This is another affordable measuring system, though not totally cheap. DAC used is Topping DX5 (both XLR and RCA output) and the digital input for the DAC is via Toslink optical cable. This is to avoid USB ground loops that destroy the measurement (both S/N and THD+N). ADC is E1DA Cosmos. Topping D10s is used as a USB/Toslink bridge to feed the DX5 with bit perfect digital signal.
Balanced inputs and outputs are used, measuring bandwidth used is 20Hz-45kHz if not stated different. Cosmos input range set to 4.5V.
Some graphs below:
Noise density
Total system noise 22Hz-22kHz is 4.243uV
Noise in dBFS
Total noise level is -120.5dBFS (22Hz-22kHz), -122.8dBFS A-weighted
THD and THD+N at 1kHz
THD+N = 0.00017% at @BW45kHz. Distortion components all below -130dBr.
Photo
When measuring power amplifiers, I use balanced dividers in front of the E1DA Cosmos ADC.
Balanced inputs and outputs are used, measuring bandwidth used is 20Hz-45kHz if not stated different. Cosmos input range set to 4.5V.
Some graphs below:
Noise density
Total system noise 22Hz-22kHz is 4.243uV
Noise in dBFS
Total noise level is -120.5dBFS (22Hz-22kHz), -122.8dBFS A-weighted
THD and THD+N at 1kHz
THD+N = 0.00017% at @BW45kHz. Distortion components all below -130dBr.
Photo
When measuring power amplifiers, I use balanced dividers in front of the E1DA Cosmos ADC.
It is truly better than most commercial HiFi gear I work on. I recently bought a second QA403 unit for a portable station, along with the programmable load (QA-451B).
A QA forum member who is also a very talented programmer, has produced a rapidly evolving independent piece of software which is a joy to use.
https://github.com/MZachmann/QA40xPlot
Example:
Related to what full scale voltage? Loop or ADC only? These questions must be asked to know what is the usable SNR and dynamic range. That is why I would like to see the FS 1kHz as well. In case of measurements we must be very specific, to get meaningful and comparable numbers.
Sure Pavel 🙂 But full scale with respect to what? The QA403 can output 18dBV balanced (7.94/15.88V) max. With several hardware output attenuators/gain. Bear in mind the software in a hardware analyzer is driving the D/A in its sweet spot for both noise and THD and using the hardware attenuators/gain stages to produce the best outcome. I can't send it a xdBFS request. I send it a dBV, dBu or Vrms request.
Example:
Best noise floor is on the 0dBV input range, best THD is on the 12dBV and 18dBV ranges with hardware gain/attenuators.
But your settings above are at -1.14dBFS, not "full scale". What is the output from the Topping DX-5 in volts at -1.14dBFS?
The software written by MarkZ does have drop downs for various units, but I haven't confirmed what does what as the software is evolving so fast. I do not claim this software is 100% accurate, but neither do I claim any software is. The results align very well with the official Quantasylum QA40x software which also is continuously evolving.
This is the alternative QA40x Plot software A/D noise floor with shorted inputs (Fs 192kHz)
This is the same, except the outputs from the generator (set to zero) are connected to the inputs. Note, the generator does not isolate/mute itself with relays at zero output, so the comparison with your DX-5 is fair.
This is a 2V (4V bal) loopback on the alternative software
As you say, there are so many variables to consider. The QA403 also has the benefit of using cross-correlation (in the QA40x software) to push the noise floor down to ~450nV for absolute noise floor measurements. Truly it's an incredible time for affordable test gear.
Your system appears to have a ~9dB advantage on THD which is amazing. 🙂 But not when it comes to N. That's because you run into the converter noise at low levels when using a designed for HiFi D/A.
Last edited:
A really nice feauture to the UPV (and a modest one, albeit not seen on other AAs to my knowledge?) is the encoder wheel. You can click on any menu parameter box and increase / decrease by whatever increment you've parked the cursor in, i.e. 1V / 0.1V or 1KHz / 0.1KHz. It means you can find out info such as clipping points or specific frequencies of interest in a few seconds. Likewise, the random noise FR feature scans 20Hz to 20KHz in real time, with >0.1dB accuracy (you can centre a balance control knob in seconds). These are just 2 reasons that make the UPV such a workhorse. If another AA offered such practical features I'd be interested in knowing (R+S still service the UPV btw).
edit - what about a hardware remote controller for other AAs? Can anyone think of one? A small box that gives basic parameters for an AP or QA, with a handul of knobs and basic display would be a seriously cool addition to the bench (a little like those programmer boxes that came out in the 90s for synthesisers, which otherwise felt akin to painting your living room through the letterbox due to being purely menu and button-based).
edit - what about a hardware remote controller for other AAs? Can anyone think of one? A small box that gives basic parameters for an AP or QA, with a handul of knobs and basic display would be a seriously cool addition to the bench (a little like those programmer boxes that came out in the 90s for synthesisers, which otherwise felt akin to painting your living room through the letterbox due to being purely menu and button-based).
Last edited:
I almost picked up a R&S UPV with the digital option a few years back, but balked at the last second due to issues with the entire "hanging on to an ancient OS" deal. I still have a funny old DOS based audio hardware device built in 1995 that I can't let go of and it needs Win95 (direct port access) to run. It was designed in the DOS 5.xx era. Honestly, since getting the QA, it's now just taking up shelf space.
The quick options on the UPV for increment/decrement are awesome. The Standford Research SR-1 is the same, but it ran on embedded WinXP and used an encoder wheel+enter IIRC. I played with it once and thought it was a glorious piece of gear, albeit one half of the price of an AP.
At low level and using the advantage of autoranging, QA403 has a big advantage. I must do such measurements per partes with Cosmos DIP switches. However, THD+N at single sensitivity setting is better with DX5+Cosmos. To compare apples with apples, I am attaching THD+N measurement with 20Hz-22kHz bandwidth, flat unweighted. Measurement is calibrated and levels and values are readable from REW output in the upper left corner.
Below my limits, from the measurement here above:
THD+N is -117.72dB (20Hz-22kHz flat)
THD = -128.4dB
SNR = 118.3dB
measured at 4.5V input range
The rules are simple:
- THD and THD+N without BW specified are pointless numbers,
- measurement must be calibrated and we must by able to read spectrum amplitude,
- noise is to be shown as a noise density (units V/rt(Hz) which is independent on sampling rate, FFT number of points and number of averaging. Tricks to shift "noise floor" down are just marketing tricks.
Example of real world measurement, Loopback of the described system vs. DUT in the loop - DUT (with gain) stresses the DAC output less than the ADC, because of higher input impedance. This affects crossover distortion artifacts (5th order and higher), though they are below -140dBr.
Last edited:
You're right. The QA has the advantage on automated tests, where it can control its ranges, but for spot tests with the official QA40x software, we have to select ranges (not hard you get used to what is the best range for what you are looking at).
The member written new software is awesome in that it tests and controls the attenuators for spot (not automated) tests. It's not perfect and you need override its choices sometimes, but it's a step in the right direction and something that makes the entire unit more friendly for new users.
All I know is right now, I'm chasing down 50Hz spikes that weren't there last week! Nothing has changed...🙂
Why is the ancient OS a deal-breaker? You can upgrade the UPV to Win-7, but we haven't bothered on any of our UPVs. What more do you want from an AA that modern OS offers? You can export most data formats out of it. The FFT on the UPV is way more legible than one out of the 555...(the green / black format has strong contrast). You can see it across the room. Another tick for the UPV.
I think I should point out the two different paths for audio analyzers which should help explain some choices. The bulk of the usage is production test. A quick THD+N test can be a quick go/nogo eval for an audio channel. Incorrect parts, missing connections, dead power supply pops up immedately and prevents an expensive customer return.
Keithly sold a lot of 2015 DVMs in China to cell phone lines for just this task. Its also where Quant-Asylum got started. AP's multi channels switches help a lot testing AV receivers with lots of I/Os. Most of this is automated with special programs. Ideally with statiscal analysis to catch degradation before its serious.
For the hobbiest a THD test serves the same essential function- it aassembled right?
The other path is R&D which is very different. A computer control system can slow things when all you want is plug in and check the number. Then (really challenging) probe in the circuit with an audio analyzer (not recommended with soundcards). AP has done a reasonable job of quick tests but nothing like switching on the Shibasoku AM70A and measuring harmonics down to -130 almost instantly.
Still the QA40X and the Cosmos (with autoranger?) are great, very cost effective tools.
Keithly sold a lot of 2015 DVMs in China to cell phone lines for just this task. Its also where Quant-Asylum got started. AP's multi channels switches help a lot testing AV receivers with lots of I/Os. Most of this is automated with special programs. Ideally with statiscal analysis to catch degradation before its serious.
For the hobbiest a THD test serves the same essential function- it aassembled right?
The other path is R&D which is very different. A computer control system can slow things when all you want is plug in and check the number. Then (really challenging) probe in the circuit with an audio analyzer (not recommended with soundcards). AP has done a reasonable job of quick tests but nothing like switching on the Shibasoku AM70A and measuring harmonics down to -130 almost instantly.
Still the QA40X and the Cosmos (with autoranger?) are great, very cost effective tools.
That the one Richard Marsh had/has? If so, IIRC it needs an input signal at exactly 1v, and maybe only works at 1kHz? BTW, also IIRC, I think Richard improved the sensitivity of his over stock (to substantially better than -130dB almost instantly - maybe it was more like -140dB?). He showed me how he could null out H2 or H3 in an input stage with it. However, the null was only good for one of the harmonics at a time, and the input stages could only be optimally nulled at one particular signal level. I didn't quite see the applicability to general purpose audio reproduction.
Also, overall it kind of looked to me like it was really intended for production line use.
Last edited:
Richard had a 725D which I now have. I started it with a 725 I modified with improved opamps. He and a friend transposed the mods to the later units. The AM70 is the ultra low ditortion generator and the analyzer in one box. It also incorporates SMPTE IM. The relay salesman made out big on these- they must have over 100 special ($$$) relays in each unit. Mine does not have that 1V limitation. Its still really low THD+N at 10V. As you increase there is a noise penalty from the input attenuator, but thats true of all audio analyzers. It uses 16 relays for input gain scaling.