• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

How do DAC chip makers verify their claims?

Tks

Major Contributor
Joined
Apr 1, 2019
Messages
3,221
Likes
5,494
In other electronics there is a reference design that is sometimes made available outside of the company to partners or concern parties, or even consumers. But in this industry I am left scratching my head. They come up with the chip, and implement it in what device, and of what design to validate their claims exactly? By this metric, they should be outperforming measurement gear? (Unless of course measurement gear makers also do their own custom DAC chips that aren't offered anywhere likewise).

Basically where does this chain begin, and how are these devices actually benchmarked when a new chip is released that purports the latest unheard-of flagship-level specs from the product offering?
 

House de Kris

Member
Joined
Apr 24, 2018
Messages
75
Likes
116
Location
Texas
Chip makers typically don't make a final product usable by consumers. Chip makers verify their claims via test equipment. Many make demo boards for 'typical' applications so makers of consumer DACs can play around with them first. All line items on the datasheet of the DAC chip are individually verifiable by testing on the chip without additional circuitry.
 

LTig

Master Contributor
Forum Donor
Joined
Feb 27, 2019
Messages
5,760
Likes
9,442
Location
Europe
Chip makers typically don't make a final product usable by consumers. Chip makers verify their claims via test equipment. Many make demo boards for 'typical' applications so makers of consumer DACs can play around with them first. All line items on the datasheet of the DAC chip are individually verifiable by testing on the chip without additional circuitry.
These demo boards are called evaluation boards, so the hardware developers can evaluate wheter the chip is fit for the intended usage. You get them also for op-amps and other chips.
 
  • Like
Reactions: Tks

Cosmik

Major Contributor
Joined
Apr 24, 2016
Messages
3,075
Likes
2,180
Location
UK
Maybe the OP is referring to 'boutique' (i.e. dodgy) chips. Evaluation boards are available for reputable brands of chip:
https://www.digikey.com/catalog/en/partgroup/cs4334-evaluation-board/28956

1558468442740.png
 
OP
Tks

Tks

Major Contributor
Joined
Apr 1, 2019
Messages
3,221
Likes
5,494
Lets say for example someone wants to make a DAC based on the AKM4499 or something, can they for instance ask for a schematic from AKM requesting what they need to do in order to reach the advertised specifications of the chip that they're selling to that person? And if they can request this, why not simply copy that design, and then simply add the finishing touches of staying within that design while also adding on personal features (like lets say this will also be a DAC you're selling that will support bluetooth).

Or the other example basically being, what exactly is used to verify AudioPrecision stuff for example? Also would it make sense for them to start offering solutions or "chips" or whatever, that are used when running the DAC portions of such measuring equipment. Or is this a case of "no this doesn't make sense for AP to use, because they can simply have their own products for sale, and since they are the cleanest, they can offer it at whatever price they want since no one else can come close to them aside from one or two other companies"?
 

SIY

Grand Contributor
Technical Expert
Joined
Apr 6, 2018
Messages
10,383
Likes
24,749
Location
Alfred, NY
Lets say for example someone wants to make a DAC based on the AKM4499 or something, can they for instance ask for a schematic from AKM requesting what they need to do in order to reach the advertised specifications of the chip that they're selling to that person? And if they can request this, why not simply copy that design, and then simply add the finishing touches of staying within that design while also adding on personal features (like lets say this will also be a DAC you're selling that will support bluetooth).

That is indeed how a great deal of design is done.
 

RayDunzl

Grand Contributor
Central Scrutinizer
Joined
Mar 9, 2016
Messages
13,198
Likes
16,981
Location
Riverview FL

tourtrophy

Member
Joined
Apr 17, 2019
Messages
9
Likes
21
Back in the days when I worked at UltraAnalog, there was an in-house developed test board for taking measurements. It 's typically a controlled environment with a linear bench power supply. We used Audio Precision System One and in-house developed test equipment for jitter analysis (That 's how we measured the AES21 SFDIF receiver jitter on the word clock). You could write scripts to automate the tests and set threshold points for failures in the production line. I don't quite remember whether we had a temperature chamber for stress tests as UltraAnalog was a small shop and hardly could survive after it lost the deal with Sony broadcast business. Since the DAC is a ladder resister type, we had a technician to run the calibration test and install the resistors manually on the PCB module. The DAC module was then encapsulated with epoxy and baked before final test was performed. Most DAC chips today should have some kind of internal ROM storing a sine wave table to sweep across the audio spectrum. Most likely there are unpublished strap pins on the DAC where you could put it into test mode. This allows you to decouple the analog section from the digital interface so that you could just spec' the analog output. I used to work at another semiconductor manufacturer for DVD applications. We built multi-layer test boards for post-silicon validation and reference designs down to two-layer PCB (to target Chinese manufacturers). We also put the boards in temperature chamber and run tests on the skew lots (voltage, temperature, and process corners). At the end of the day, our reference designs would test against Dolby compliance specs and CISPR13 EMI specs.
 
Last edited:
OP
Tks

Tks

Major Contributor
Joined
Apr 1, 2019
Messages
3,221
Likes
5,494
Back in the days when I worked at UltraAnalog, there was an in-house developed test board for taking measurements. It 's typically a controlled environment with a linear bench power supply. We used Audio Precision System One and in-house developed test equipment for jitter analysis (That 's how we measured the AES21 SFDIF receiver jitter on the word clock). You could write script to automate the tests and set threshold points for failures in the production line. I don't quite remember whether we had a temperature chamber for stress tests as UltraAnalog was a small shop and hardly could survive after it lost the deal with Sony broadcast business. Since the DAC is a ladder resister type, we had a technician to run the calibration test and install the resistors manually on the PCB module. The DAC module was then encapsulated with epoxy and baked before final test was performed. Most DAC chips today should have some kind of internal ROM storing a sine wave table to sweep across the audio spectrum. Most likely there are unpublished strap pins on the DAC where you could put it into test mode. This allows you to decouple the analog section from the digital interface so that you could just spec' the analog output. I used to work at another semiconductor manufacturer for DVD applications. We built multi-layer test boards for post-silicon validation and reference designs down to two-layer PCB (to target Chinese manufacturers). We also put the boards in temperature chamber and run tests on the skew lots (voltage, temperature, and process corners). At the end of the day, our reference designs would test against Dolby compliance specs and CISPR13 EMI specs.

Always wondered, did you guys ever consider making a retail product? I always am curious to see if that is some sort of economic reality that prevents companies involved with reference and measurements from entering the product sphere as opposed to the service/validation sphere, or if it's simply a preference thing. Is it simply a fact of the matter that the things you guys made had no bearing on being products simply due to the staggering cost, and then having to deal with applying it in an enterprise/consumer orientation or product SKU creation?
 

sergeauckland

Major Contributor
Forum Donor
Joined
Mar 16, 2016
Messages
3,440
Likes
9,100
Location
Suffolk UK
That's how I got my first ADC. I bought a Crystal Semiconductors evaluation board, put it in a box with a power supply, and got a, then, very good 20bit 48k sampling ADC. I used it for years as an analogue front end for my Meridian DSP 'speakers which had digital inputs only. I still have it and it's still pretty good.

S
 

tourtrophy

Member
Joined
Apr 17, 2019
Messages
9
Likes
21
Always wondered, did you guys ever consider making a retail product? I always am curious to see if that is some sort of economic reality that prevents companies involved with reference and measurements from entering the product sphere as opposed to the service/validation sphere, or if it's simply a preference thing. Is it simply a fact of the matter that the things you guys made had no bearing on being products simply due to the staggering cost, and then having to deal with applying it in an enterprise/consumer orientation or product SKU creation?
At UltraAnalog, we were also a design house for Adcom, Parasound, Classe, Sonic Frontiers, Threshold, PS Audio for some of their DAC processors using our modules. We only had a few design engineers including myself. Our VP, Remi, actually developed the first HDCD ASIC for Pacific Microelectronics (I remember one day driving myself to Berkeley to pick up the first lot of prototype chips). So yes, we were developing retail products but not all of them due to the limited engineering resources that we had. And we took measurements of all of the retail products we developed. We did all the tweaking and modifications if they didn't measure up. Production test was done by us as well since we had all the test jigs for DAC measurements. During my 2.5 years at UltraAnalog, I did the PS Audio Ultralink2 including retrofitting it with HDCD, Sonic Frontiers SFD-2, and Theta Chroma DAC. The latter didn't even use our own DAC module but a Burr Brown DAC in order to meet the cost target. At least one other Adcom DAC was using Burr Brown as well. Unfortunately, our company didn't have the economic of scale and did things rather conservatively. When I first joined, I couldn't believe the company didn't want to invest in schematics CAD tools and we had to draw our schematics with pencils and drafting paper!!
 
Last edited:

Sawdust123

Member
Joined
May 3, 2019
Messages
49
Likes
159
My biggest customers at AP were chip developers. Unlike discrete circuits that can be bread-boarded and are relatively cheap to iterate, chips are designed wholly in software. When silicon is finally available, it gets distributed to multiple verification teams that each focus on different parts of the chip. For instance, the analog and power stuff will be verified by different teams than the radio and phone stuff. Simultaneously, there could be other teams working to develop the eval boards for customers. Some of the cell phone chips are so complicated that the manufacturers basically provide reference designs that can be used as-is in the final product.

I like to point out is that the average chip is tested far more rigorously than the average piece of gear it gets used in. A precision op-amp data sheet comes with about 40 graphs, each with multiple traces that show behavior under different conditions. The typical audio gear is given a single frequency response and distortion test at one condition. Why do the finished goods require less testing than their parts? IMHO, the answer is they don't, we just don't want to pay the expense.

As for calibrating the APs, it is pretty simple albeit labor intensive. The only quantities you can really calibrate are frequency (time) and level. The other measurements are derived from that. So basically you need a precision volt meter and frequency counter traceable to NIST and you choose test points in every voltage and frequency range. AP provides calibration data with new instruments and you can see the many tests that are performed. Of course, calibration has to be done in a temperature controlled lab with regulated power and every last detail is documented to ISO 17025 standards.
 

shanecoughlan

Member
Forum Donor
Joined
May 13, 2019
Messages
72
Likes
85
Location
Japan
Lets say for example someone wants to make a DAC based on the AKM4499 or something, can they for instance ask for a schematic from AKM requesting what they need to do in order to reach the advertised specifications of the chip that they're selling to that person? And if they can request this, why not simply copy that design, and then simply add the finishing touches of staying within that design while also adding on personal features (like lets say this will also be a DAC you're selling that will support bluetooth).

Or the other example basically being, what exactly is used to verify AudioPrecision stuff for example? Also would it make sense for them to start offering solutions or "chips" or whatever, that are used when running the DAC portions of such measuring equipment. Or is this a case of "no this doesn't make sense for AP to use, because they can simply have their own products for sale, and since they are the cleanest, they can offer it at whatever price they want since no one else can come close to them aside from one or two other companies"?

In modern electronics there are various discrete components that a company can elect to use. One component is a DAC, a chip specialized in digital to analogue conversation, and this is usually provided by a third party such as Asahi Kasei. This third party will provide an evaluation board or similar during the prototype phase of building a new product. Their evaluation board will come with very detailed information, schematics and test suites to review, test and confirm each advertised feature.

There are plenty other dedicated components in a finished product. You referred to Bluetooth. This involves a combination of a Bluetooth chip (a radio chip) which may also cover other types of radio activity such as WiFi or wired networking. The more it does, the more complex the chip, and the more complex the software running that chip tends to be. A lot of product categories in the technology industry use Linux-based networking stacks to address this.

This networking capability will have its own evaluation board for the prototyping stage.

And now we move to other chips. General I/O (input/output) around digital connectors such as USB. I notice XCOM is popular in the audio world. Another evaluation board. This evaluation board may also have some more generic I/O to address things like optical (another chip) and so on.

Now you have a bunch of evaluation boards provided by various companies. Each board has a defined set of activities. You string them together, with the interconnecting glue either provided by hardware or - in more sophisticated setups - via software. You test out the combination to make sure everything works. You probably have to make adjustments. Let’s say the XCOM is making a call to a register that the networking chip is also calling. Change the register for one, no more congestion. And so on.

When you have everything working, and your input/output materials meet your requirements, you now know how to build your device. You send a prototype order to a manufacturer and they send back a completed combined circuit board. You test, fix and revert.

When all this is done, you have a product working according to spec. You can then move towards production.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,386
Location
Seattle Area
A DAC eval board is unlikely to have USB input. So you have to source and integrate that subsystem as a minimum to have a real product in today's market. Doing so will usually mean a new PC board and then all bets are off with respect to performance.
 

tourtrophy

Member
Joined
Apr 17, 2019
Messages
9
Likes
21
I have seen DAC eval boards from Crystal and TI with SPDIF receivers as they make their own SPDIF and you could inject a digital sine wave sweep to measure the performance of the DACs. USB is pretty much Xmos exclusively and if SPDIF is available, it doesn't make sense to add USB to showcase the DAC performance as it adds complexity to the eval board and PCB design. That said, when I worked on DVD reference designs to show case our DVD codecs, I had to choose other third party components such as audio DACs, NTSC/PAL video decoders, opAmps, all the way down to power supply and DVD transport in order to complete the whole system for showcasing our applications and the bill of materials. Without these other components, our end customers would not be able to evaluate our codec and might not know the overall system cost.
 

JohnYang1997

Master Contributor
Technical Expert
Audio Company
Joined
Dec 28, 2018
Messages
7,175
Likes
18,292
Location
China
I almost felt like I can't read English.

For reference design, yes, it's evaluation board. And most provide separate datasheet for evaluation board like akm and analog device etc.

For the measurements equipment they used to produce datasheet. It's relatively easy to measure individual performance aspect using different specific measurement circuits.
For example, testing noise, use a very low noise input opamp and setup gain of 30db then convert back after adc in digital. For distortion, just use a notch filter before adc.
Audio precision is a completely different beast. Ap does every thing very well, extremely well for something even. But for example the input noise is specified as 1uV maximum, irl 0.6uV typical. Simple opa1612 can do 0.3uV, if you make massive paralleled opa1612 it's possible to achieve sub 0.1uV noise. It's different for manufacturer just using ap to measure everything with Audio precision doing their own R&D with different test circuits. It's like whether we have egg or chicken first paradox which isn't a paradox. They just use different circuits to maximize the measurement resolution and progress over time.
 
Top Bottom