Sneak Peek: The QA402

Hi @Claudio, isn’t a USB hub effectively the same thing? You have a wall wart that goes into the QA402 or it goes into a hub. The only penalty is the hub (versus a QA402 that accepted external power directly). BUT, that’s not even a penalty really because it lets you power a bunch of other boxes. And since most laptops have just a few USB ports, you really need a hub anyway if you are trying to measure something with a few QA boxes.

Plus, the hubs are ubiquitous and cheap. The USB3 hub below is $17 from Amazon, and it’s 5V/2.5A

image

The USB2 hub below is $20 from Amazon and it’s also 5V/2.5A. These are both really reasonable, and if you want a test station with several boxes connected, it’s a very economical way to power everything.

Hi all, is there a time frame on this, the thing is

I do have a 400 but it sits idle while I do not have a computer that supports it (any more)
I also have a 451 and a 480 sitting useless
But I am not going to get a 401 now a replacement is on the corner

Frans,

I guess, no news is good news, so this quote from the original ‘Sneak peek article’ is holding.

… In short, right now the expectation is shipping in March 2021 …

Still looks reasonable. Here’s some loopback data comparing QA401 to REVJ QA402:

The sweet spot on the QA401 for THD is around -18 dBV, and the sweet spot on the QA401 for THD+N is around 0 dBV.

First, note that as you move away from the sweet spot on the QA401, the numbers on the QA401 degrade. This is normal and expected because the QA401 doesn’t have a bunch of relays to keep the ADC fed at it’s optimal point. So, the upshot is that at normal audio levels around 0 dBV the QA401 best case THD is below -110 dB. And as you drop to -30 dBV loopback, the best case THD moves to around -100.

On the QA402, note that THD is generally better than -112 no matter what your input level. And this is loopback. With an ultrapure input such as the QA480 take a look at the THD performance at +18 dBV input. This is measured at -122. The 2H level is around -110 dBV, which is -128 dBc

And the QA401 at that level is around -90, or about 30 dB worse.

image

I’d like to make a point about an ADC being well behaved (as the PCM4220 appears to be). First, take a look at a -1 dBV signal from the QA480 in the plot below. This is a THD of -98 (the QA480 is probably around -135 at this point). Now, your first thought might be “Hey, you just told me the THD I should expect at 0 dBV input is much better!” And that’s true. But this measurement is being made at 0 dBV max input, so we’re right at clipping.

Note in the picture below the measurement shown as “N-D”. This is the measurement of all the noise EXCEPT for the fundamental and distortions. In other words, measure the noise in all the gaps (from 20 to ~900 Hz, from 1.1 kHz to 1.95 kHz, from 2.05 kHz to 2.95 kHz, etc).

Note that figure shown here is -115 dBV. For a THD+N measurement, this would be the noise part of that measurement.

Next, let’s short the input and check the noise. Here we see the noise (20 to 20k) is -115.2 dBV. So, the ADC only degraded the noise measurement by <1 dB. That is outstanding.

Finally, let’s move to a higher input level (+18 dBV max) and measure the exact same -1 dBV signal from the QA480 (now at -19 dBFS). Now you see the THD measured at -121.8 dBc. But more accurately, we see the 3H is at -125 dBc. But the noise (N-D) is much higher. What on earth is going on?

In short, due to the number of relay ranges available to us on the QA402, and due to the fact that the PCM4220 isn’t showing a dramatic rise in noise between the harmonics when near overload, we can “piecemeal” a THD+N measurement. That is, we use one range near clipping to measure the noise, and another range far from clipping to measure distortion. And by putting them together, we know that the signal’s noise + distortion is roughly -115 dBV (noise) + -125 dBV (distortion). Longer term, I hope there will be some tools available to exploit this and allow you make measurements automatically: The QA402 will step through its ranges and let you make THD+N measurements approaching -115 dB.

In short, I’m really impressed with the PCM4220 ADC.

Some other last minute tentative changes…

THD is being prioritized over noise. The QA402 will be able to measure noise at the same level as the QA401. But it’s easy to use an external amp to gain another 10 dB of noise performance. It’s not easy to improve THD via external means.

The XLR will probably be reverting back to BNC due to size constraints. The input caps on the QA401 were 50V, unipolar (hence the negative DC limitation). The input caps on the QA402 are moving to 100V bipolar. These are much, much larger.

The input ranges have shifted again. Note the addition of +18 dBV without the atten active, and +42 dBV with the atten active. Now, +42 dBV = 126 Vrms. But normally you will operate 20 dB below the max input range to achieve the better performance. That is, if you are measuring a +10 dBV signal, you’d want to be in the +30 dBV input range. At some point, an autorange button will be added.

image

The regulatory bodies around the world are less discriminating (in terms of certification burden) when AC voltages are under 50V (see Low Voltage Directive). For this reason, if you inject a voltage in excess of 50Vrms, it will be capped at 50Vrms = 34 dBV = 70.5Vpk = in the software. Of course, testing and more review will be done to ensure clearance/creapage, insulation check at 250V DC for signal, 1000V for ground, etc.

Another point is that the front-panel I2S might only work at 48K due to processing limitations. Or the front-panel I2S might not be part of the March release and another product would be added later that had the front-panel connector. I don’t fully understand the issue yet as it’s explained to me.

Thanks

1 Like

Very well! I find the BNCs more comfortable since you can very simply use the oscilloscope probes directly without the need for adapters

Continuing the discussion from Sneak Peek: The QA402:

Hello Matt!
Disappointed about the AKM destruction…i was really waiting for the 32 bit ADC/DAC but…
I have a question, the 402 will have ASIO driver?
Thanks

An smart and enterprising person could write it and we’d support (as we did for the QA401). Longer term, I think there will be an option for standard USB audio class, with HID support for setting relays. But in both cases, the number of relays complicates things and it means the software you want to use with the QA402 with will be hard to use unless they have added support for the QA402.

Great!

When will it be possible to book the QA402?

What will the DC limits be with the new input caps? I always found the 5V limit on the QA401 a bit restricting not that i expected to be able to measure 300VDC PSU noise (would be nice though)

Having I2S and SPDIF-AESEBU I/O would be great but a separate box would probably make more sense. Maybe use the connection for a Clock sync between the QA402 and the external digital format box. You may be able to build it with a simple pairing of the CMedia CM6631 and any of the SPDIF/AES chips. You have already demonstrated software that can keep them in sync.

I do use the digital analysis capability of the RME stuff occasionally to see what the digital data is and a basic digital I/O could be really useful. For those who don’t need it this may be a place to save.

A separate thought as a work around the 50V limit and its associated overhead- outboard a USB interfaced attenuator that can deal with the high voltages. You may not need to get the whole unit certified, just the add-on box. Insulated banana in (meet safety requirements) BNC out, autoranging and internal protection, USB control for reading the attenuation. This would also be the right place for AC/DC issues for reading noise on a power supply.

1 Like

Looking at the first two plots comparing the '401 and the '402, it appears that the noise spread at the base of the QA480 fundamental tone is a fair amount less on the '402 sample.

Do you think that’s from the lower phase noise clock used in the QA402? No synchronized clocks for the converters in this case, obviously.

Per Demian’s earlier point, perhaps that’s also part of the better performance for the PCM4220.

In any case, it certainly appears that the QA402 is a step forward in features and in performance. Too bad that the AKM situation put so much pressure on designing a replacement for the QA401.

Do you think that’s from the lower phase noise clock used in the QA402?

No, I don’t think so. I don’t think you’d be able to see any jitter in an audio spectrum plot if the system is already fully synchronous and without any PLLs. Now, you can see jitter in the 24.576 MHz clock, but you’d need a very expensive instrument to measure it.

The big concern for jitter in telecom is when your reference oscillator is much lower than your frequency of interest. For example, you might use a 100 MHz reference oscillator with a PLL to mix that up to an RF frequency of 1.57 GHz in a direct conversion GPS receiver for example. When you go up in frequency, the jitter of your reference is multiplied. So, if the oscillator has 1 ps of jitter at 100 MHz, then your RF has roughly ~16 times that. Intuitively, that should make sense because each cycle of jitter of the reference results in 16 cycles at RF–the synthesized waveform is bouncing around a lot more than the reference.

Now, for audio, we’re going the other way. We start with a 24.576 MHz reference, and from that, we have a 1 kHz output. The average jitter of an oscillator over an infinite number of cycles must be zero (otherwise you’d get drift). In the case of 1 kHz, we average ~25K cycles of the 24.576 MHz reference oscillator. So, 1 ps of jitter at 24.576 MHz becomes 1p/25K = a vanishing small number.

The above is rough and conceptual. But below is a link from Analog Devices that will let you play with this in a spreadsheet if you want.

Take a look at figure 2 which shows how SNR degrades as a function of input frequency and jitter. A typical 24.576 oscillator might have 1 pS of jitter. EQ 1 shows -20 * Log10(6.28 x 1 kHz x 1ps) = -164 dBFS SNR. If you took the jitter up to 10pS, our SNR degrades by 20 dB to -144 dB. Maybe if the clock had 100 pS of jitter would we start to see it impact things visually. But that’s an insanely high level of jitter from a crystal (but not from a PLL)

In short, if you want to measure audio oscillator jitter, you need to do so at the reference frequency using some very specialized equipment and a lot of signal processing.

1 Like

Should be in March of this year.

Yes, that limit will be eased considerably. I think for noise measurements +/- 50V DC should be OK.

1 Like

Having I2S and SPDIF-AESEBU I/O would be great but a separate box would probably make more sense.

Yes, agree a lot more could be built here. But it all gets pretty specialized pretty quick. The big target for the QA402 is to try and address the i2s digital-in parts like bluetooth and smaller class D amps.

Matt
“The XLR will probably be reverting back to BNC due to size constraints.”

Nooooooo! XLR’s please!
Makes life so much easier for my kind of work. BNC to XLR adapters are cheap and and available.

1 Like

I vote for BNC. Easy to upgrade from QA401, and less chaining of adapters.

1 Like

I have QA401 and Im very pleased with it. 402 sounds like a good update to me. Subtle changes but awesome results. My vote is to keep BNC’s if there is a vote.

Im interested in the power conversion in the two products - obviously opamps have really good power rejection ratio but did you have to do anything special to get good clean split rail supply’s?

BNC is preferred by me, and I will pe-order now :slight_smile:

1 Like