Audio Precision APx516B Review

[pma]

Yes I know, it is probably about output calibration (0dBFS out is in fact 4 or 4.2V and not 1V), but I do not sit near the workbench now so cannot confirm. ADC Input is calibrated, the 3.94V you see is correct. Digital generator level was 0dBFS. 3.94V is because of Cosmos 1.66k input impedance that makes a divider with DX5 output impedance. This is a nit- picking in fact, all I wanted to say was that IMO -105dB THD+N limit of AP516 is not good enough nowadays for a $9000 instrument, it is overpriced similarly as audio high-end .

 

[IVX]

AP sys2 getting cheaper and cheaper but its analyzer residual THD+N is about -120db@1k 4-5V(the loopback -115-118db, depending on range). I guess you can buy 2pcs SYS2522 for $9000.

 

[MC_RME]

I am missing a cross measurement to evaluate the performance of ADC and DAC path separately. Plugging the 516 output into the 555 input will give you the true performance of the 516 generator, as the 555 input has very low THD and more than good enough THD+N/SINAD. The other way round, 555 output into 516 input will reveal how much worse the 516 input is compared to the 555 loopback measurement. In the end it might show that the ADC side of the 516 is much better than its generator (maybe)...

 

[amirm]

I had to send the 516 back already as it was on a short fuse. I thought about using APx555 as a source for the 516 but then thought no one would be in a situation to have both and use them that way. So instead, I ran the DAC test which only used the ADC.

 

[JakeK]

As this measures beyond the level of audibility I would say it's good enough for reviewing most audio products and spotting anything that is flawed enough to be avoided. Of course the better measurements that Amir does are very much appreciated. I think reviewers shouldn't let the fact they can only afford this unit and not the higher end analysers stop them from wanting to do such tests and go right ahead and get one.

 

[dzerig]

Performance over the ability of a human to hear is an unnecessary waste of components and energy (assuming a listening unit, and not something for production).

I don't know what those limits are, but it seems comparing numbers to one another without a reference point of what at least the author can hear in a blind test lacks important context.

 

[DualTriode]

I don't know what those limits are, but it seems comparing numbers to one another without a reference point of what at least the author can hear in a blind test lacks important context.

The limits are know, if you know it or not.

 

[IamJF]

Limits of ability to hear ... it's not that easy cause the SYSTEM has to fulfill these limits. And we are measuring components. And there is the room ...
Simple example - noise. We can hear down to about 0dBSpl noise.

When you have a bookshelf speaker with 83dBSpl @ 2,83V sensitivity the WHOLE SYSTEM is allowed to produce 200uVrms noise at the output of the amplifier. With a large floorstanding speaker and 90dBSpl sensitivity it reduces to 90uVrms ... that's already a harder to meet point.
Active speaker, 97dBSpl 3" midrange dome - 40uVrms. That's already VERY hard to achieve, Hypex active Module has 50-80uV(A)rms with digital input. Now use a 110dBSpl compression driver ... (9uVrms).

But most rooms don't have 0dBSpl noise level, even at night. When you allow your speaker to produce 20dBSpl noise level in 1m the amp for the bookshelf speaker can have 2000uVrms noise! Even cheapes D/As and receivers will not produce noise in this scenario. Floorstanding speaker in a quiet environment and 10dBSpl noise would be 283uVrms - that already needs good components.
My listening rom is dead silent (double walls and in the middle of nowhere), I have an active speaker and use Hypex plate amps (which have less noise as any AV Preamp/Power amp combination) - I can hear a liiittle noise at listening position. In my living room cinema I use a 16R compression driver - but never heared any noise, the room is simply to loud.


So, the most basic measurement - what's the limit of noise for a D/A converter?
Don't even start with THD ...

 

[BEG]

Absolutely terrific review and discussion on the Audio Precision APx516B. Competition is the driving force behind technological innovation and creation. It’s apparent in this thread.

I’ve attached two files of measurements of my Topping E70 (ES9028PRO version). The first one is using my QuantAsylum QA403 with E1DA Cosmos Notch. The second one is using my Cosmos Notch+Scaler+ADCiso (Grade 0). As you can see both of them agree closely.

 

[Listener#1]

$9000?? Are all those parts really worth 9k? I guess price really is relative, huh.

 

[amirm]

$9000?? Are all those parts really worth 9k? I guess price really is relative, huh.

A lot of that is the HDMI adapter. I think they want $3000 to $4000 for that option alone!!!

 

[Dave Zan]

I see no reason to combine all devices....

A combined DAC and ADC with a common clock in an audio interface would simplify synchronisation.
That's important in speaker measurements, maybe not an enormous market but I will definitely buy one if you make it
It would simplify life to have one interface rather than your excellent performance unit for electronics and a lower performance but synchronised one for acoustic measurements.
I suspect others feel the same way.

Best wishes
David

 

[Roland68]

$9000?? Are all those parts really worth 9k? I guess price really is relative, huh.

This is not aimed specifically at you, but at everyone who asks a question like this and relates the value of the parts/parts/components to the purchase price.
Hand on heart, if you work several hundred or several thousand hours for something and do a good job, do you want to be paid appropriately or do you work for free?
Unfortunately, it is very common to have no appreciation for the development of devices, which always amazes me. Because none of these devices would exist without development work.

The development work is a very cost-intensive part of the calculation, and if it is not paid for, there will be no future devices and the manufacturer will no longer exist. The manufacturer also bears the entire risk of the development, because not every development is successful and not every device covers the costs incurred. The lower the number of units, the higher the proportionate development costs per device.

A second point is that most people have no idea how much, or how little, of their money goes to the manufacturer. For devices sold through normal retailers or sales platforms, it is between 25 and 50%, but closer to less than 40%. In direct sales it is around 70-75%, but do not underestimate the effort and costs of direct sales. But in principle, whenever there is a sales partner or a foreign branch, the revenue per device for the manufacturer is significantly reduced.

The third point concerns the guarantee and service after purchase. This also has to be financed by the sale of each device, from the money that reaches the manufacturer. The costs for this are in the 1-2 digit percentage range and most manufacturers have to make provisions for this every year.

 

[weme]

I am missing a cross measurement to evaluate the performance of ADC and DAC path separately. ...

I had to send the 516 back already as it was on a short fuse. I thought about using APx555 as a source for the 516 but then thought no one would be in a situation to have both and use them that way. So instead, I ran the DAC test which only used the ADC.

I have received feedback on this review from students (first year electrical engineering students): 'They would have worked through the night to fully test the APx516B with the (calibrated?) APx555!

This is your forum, you determine the quantity and quality of your measurements. This review also shows us your progress in testing devices. Here is just one example of how you manage measurement errors:

Output impedance was excellent at just 1.1 ohm:

I'm puzzled by the 1.1 Ω output impedance measured here. That's way higher than the 35 mΩ I measured. Is part of the wiring harness used for testing included in that 1.1 Ω by any chance?

Yes, the best case scenario on my setup is around 0.9 ohm so 1.1 ohm is almost as good as it gets. I have not tried to optimize this for consistency and the fact that it shows what a typical headphone would see.

If you are happy that many people today seem to buy Audio-components according to the "SINAD" table and no longer, for example, according to the output power "watts" as in the past, then that is fine with you. But, as you have already read here several times, there are people in this forum with very specialised knowledge: People who could help you take your measurements to a higher level and, above all, develop new, additional (!) measurement methods. Only then would a presentation at an AES (Audio Engineering Society) conference be logical and useful.

I would be very happy if I could read it one day:
... developed the new measurement methods with members of the ASR Forum ...

BTW:
Elsewhere here, you started a discussion about the planned measurement of MIC preamps. I have the feeling that you are (again) driven by the marketing aspect, i.e. that there will be an "EIN" table *) that is not based on the standard dBu but on dBV - simply because the dBV figures have higher values.
*)
Equivalent Input Noise table:
The input is shorted to 150 ohms, for example, to simulate the typical source resistance of a microphone. If a preamplifier does not add any noise, only the noise of the resistor (here 150 ohms) is measured:
150 ohms, 23 °C, 20 - 20,000 Hz results in -130.87 dBu or -133.09 dBV

 

[IamJF]

dBu is a horrible standard ... I'm happy with dBV. Also A weighted measurements make sense - esp. if you have also unweighted ones (so you see some of the ratio low frequency noise to broadband noise).
It's still a forum to compare consumer stuff ...
(but yes, it's important to define the measurement routines and setups. This could be made even more transparent?)

 

[JakeK]

This is not aimed specifically at you, but at everyone who asks a question like this and relates the value of the parts/parts/components to the purchase price.
Hand on heart, if you work several hundred or several thousand hours for something and do a good job, do you want to be paid appropriately or do you work for free?
Unfortunately, it is very common to have no appreciation for the development of devices, which always amazes me. Because none of these devices would exist without development work.

The development work is a very cost-intensive part of the calculation, and if it is not paid for, there will be no future devices and the manufacturer will no longer exist. The manufacturer also bears the entire risk of the development, because not every development is successful and not every device covers the costs incurred. The lower the number of units, the higher the proportionate development costs per device.

A second point is that most people have no idea how much, or how little, of their money goes to the manufacturer. For devices sold through normal retailers or sales platforms, it is between 25 and 50%, but closer to less than 40%. In direct sales it is around 70-75%, but do not underestimate the effort and costs of direct sales. But in principle, whenever there is a sales partner or a foreign branch, the revenue per device for the manufacturer is significantly reduced.

The third point concerns the guarantee and service after purchase. This also has to be financed by the sale of each device, from the money that reaches the manufacturer. The costs for this are in the 1-2 digit percentage range and most manufacturers have to make provisions for this every year.

Yes that is a very good point. I've no idea exactly how many units they sell of each model but it's not that many. I had a quick look and they have a Revenue of
$22.7 Million which might sound like a lot but it's nothing compared to Apple and if their most popular model costs ~$50k that's just ~450 units per year across 9? models so maybe 50 units per year of each model. Not very many unit sales per model to share the development and all other costs across. Plus a share of that revenue will be software sales so maybe even much less than 50 of each model of physical units are sold.

3-year standard warranty, or 5-year extended warranty
Accredited, ISO 17025 calibration from the factory
World-wide, direct support
Every new analyzer comes with a perpetual license for the current and next major software release, including a limited number of out-of-the-box measurements.

How much value do you put on that level of support as well? As a business user I would put a high value on good support like that.

If you're a hi-fi manufacturer selling thousands or maybe millions of units each year $9k is a tiny amount to pay for the ability to test your designs properly. If you're so small you can't afford that, then Amir or others may be willing to help you with testing before you release a final product?

Anyway. I agree it's not fair to try and cost such a product based only on it's bill of materials.

 

[Murray A]

As an APX555 owner I'm curious about the performance of the "B" version.

Amir, is your 555 a "B"version? What version of software are you running?

Being retired from the industry for the past few years, I'm seeing that equipment technical performance appears to be loosing relevence.

 

[amirm]

I have received feedback on this review from students (first year electrical engineering students): 'They would have worked through the night to fully test the APx516B with the (calibrated?) APx555!

Very well. Go and ask AP to see if they loan you both machines and then put said students to work. Don't come here complaining about my work after all the data that you just got without you or your students doing a thing.

That aside, you should have explained what I did to them. That a high performance DAC's output essentially played the same role as the APx555's generator so the same was accomplished without that work. I can expect your students to not realize this but what was your excuse?

 

[amirm]

As an APX555 owner I'm curious about the performance of the "B" version.

Amir, is your 555 a "B"version? What version of software are you running?

Mine is the non-B version. I know companies that have both and they reported they are identical in performance in the way I use them (there is some obscure optimization that I don't recall). I checked with AP recently again and they repeated the same. I wish the B version was better in which case I would upgrade immediately.

On software, I am running the latest 8.X which was actually required for this testing. I think I paid $2000 to get the latest software about a year ago.

 

[amirm]

If you are happy that many people today seem to buy Audio-components according to the "SINAD" table and no longer, for example, according to the output power "watts" as in the past, then that is fine with you.

That is a talking point created by anti-measurement folks. Strange to see a member here falling for it. If it were fine by me, I would not be spending huge amount of time running the rest of the tests.

But, as you have already read here several times, there are people in this forum with very specialised knowledge: People who could help you take your measurements to a higher level and, above all, develop new, additional (!) measurement methods. Only then would a presentation at an AES (Audio Engineering Society) conference be logical and useful.

AES presentation was already made: https://www.aes.org/e-lib/browse.cfm?elib=22278

The membership here has been instrumental in helping me develop the full suite we have today. Original DAC tests were far less than what I run today. If you are talking about inventing measurements, no such offer has been made and even if it had, it would not be what AP can run as its functionality is essentially fixed.

What we run here is extremely comprehensive. It routinely catches design problems. Running more may impress you or your students but won't provide more data. It will help to deter people from learning them. Nothing causes people to skip over everything than having 50 tests in front of them. Heck, I don't read measurements that are graph after graph. We need to run enough tests to figure out the nature of the product and that is that. Anything more falls in the category of measurebating which I have no interest in.