Looking forward to purchasing but wanting to check that it isn’t simply shorting the negative connections like some passive adapters etc.
Hi @GoldenSound, just to clarify, the test above with XLR was run on a previous version of hardware (with XLRs) since it was expected to be destructive. The later versions of HW moved back to BNC so it will be just like the QA401 (BNC are rated for 500Vrms, 1500V dielectric, so they aren’t the weak link)
Yes, correct! The test board uses a TLV320AIC3110 and it requires 10-20 registers to be programmed over I2C before it will do anything. So, this should eventually allow sophisticated control sequences to live inside Tractor to put your hardware in a specific state. This would allow you to test standby current by forcing the DUT into low-power mode over I2C, program different gains to help with audio test, etc.
As noted before the voltages aren’t programmable, and so the “programmability” will come by placing an isolator on your DUT test board. The QA402 will power its side of the isolator (~3 mA) and then your hardware will power the other side of the isolator with whatever IO voltage you need.
The testing that just finished used a 6" ribbon cable with a 12.28 MHz MCLK with no issues (none were expected). Longer term, there’s a lot of programmability that can be offered–MCLK, BCLK, LRCLK, bits per word, polarity, etc (QA402 must always be master). But it will take time for this feature to grow
Ah damn, does that mean the QA402 won’t support balanced input? 
Ah damn, does that mean the QA402 won’t support balanced input?
No, it supports them just as the QA401 does via BNC. Each input/output has a + and - output. On the inputs, if you want single-ended, just cap off the - input with a BNC terminator.
In balanced mode, the QA402 can generate +24 dBV into reasonable loads. Below is spectrum in balanced mode with generator at +18 dBV SE (=+24 dBV balanced). There’s a 240 ohm hanging off L+ OUT and another hanging off L- OUT. The QA402 has 100 ohms of output R. At this level, the THD doesn’t improve much if the load is removed as things are bumping up against internal rails. But, sometimes it’s nice to have a lot of drive even if only useful for higher impedance loads.
It sounds fantastic - great feature 
I’m thinking about designing a board with SRC4392 to expand the QA I/O for fully featured SPDIF/TOSLINK functionality. For the full functionality it requires I2C. Programming SRC4392 you can not only configure the bits, clocks etc. but you can also change master/slave, so it should be able to receive and analyze SPDIF signals which are always master 
I cannot wait to see the analyzer 
Hi. When are the first deliveries expected? and the price has been fixed?
Thanx
Wondering the same thing
How are things with QA402? Is there any news about release date/price?
Matt, would it be possible to see the loop measurement - residual THD and THD+H plot over the amplitude - QA402 audio-out to audio-in sweep for both channels? Can you tell me what THD/THD+N channel match do you expect? something like 1dB, worst/better?
Hi @ChUml, I’ve attached a plot of what is representative of performance in loopback, from -50 dBV output to +18 dBV output.
First, I’m going show the curves with automatic range selection active. There are 3 families of curves. In each family, there is an L and R plot as the same color. This is so you can see the channel to channel spread. And as I think you’ll see, wrt channel-to-channel, things look fine.
The 3 different groups are distinguished by a concept of “adder”…so you you see “25 dB” adder and “20 dB adder”
What this signifies is the input range that is selected for the given output range. That is, if you have specified a “20 dB adder” in the automated test, then for -10 dBV QA402 output, your target input range will be +10 dBV, which doesn’t exist, but will be rounded up to +12 dBV input (there are 8 input ranges are 0, 6, 12…42)
Now, specifying this “adder” might seem like a curious step. But it is necessary with so many input and output ranges (the output relay selection is done automatically). For best THD, the offset wants to be around 25 to 20 dB (that is, for a -20 dBV output you want to be at 0 dBV or +5 dBV input), for best THD+N, it wants to be around 15 dB. Eventually, there will be a button that says “optimize for THD” or “Optimize for THD+N” and the system will figure out the best gain range for each. But I think for most measurements the 20 dB rule will work fine. Below you can see where you tell the tests what offset to use.
For the QA401, we can run a similar plot in loopback:
Looks much less chaotic without all the relays, for sure. Roughly, from the plot above, we can see THD+N is better than 95 dB from -15 to +5 dBV input range.
The QA402 is able to extend that range, from about -18 dBV to around +18 dBV
We can replicate the QA401 settings on the QA402 and use a +12 dBV input range. You can see a flat spot at -10 dBV input level–that is the output DAC gain relay changing. And the sharp rise after 0 dB mirrors what you see in the PCM4220 ADC data sheet (see figure 6). The QA401 didn’t have the sharp rise. But we can hide that with the relays.
Now, one observation is that “Hey, this is supposed to be a -108 THD+N DAC, you are 10 dB away from that!” which is true. But this is loopback. So, you have the DAC N+D, the output amp noise and the input amp noise contributing too. Using a QA480 as the source will close that gap by a few dB, leaving the balance of degradation related to the QA402 input stages. That was a deliberate choice, because (as I noted a few posts ago) the input THD was prioritized over noise.
Now, a really big challenge today is measuring the newer “self-oscillating” class D power amps. These are delivering killer THDN performance that is so good it can’t really be measured accurately even by super expensive analyzers. Roughly, the THD+N of these amps is around -120 dB, and you’d need a -130 dB THD+N analyzer to measure that with confidence. But those analyzers don’t exist. But -120 dB THDN analyzer do for $$$.
Concurrently, we’re seeing a shift to ADC converters that appear to be getting their wins by paralleling stages and compensation. Paralleling is used to reduce noise, and compensation is used to eliminate harmonics in the converter and input/output stages. Both techniques can be used standalone, with physical hardware being replaced by time. I mentioned in an earlier post in this thread on the QA402 ADC being “well behaved” and showed some examples above how a THDN measurement can be made in several steps–one to establish the noise and another to establish the harmonic levels.
There’s a new algorithm in the QA402 software that will permit cross-correlation measurements to be made. This is a two-channel measurement that will allow you to measure noise in a DUT below the noise floor. It’s commonly employed in phase noise measurements, where 10 minutes can be spent on a measurement. Because you use two channels, noise and spectral components that are common to the DUT appear on both the left and right channels. Noise that isn’t from the DUT is seen differently by the left and right channel and thus can averaged out over time. For every 10X increase in correlations, you get 5 dB improvement in noise. So, 10 correlations gets you 5 dB, 100 gets you 10 dB, 1000 gets you 15, and 10,000 correlations get you 20 dB. The phase-noise folks are routinely getting 20 dB.
If I measure a QA480 (THD of roughly -140 and THDN of roughly -120) and let it cross correlate for 1000 samples, I get a plot as below (ignore the amplitude–there are currently some scaling issues). The measurement NmD reflects the noise minus the distortion and fundamental. That is -98.7, and the signal peak is 22.2, so that means the noise is -121 dB below the signal. This measurement was done a +6 dBV input range with a +4 dBV signal). Note the harmonics are around - 100 dBc
Now, keeping the QA480 amplitude the same, and switching to +30 dBV input and running that for 200 correlations shows actual THD of DUT at -120 (the -100 shown earlier was pessimistic and due to being so close to the 0 dBFS of the ADC.
So, as mentioned a few posts back about making a series of measurements using the attenuators with THD+N approaching -115. But that wasn’t with cross-correlation. I think with cross-correlation that can be extended even further.
In short, I think the range of attenuators will facilitate a host of enhancements from signal processing. And these will all work without a notch and at a range of frequencies. Using cross-correlation to dig deep combined with attenuators should also allow the automated determination of compensations to quash 2H and 3H inherent in the QA402. These are all very long-term items to refine and I don’t know precisely what the long-term wins might look like. But I’m really excited about the hardware framework that is in place with the QA402 and what is enabled with the 8 input ranges.
The QA401 had a detailed 10-page doc on measurements, and the doc for the QA402 will have even more when it’s ready.
PS. As for timing, originally we’d hoped for shipping in March. But with the front-panel I2S problem found and noted above, that required another PCB. That was released last week and hopefully it checks out fine. I think a reasonable date for sales starting would be May 1 at this stage. We’re not usually this forthcoming with dates, but with the AKM problem I felt we needed to share more to ensure folks understood what was coming. Unfortunately, it also means slips in schedule are more scrutinized. So, my apologies in advance.
The price for the QA402 will be $499.
Very cool! Thanks for the update Matt. I would be pleased with a pre-order if that could work for you. Seems like you have a lot of clients for the QA402 on the wait.
I guess a variation of the above could be used to implement a digital version of Bob Cordell’s distortion magnifier. One channel to measure the source and a second to measure the output of the device under test. They’d need to be time aligned through a (possibly digital) delay in one path and then subtracted in the digital signal processing. Plus some compensation for the differences between the two converters.
That would allow for even more, ahh, magnified and source distortion free multi-tone tests, albeit at some expense of measurement time. You could even add the approximated test tones back in for a better display, similar to how that’s done when using the QA480 and its notch. Not quite the same as converters with 160 dB SFDR, but life is full of compromises.
And all for less than $500… Pretty amazing.
Hello Matt,
I would be pleased to do a (pre) order now. Please put me in the right direction.
XLR is appreciated.
Hello Matt,
I would be pleased to do a (pre) order now. Please put me in the right direction.
XLR is appreciated.
Hey Matt! I just learned about and ordered a QA480 while checking for new QA401 software, then spotted these blog posts and discovered this (QA402). The 402 is finally going to happen! Awesomeeee
Sounds like you have hit all the major points from the past DIYA chats - those input capacitors on the 401 and moar power with the wider rails. Ha thank heavens you moved back to input BNC over XLRs (just my opinion). BNC rules! Just say “NO” to adaptors and break-out cables.
My one suggestion, as always, USB 3.0 for the power: “The amount of current draw for USB 3.0 devices operating in SuperSpeed mode is now 900 mA, resulting in an increase in total power delivery from 2.5 W to 4.5 W (at 5 V).” Future proof, designed for 800mA+, no hub-sorting required, cables sized for the current with limited vdrop and all that jazz. But that is just me. 
So what is up with all these Japanese foundries burning down? 
The Renesas one last week too. Coincidence? I think not → Godzilla (lol j/k). Great work on the 480 and 402 designs (and software!) as always.
I’ll be eager to get my mitts on that QA480. I did/posted a couple of notch filter PCB designs on DIYA year(s) ago for use with Victor’s oscillators, but just ran out of time to mess with them further. Looks like you’ve nailed a great oscillator + notch design, from the results! I fully understand about hand-sorting & swapping pricey high-precision resistors and caps during the build to tune that single-frequency notch. I had to do the same and yeah, time consuming!
Have fun! agdr
Hey Matt,
Please tell me how can I do to preorder or order this fantastic Q402, I want it.
Hey Matt! This looks like exactly what I’ve been wanting! Is there any info on how to preorder? I am also curious on potential Mac support. I have an unused Mac mini that work great for tech bench roles.
Thanks!
Hi Matt, all look great! Nice design however the feature I love most is the I2S port Other audio analysers have this feature only as an expensive option!. What performance can we expect using the QA480 and sweeping QA402 input with QA480 output? Have you got such a graph with all steps to see how auto ranging works? Thanks, Tomasz
Hi @ChUml,
Below is a representative plot of THD.
Below is a representative plot of THD+N
Compared to what was previously posted you can see a lot of clean-up around discontinuities. This is from smarter switching points for the relays.
For loopback, here is THD
And for loopback, here is THD+N
One issue is that the input and output relays need to be configured differently depending on whether you are wanting to optimize for THD or THD+N. For best THD+N, you want the DAC running hot (roughly -5 dBFS for given relay range) and the ADC running at -14 dBFS for given relay range. For best THD, you want the DAC running around -18 dBFS and the ADC running at -20 dBFS.
I think the way this is initially surfaced is that when you select the measurement type (THD or THDN), the last one you select will determine which will be optimized AND you see an annunciator on the screen indicating “THD OPT” or “THDN OPT” for example. But in short, you need a way to tell the system if your goal is to see best THD or best THDN.
Below is a generic breakout board for I2S. On the left side is the connector back to the QA402 (2 row, 1.27mm pitch), in the middle are isolators, etc, and on the right side are the breakout signals on 0.1" headers that you can jumper to whatever you are testing. The IO can be anywhere from 2.5V to 5.5V. We will provide these separately with the SMT parts soldered, and the connectors to be soldered by customer.
Hi @matt thanks for your very quick respond - all looks great! I just wonder why THD+N sweep over amplitude looks practically identical for QA480 as source and for the QA402 as source in standard loopback configuration 
Does it mean the THD+N performance is limited by the audio input of the analyzer? So, performing the sweep with QA480 we can practically see the performance of the analyzer audio input?
Generally the performance of the QA402 in the loop looks very good - really good device for the price. The audio performance is then comparable with the NTi XL2 which is over 3 times more expensive and has just one channel Nevertheless I still wonder why having a PCM4220 converter with a typical distortion of -108 dB THD+N, and I know that TI gives the parameters very carefully and often the ICs themselves achieve better performance, THD+N of the analyzer input are above -100dB 
what is the limiting factor here? noise?
I take the my RME ADI-2 Pro FS as example. RME uses AKM4490 DAC specified for typical -112dB THD+N and AKM5575 ADC specified also for -112dB THD+N. Performing the loopback THD+N measurement using the REW SW the system reach -113dB THD+N in the sweet spot @ -1dBFS. The THD values are spectacular.
You can also see how RME ADI-2 Pro FS auto-ranging works to keep ADC/DAC optimally fed with signal amplitude. RME ADI-2 Pro FS is much more than a audio interface.
The breakout board for the I2S port is a great idea! Do you plan something like SPDIF in/out extension card for the I2S port - would be a nice feature.
