. The only Windows computer I have at the moment is that one running Windows 7, and it's been very stable.
The potential issue with Cosmos is its lowish and variable input impedance but for testing a power amp or any other device with an output buffer and 50 to 600 R series resistance, it shouldn't matter or just result in a slight gain error. If you really need to look at test points within an amplifer, there should be enough good op amps these days to use as a buffer.
What Autoranger is this? Jan Didden's?
I am absolutely amazed you can achieve -147 dB 2nd harmonic with the QA401. I had the AK5397 pegged for low noise but highish distortion. Maybe some subtraction of opposite sign harmonics going on?
As you said, the other measurement cannot be compared. The harmonics could easily be in the oscillator sans notch, but they are also in the range of what Ivan reports. It would be interesting to see what you measure for VicOsc --> QA480 --> Cosmos.
I've been using Win10 for years, and it's been rather less of a hassle than Win7. Some folks report that reinstalling drivers for the E-Mu 1212m or 1616 is a challenge with each feature update but I am not running that. Get yourself a nice used notebook or tablet with Win10, you can never have enough of those. Many newer models will only have USB-C, so in that case you need a hub.
I really need to stop posting comments to the internet. My communications skills clearly suck.
Anyway...
The first plot is the Viktor oscillator run into the QA480's notch filter into the QA401. With the 40+ dB notch, I think this displays the actual harmonic output of the Victor oscillator. As I said, my sample of the oscillator is more than eight years old, and I know his newer efforts give better performance. Amazing work by him.
The second plot is that same oscillator run directly into the AutoRanger, which in turn is directly connected to the Cosmos ADC. So, I tend to think that shows what the combination of the AutoRanger and the Cosmos ADC can do, without a notch filter.
Some of the measurements can't be directly compared between the two plots because the filter settings are not the same. For example, the "base" width of the oscillator is different in the two plots. That could be attributable to the phase noise of the clocks used in the ADCs. Or, to the different filters used. Or, the software notch compensation used in the QA480/QA401 combination giving a misleading value of "base" width since the tone level is boosted in software to make the measurement easier to visualize. (The actual phase noise "base" is buried in the notch noise floor.) So, I dunno which has how much effect. I just don't want anyone to think that the Cosmos ADC clock stinks (it doesn't), even though the plots might give that impression. However, the harmonic spectral content should be comparable.
I made these plots at different times for different reasons. For the first, I just wanted to check the spectral output of the oscillator. Much later on I wanted to see what the effect of adding the thermal control system to the Cosmos ADC is, so I performed a before and after measurement. They just happened to be similar. I specifically showed the first so that everybody could see about what the oscillator itself looks like. I'd hoped that would eliminate the oscillator as the source of any harmonics in the AutoRanger Cosmos ADC combination. Guess not.
As for computers, I'm not especially interested in collecting them. Pretty much my entire collection of test gear has been chosen so that it fits on a single shelf in a closet when it isn't being used. The computer being the exception. More computers means more stuff. No thanks.
"I really need to stop posting comments to the internet. My communications skills clearly suck."
I understood what you posted just fine. Maybe since I have been following this all along it was obvious.
The spread at the base of the fundamental is a product of both phase noise and the windowing you are using. No analog oscillator will be as good as a digital source. Phase locking or injection locking can help but with significant overhead and some potential degradation in the distortion.
I understood what you posted just fine. Maybe since I have been following this all along it was obvious.
The spread at the base of the fundamental is a product of both phase noise and the windowing you are using. No analog oscillator will be as good as a digital source. Phase locking or injection locking can help but with significant overhead and some potential degradation in the distortion.
Nah - I do suck and am going to retire from posting. Just wanted to follow up for you.
In truth, the biggest difference between the base spread of the QA480/QA401 and the AR/Cosmos is the way the QA401 handles the software equalization of the QA480's notch. The tone is gained up at the notch frequency, but the base spread is not evident because it's lost in the notch noise, which isn't amped up. I looked and found that I had a couple plots comparing the QA401 driven directly from the Viktor oscillator to the AR/Cosmos driven by the same oscillator. Using the very same software (REW) for both, the same sample rate, and the same filter settings, the Cosmos actually is a bit narrower at the base.
Anyway, my whole point wasn't to compare phase noise of the two test set-ups. I just didn't want anybody to get the wrong idea about the quality of the Cosmos ADC conversion clock. Nor to think that Viktor's oscillator was the limiting factor in the AR/Cosmos test. The AR/Cosmos acquits itself quite well, at least in my opinion. A greater than 130 dB distortion dynamic range with no notching is pretty good. A year or two ago, you'd need an AP or something equally expensive to get that result.
By all means, please continue posing. This is valuable data.
I am still curious which autoranger you use - Jan's?
I am not concerned about the noise skirt / base spread, but I am still baflled that the QA401 with its AK4397 ADC should be able to achieve such low HD. Could you run Vic's Oscillator and QA480 into Cosmos, ideally with and without the autoranger?
Thanks
Who has QA401, QA402, QA403, idle or used second-hand i want to buy it .
please contact me , thk.
Mail: reverseking2020@outlook.com.
please contact me , thk.
Mail: reverseking2020@outlook.com.