Introducing the QA480

[BKDad]

Update time.

I experimented some more and discovered that the PowerWedge power strip slash power filter I’d been using was causing the basic 60 Hz and harmonics in the basic QA401 measurement. It has isolation transformers within, and they must just be radiating just enough. This was picked up even 4 feet away. (I’ve ordered a new power strip…) Using the USB hub seemed not to matter much. This is the QA401 by itself with the new noise floor - no notch weighting - actual Vrms.

Then, I tried just looking at the notch filter output with the input shorted.

I think this reveals the limitations for second harmonic measurement.

Now, this is the oscillator output into the notch filter, with 0 dB of attenuation.

Exact same thing with 5 dB of attenuation for the oscillator.

The absolute “hum” levels don’t change. Finally, I tried reorienting the QA480. Best performance was with the QA480 sitting on its side a couple feet away from the computer and the QA401.

That’s suggesting to me that it may be the cables, since the attenuator doesn’t change the hum levels. It’s hard to separate cable orientation and positioning from QA480 positioning.

So, if anybody has some magic solutions, I’m all ears. I think that any power amplifier I might have on the bench will probably radiate enough to get into this measurement, so unless I run very long cables and put the DUT in the next room, I may need a better solution.

[matt]

Nice detective work! 50/60 Hz is very tough to eliminate at these levels. For shielding, lower frequencies will benefit from a ferromagnetic material (steel) or at the extreme mumetal (nickel iron). Shields designed for higher frequencies don’t do much at powerline frequencies.

There are mumetal braided wire sleeves that you can slip over regular cables (link below). But it might make the most sense to use the QA480 setup for looking at THD(N), and then the QA401 by itself to look at hum (maybe combined with an external 10 or 20 dB amp depending on how good the DUT is).

Amazon.com: MuMETAL BRAIDED SLEEVE 3/4" I.D. x 3 Ft. length: Home Audio &...

[BKDad]

Thanks.

I’ll give the MuMetal sleeve a try. Why not?

But, I think your idea of using the QA480 for THD only makes the most sense.

Thanks very much for the help.

[BKDad]

More progress, of a sort.

The real BNC short - as in zero Ohms - terminators arrived over the weekend, as did the new outlet strip. This shows the QA401 with the inputs actually shorted as opposed to terminating with 50 Ohms as shown in my previous post.

Yeah, a couple spurs that were buried in the noise previously were revealed, but at a very low level. The point is that it’s worth spending a couple bucks on real shorting terminators, even if they’re harder to find.

[matt]

I think those spurs are artifacts of the 192K sample rate (inside the ADC). Generally, the 48K will give slightly better noise a few more dB of performance on THD(N). When time permits, if the spurs are bugging you, see if they go away at the 48K sample rate.

[BKDad]

Indeed, those spurs go away with the lower sample rate. The noise floor drops as well. So, now I see spurs at 8 and 16 KHz at an even lower level, which I think may be related to the USB sample rate.

But, these are all at such low levels I can’t really do much but take note of them and then sit back and laugh. This is ridiculously great performance for a modestly priced piece of test equipment that takes up almost no room either in the closet or on the bench.

Next step is to characterize the Crystal clock before and after, and then try to figure out some other tests that might be useful in actual audio gear performance.

Prior to the getting this all working properly, I was relying on some basic tests and the results of LTSPICE simulation. The latter wasn’t really verified. But, I measured a preamp I’d reworked and the tested results were within about a dB of the simulation. To be fair, I did try to include all the little variables that people sometimes ignore in simulation. But, still.

All in all, a very good week on the bench. Thanks for all your help!

[1audio]

I got my QA480 running and here are the results of the loopback test.

Once I got the hang of it the results are quite good. I’ll be comparing it to my other devices. And to what I get from the passive B&K notch. The combination should be really useful. I do get an error when I try to control click the freq resp button.

[1audio]

This is what I get using the QA480 notch with my Victor oscillator. The harmonics are a little lower as is the hum level. The Victor is running from batteries and has a drift issue so I may upgrade to a newer example. This is an early first gen board. Still shows the benefit of the notch.

[BDWoody]

Hi all, first post here, but wanted to upload my measurement results. It’s been a while since it’s been updated with new units, so figured it’s a good way to say hello.

Looks like it’s right where it’s supposed to be, if I did everything right.

Thanks for making these units…that’s really remarkable performance.

[matt]

Hi BDWoody, does your setup show any sensitivity to the 60 Hz? That is if you move away from an obvious AC sources, and perhaps re-orient the unit can you reduce the 60 Hz? It’s a pretty low level, but you can see it’s pervasive about the fundamental. If you could get rid of the 60 Hz then all the hash between the harmonics would likely clean up too. Thanks very much for sharing.

Another experiment that is somewhat telling is whether or not the 60 Hz levels drop 10 dB if you add in 10 dB of attenuation in the QA480. If so, then it suggests the 60 Hz is creeping into the oscillator OR some cabling from the osc to the notch. If the 60 Hz level doesn’t change when you add in 10 dB of attenuation, then it suggests its coupling into the notch or via post-notch cabling.

[GoldenSound]

This looks fantastic!

Is it likely that there will be a balanced/XLR compatible version of this in future?
I’m actually using an RME ADI-2 Pro as my ADC which in most situations works fantastically due to its performance and input ranging relays. But a frontend notch is something i’m needing.

If there is a balanced version of this released at somepoint (looks like the new analyzer is balanced?) i’d buy it in a heartbeat!

[matt]

Hi @GoldenSound. A balanced QA480 isn’t currently planned. But at some point it might make sense. I think the leap from single-ended to balanced is pretty easy to make mathematically in the absence of external common mode noise sources.

If you have a single-ended signal with a swing of X, then when you move to balanced your swing is now 2X, which gives you 6 dB more signal and thus 6 dB more SNR. However, you have more noise, because you are now using a second input. Generally, the noises are uncorrelated, so they don’t add–instead it’s the sqrt(2) which gives you 3 dB more noise.

So, when making a single ended measurement without common mode noise, if get an SNR of 100 dB and then moved to a perfect balanced measurement, your SNR would improve to 103 dB. The odd harmonics would be unchanged and the even harmonics should cancel completely if your DUT has very good CMRR. But if your 2H is already very low (as it is from QA480), then the measurement will be dominated by the amp’s 2H whether balanced or unbalanced.

In short, balanced has a huge benefit if you are fighting 50/60 Hz in a long run. But if you are just driving an amp and the 60 Hz isn’t an issue, then it’s pretty easy to know exactly how a measurement would improve if you switched to a perfect balanced setup: 2H goes away completely and SNR increases by 3 dB.

In the first link below there’s a paper from TI on the harmonic math. In theory, all of your even-order harmonics should disappear completely. In practice, as they show, a balanced measurement will knock down even order harmonics by about 6 dB.

https://www.ti.com/lit/an/sloa054e/sloa054e.pdf

The link below goes into the math on how you treat noise when adding.

https://www.analog.com/en/technical-articles/paralleling-amplifiers-improves-signal-to-noise-performance.html

And BTW, the ADI paper is really at the root of how modern converters work. Every time you double the number of channels, you cut the noise by 3 dB. Take a look at a modern ADC like the AK5578: This is an 8 channel ADC. A single channel delivers SNR of 121 dB. Using two channels to digitize the same signal yields 124, using 4 channels gives 127, and using 8 channels to digitize the same signal yields 130 dB.

[GoldenSound]

Thank you very much!
I’ll have to look about building a notch for now in that case (though if one is released by QA in future that would be awesome).

Balanced and SE performance can USUALLY be equated/worked out in theory, but there are a fair few products where design aspects cause the SE vs Bal performance to be drastically different. One example being the Schiit Magnius headphone amplifier Schiit Magnius Balanced Headphone Amp Review | Audio Science Review (ASR) Forum

I’m mostly seeking to sort this setup for product reviews, not for development, so having the ability to measure both SE and Balanced natively is critical.
Might it be possible to purchase two QA480’s and use them in a balanced setup via adapters? Or would the unit to unit variation create issues there?

[matt]

but there are a fair few products where design aspects cause the SE vs Bal performance to be drastically different.

Yes, that’s true. And the example you show is indeed a problem. But I think the correct way to think about it is that if your unbalanced versus balanced are that far apart, then you aren’t yet done with the design. So, you first work really really hard on the single-ended performance AND THEN the balanced performance just falls into place. If you start looking at only balanced, then there are problems being hidden from you. If your 2H rises by more than perhaps 10 dB when you go from balanced to unbalanced, you really need to understand why.

BUT, I don’t think there’s an analyzer setup out there today that cost under $30K that could accurate measure the balanced performance you linked. And, practically, I suspect what you are mostly seeing is the THD+N of the analyzer. State of the art today is around -122 dB for analyzer THD+N in loopback. To measure a DUT at -120 suggest the analyzer is the weak link.

Might it be possible to purchase two QA480’s and use them in a balanced setup via adapters?

Ideally, yes, but I don’t think that would give the result you needed because the amplitude variation on the QA480 is around +/- 0.5 dB. The correct solution would probably an OPA1632 driven from +/-15 rails and use that to go from single-ended to balanced. The notch would likely run into the same matching issues.

Overall, there’s a very big challenge with measuring state of the art pre-amps and some of the newer “self oscillating” class D amps.

[GoldenSound]

@matt

But I think the correct way to think about it is that if your unbalanced versus balanced are that far apart, then you aren’t yet done with the design.

I completely agree! But that’s exactly why it’s needed. Because when evaluating/reviewing other’s products, you need to be able to check whether things have indeed been done ‘correctly’ so to speak. There are numerous instances where this isn’t the case and so being able to measure both balanced and unbalanced outputs of a device separately is necessary.

@matt

BUT, I don’t think there’s an analyzer setup out there today that cost under $30K that could accurate measure the balanced performance you linked. And, practically, I suspect what you are mostly seeing is the THD+N of the analyzer. State of the art today is around -122 dB for analyzer THD+N in loopback. To measure a DUT at -120 suggest the analyzer is the weak link.

Also likely quite true. But at the moment i’m in a spot where my ADC itself can achieve about -115dB THD+N by feeding loopback of the DAC output to the ADC (no notch). And so I need a notch filter in order to get measurements of some of the higher performing devices available.

But yeah, unfortunately I don’t have $30k to drop on an APx555 or something like that, and so am simply seeking to get the best out of the hardware I have (ADI-2 Pro FS R).
The ADC itself has rather fantastic performance. Its just that I need to find a way to sort a balanced 1khz notch as much of the higher end audio market nowadays is balanced. And although it shouldn’t be the case, many products have performance disparities between their SE and Bal outputs.

If you (or anyone reading this) has any suggestions as to a possible solution it would be hugely appreciated!

[matt]

I think the solution here is to borrow a page from what has been learned in measuring phase noise over the years: You trade time for accuracy. Phase noise measurements can run for 5 minutes…and every time you increase you measurements by 10X you get a 5 dB improvement. The QA401 has a feature called “melt noise” that can readily show a 15-20 dB improvement in THDN. But that’s really just a start as to what is possible with lots of gain ranges and a two-channel measurement.

[1audio]

Another way to get a balanced source is to use a single ended source isolated from any ground. If you need a center tap a pair of resistors will give you a common point. This can work quite well with no issues about differences between + and - signals.

[agdr]

I received my new Q480 yesterday! I just installed the software and powered it up. So far I’m just trying out the relays (audible) with the checkboxes. I’ll hook it up to the scope and QA401 and another analyzer here over the next few days. The case build quality looks excellent. 350mA draw according to the USB tester. Thanks Matt!

[gvl]

Is the QA480 that can be purchased in the store the latest hardware revision with all the incremental improvements to date?

[matt]

Hi @gvl, yes, correct!