*I decided to go into a lot of details when examining this instrument, both in the teardown as well as the experiments. Please let me know if this level of details is too much!*
Nice review! Modern electronics witchcraft. Puts my 1980s HP 9803A audio analyzer to shame. And I can’t believe how affordable it is! This must set some new record in the bang for the buck category. Or utter silence for the buck in this case…
I just got one of these just to measure path delay in AES67 audio networks. I was very happy to see your feature - I know it's from the same batch i got mine from! A++ work sir, your tutelage is amazing.
Not the kind of instrument I'd use, but I'm very surprised at how good it is and how little it cost. When at the end of the video you showed the price I was very confused. For a niche instrument like this with this amazing quality and software I expected a price on the order of multiple thousands of dollars. The guys who made this should be proud!
Very thorough, informative walkthrough. Thank you! I do have a question about this unit that doesn’t seem to be dwelt with much either here or on the QA website. There is a generator built in but, there aren’t any specs on the QA website or in this video. It’s nice to have a nice low noise input to the analyzer but, it’s even more essential to have a low noise and distortion signal generator. QA always seems to use this (?) generator with a notch filter or show the analyzer capabilities with another, external oscillator. You show the two sections in loop back but, you don’t really qualify the parameters that matter. Please advise.
At a $599 price point, this is a no brainer for anyone testing audio equipment, DACs or speakers. The top of the line is the Audio Precision APX555 which runs around $30k new.
Many thanks for this video! It's a pity, that you obviously can't synchronize to external frequencies. Is it possible to precisely control / trigger on the phase between internal oscillator and ADC acquisition? You might compare this impressive low THD performance to your HP3458A, which has an ultra-linear and low THD ADC as well. In 1990, I built an FFT analyzer with this instrument. The 195Hz ac signal was in the µV .. mV range, amplified by a low THD, low noise x 100 pre-amplifier (better than the SR560) and the front-end of a LI, using a very weak LPF @ 50kHz, i.e. using a rectangular window. The 100kHz sampling frequency was phase and frequency locked, so it was possible to make a complex FFT on the magnetic response of High Tc Superconductors (X' + iX'' and higher Harmonics) with high accuracy regarding amplitude and phase (+/-50ns), and no frequency leakage. The NIST and PTB still use a similar method to precisely measure audio ac signals, by extracting the internal 10MHz ADC clock of the 3458A, and phase-synchronizing that to the input signal. accuracies
No phase measurements, I guess the dac/adc do not use zero phase halfbands. I am also concerned about what is going on above 20KHz. A lot consumer audio interfaces have a +36dB bump in noise above 20K (when running at 96/192K). My guess this is how they do asrc to sync the clock domains, by using clock jitter. This sort of out of band noise can affect a measurement of an opamp. You may want to measure the noise floor of the dac using a different instrument that displays to 96KHz.
The Wandel & Goltermann Audio Analyser NFA-1 was the gold standard in the eighties and nineties of the last century. It comprised of 16 bit DACs and ADCs and a bunch of lumped element filters. It's price was that time >50.000 DM (around 25000 $ @1990). Now this for 700 €. Tempus fugit.
The standard instrument for this application is the Audio Precision APx555 which is around $30,000 list price. Would be great to see a comparison. An immediate disadvantage for me is that the QA unit has a much lower input voltage range, whereas the AP inputs are rated up to 230V peak. This is useful as you can directly measure the outputs of high power audio amplifiers like a differential probe. Only an external LPF is required for PWM amplifiers.
Finally a piece of test gear I have at home ;) Since the SR560 also makes an appearance here: I built a small front-end clone PCB of that (due to it being covered in The Art of Electronics), same topology, better noise performance, all on a credit-card sized PCB. Fun little project.
@@kafinmufid994 TH-cam has been deleting my comments with links recently, unless they were other YT videos. I used JFE2140 instead of the original JFET and reduced the resistor values as already suggested in TAOE. Changed the Vds that is forced across the JFET pair to 5.1V instead of 6.2V, will eventuall go even lower (maybe try a 3.3V Zener), also slightly lowered the bias current to between 2-3mA per JFET. Everything else is pretty close to the TAOE/original Schematics floating around on the web (I was experimenting with drain resistors and JFET stage gain but was lacking some footprints for compensation components). I used 1206 (imperial) resistor footprints throughout the signal path and fitted Vishay MMA0204 MiniMELFs due to their comparably low excess current noise (see "Resistor Current Noise Measurements" by Frank Seifert). I also replicated the front-end circuit on a single layer home-etched board with TH parts and LSK389 for the input JFET, works nicely too but the single layer construction comes with its drawbacks regarding EMI etc.
@@kafinmufid994 I was going to make a PCB revision 2, fixing all the errors of revision one (and adding relays for the selective input grounding, input reversal and AC/DC coupling on the PCB) but got it got bumped down the priority list. As it is the board is not really optimized for hand assembly, cleaning or BOM cost. I want to change the miniture trimpots and latching relays to TH to allow for an ultrasonic bath after full SMD assembly for example. And some other stuff...
Nowadays they all offer higher sampling rates and more bits, but I can't see any significant improvements in the last 10 years. But to be honest, there are usually much more important parts in the audio chain that need to be improved, so here is no real need, except for the marketing team. In terms of audio this is possible since at least 2005? In a loop test AD(CS5381)DSP(ADAU1452)DA(CS4398) (A-weighted) I have seen 0,000627% THD+N (-104dB) at 2V RMS (differential). The OpAmps used in this application are OPA1612 as well, they are available since 2009? AD/DSP/DA run synchronous clocked by a CCHD-957.
@@fio6098 agreed. I'm not that into audio anymore, but I remember when the Wolfson WM8741 came out in 2007 with 128 dB SNR and slightly worse THD+N than the Cirrus Logic. It's a trade-off of course, but the SNR has been up there for a while.
Thanks for the thorough and detailed video. Having designed a device using this DAC before, I found it especially interesting seeing how a ‘pro’ approached a similar problem! Also it’s nice to have a break from spectrum analysers 😂
This is quite a device. I've spent some hours with Audio Precison 525 and d-scope products, and while they have their place, this looks useful. Basic io done very well. Thnaks for sharing and for your thorough review.
I wanted to purchase the QA403 for a long time but then the component shortage came and the unit became unavailable for a long time. Then in EU (UK - well, almost EU…) Elektor started selling the unit a few months back. I bought one for my audio projects….yet maybe not the same as an AP but for a fraction of the price it is a really useful tool. Finally the QA404 should have come out this year but this never materialised…
These reviews are awesome never to much detail just perfect ! I am very interested in audio but have a long way to go before I have a workable skillset but I love learning about it. I have been using high end opamp's to build 1kHz sine waves stabilized by bulbs and based around the Wein Bridged oscillators as I am trying to build a practical board as part of my test equipment. Do you think this is instrument is a waste of money for a guy who's just starting out or is it perfect because I can grow into it please ? Anybody else please you thoughts are most welcome...cheers
Nice hardware. Too bad they do not provide their application on more professional operating systems. I am done wasting time fighting with Windows for lab work.
I wonder if with a firmware update the QA403 could be used as an Audio VNA. There is a lot of audio snake oil that needs a VNA graph to debunk with data, and on the flip side wonderful and interesting responses in things like tube-amps that also would show up. Like "warmth" only appearing when you do broadband chords that won't appear with pure tones.
@@tommihommi1 vector network analyser. Looking at the signal reflections between amps and speakers. There are some interesting interactions at the system level.
@@tommihommi1 No, VNA is about sensing current and voltage on a port separately, instead of only measuring one and inferring the other by multiplying/dividing with the nominal impedance. E.g. speakers are typically highly inductive, and often speaker effects/distortions could be reduced if the amplifier could be more of a current source than a voltage source. But even that is very idealized, of course. Hence the desire for VNA tactics.
That's a valid point. Lot's of test on ASR are just useless as they don't show whole truth what's going on with signal. I'd be very interested in such use case as I always believe in audio hearing is believing, not just limited measurements as indication of DUT sonic quality. How is that tube devices sounds so good given higher THD etc versus uninvolving, clinical sound of newest silicon DACs, AMPs etc with 0.0000x% THDs. Also, such Audio VNA would be great to measure difference that different cables (signal, power) brings and are easy to spot (to hear) with high resolution audio setups but are tricky to measure using standard tests. If anyone knows of such device and test designed to show such differences please let me know.
To get rid of 60hz spike just connect USB GND to signal in GND of the interface. It is isolated system meaning there is no path fro 60hz signal to go, but as soon as you connect it to USB ground it will remove it.
Might I add that a competing product, also using the ESS9822, but without a generator, is available for approximately half the price. It’s called the E1DA Cosmos ADC.
I was just gonna mention this as well. There is also a Cosmos DAC. That way 3rd party software can be used. Obviously the Q403 is a more complete package incl software. Both way more affordable options compared to Audio Precision.
Yep, I use the E1DA Cosmos ADC for all my measurements -- phenomenal for the price. E1DA has a whole line of crazy high-performance AF measurement gear, but you don't get the software integration of the QA stuff. I just use REW, but having a dedicated software package would be quite nice.
€800, that is a competitive price, the only downside is, you have to use a screen, and where place is an issue, this can cause problems. Or removing other test equipment. Great video tho.
Hi thank you so much for your extremely valuable analysis I missed the part number of the opamps used inside the unit Their performance looks state of the art ? Unfortunately the best figures are around 6V output A little too much to drive a power amp Maybe with a voltage divider at the output ?
9:53 it would be interesting to know why they implemented the power supply like that with transformers instead of using off the shelf isolated regulators... price/performance most probably but still a good question
Great question! My guesses, they wanted to synchronise the switching with the ADC clocking, and have this fully software defined to allow optimisation? Many smps chips will allow external clocking or sync but not full control of the feedback loop
@@ralfstocker7742 OK thanks. So apparently they are driving a center tapped transformer to get split supplies and regulating them with LDOs. They claim really good noise performance but remain somewhat reserved in terms of cost efficiency.
14:33 to lower the DANL by another ~10dB, QuantAsylum should add a button to mathematically subtract the no-signal noise floor from the measurement, as is commonly available on RF spectrum analyzers from Keysight, Rohde & Schwarz, and others. Or use a min peak detector with enough sweeps to push the noise floor down, as a debugging tool to identify spurs hidden within the noise. The residual noise visualizer from 20:14 is of course another way to learn more about the distortion characteristics of the DUT, and would help with this type of analysis.
You should do a circuit board analysis of the Kitsune KTE Holo May DAC, which is an r2r ladder DAC. I think it would impress you. It’s my favorite DAC.
Would be nice to do some R2R DACs comparison and measurements. r2r ladder is one thing, other one is good I/V stage to achieve good quality sound from DAC.
Yeah its an end game level DAC in terms of sound quality. I own one and I wouldn’t trade it for any DAC at any price. But from a circuit standpoint its a unicorn because it surpasses the limitations of the traditional resister ladder where it can rival DACs using FPGA processors like say a DCS Bartok, Mola Mola Tambaque, or Chord Dave. The circuit board layout is a work of art. Same goes for the Benchmark AHB2 power amp which rivals the THD of that Stanford Research amp this channel did last video. The Benchmark AHB2 has a -137db THD/N bandwidth. And the Holo May is a perfect match for that amp.
wow .impressive .Here I sit with an analog Lindos audio test set from the early 80s it can just manage -100 db measurements with careful grounding .designed for audio test gear .
Nice product, although I wonder if I would make my design so dependent on TI stuff..... Look how bad they have performed during corona time. Lots of companies were brought to the edge of existing due to this case, like ours.
I’ve had this instrument on my want list for awhile now, but I’ve been holding out for the last year or so for the QA404, but it keeps having setbacks which I’m guessing are due to accessibility to parts. I’m wondering if I should just bite the bullet and grab the QA403, though with my luck I’d buy it and then they’d announce the release of the 404. Has anyone been in the same boat, waiting for the 404 and just decided to get the 403? Curious if I should keep waiting. At the moment I have no analyzer. Out of curiosity, would this work as an analyzer to where I could check the IF alignment on radios and such?
Very nice video as always. This QA403 device is not something I am interested, but looking at their website I saw the QA351, that one picked up my attention. Thank for sharing.
*Summary* *Introduction and Instrument Overview (**0:00** - **1:00**)* - Introduction of Quant Asylum QA 403, a 24-bit audio analyzer with high performance capabilities. - Notable features include over 120 DB of total harmonic distortion (THD), two-channel generators and analyzers, and USB connectivity. - Discussion of the need for such an instrument in measuring ultra-low distortion generators. *Internal Architecture and Specifications (**1:06** - **8:49**)* - *Detailed Teardown and Architecture Overview (**1:06** - **2:15**)* - Internal inspection reveals a single PCB with digital components on the left and analog circuitry on the right. - Discussion about how the design maintains dynamic range and linearity in the analog circuits. - *ADC & DAC Specifications (**2:38** - **4:16**)* - Both ADC and DAC are 32-bit architectures by ESS with high dynamic range and low THD. - ADC operates at 192 kHz sample rate with potential for future firmware updates to double that. - *Front-end Design and Components (**4:29** - **5:36**)* - Initial op-amps are highly linear and drive the input to the ADC. - Mechanical relay and solid-state MUX are used for signal attenuation. - *Signal Chain and Op-amp Specifications (**5:49** - **6:40**)* - Detailed look at the various op-amps used, their THD performance, and how they're placed in the circuit to maximize performance. - *Output Stage and DACs (**7:55** - **8:49**)* - Similar to ADCs, the DACs are high-performance, followed by a series of op-amps. - Output stage also uses relays for attenuation. *Board Features and Capabilities (**9:08** - **10:57**)* - 9:08: Deck runs coherently, loopback maintains frequency and samples line up. - 9:19: Some correlation between channels. - 9:25: Processor interfaces with USB. - 9:30: PCB cut separates two halves for galvanic isolation to avoid ground loops. - 9:46: Expansion port with capabilities like I2S. - 9:53: Transformers pass power to the left circuit side. - 10:04: Optoisolators send control signals back and forth. - 10:16: Loop-back test used for performance verification. - 10:27: Two channels allow double measurement; single-ended test reveals harmonic distortions. - 10:57: GUI lightweight and regularly updated, includes comprehensive help file and Wiki page. *Software Interface and User Experience (**11:26** - **20:18**)* - 11:26: Despite being from a small company, attention to detail in GUI is high. - 11:44: Reports USB voltage and current, REST server for API runs in background. - 12:04: Sample rate and FFT size adjustable; impacts frame rate. - 12:29: Noise figure influenced by attenuation. - 12:47: Markers can be placed for measurements, some limitations exist. - 13:15: Signals very small, around -139 dBV. - 13:26: ADCs and DACs have separate controls in GUI. - 13:38: Multiple window types for FFT available. - 13:56: Dynamic range optimized by adjusting attenuation. - 14:25: Output from generator as expected; harmonics buried in noise floor. - 15:04: Compromise needed between noise and nonlinearity. - 15:30: Averaging reveals harmonics. - 16:16: Extensive built-in measurements, including THD. - 16:55: THD measurements are excellent; some variations between channels. - 17:26: Nonlinearity coming from DAC, not ADC; due to semiconductor variations. - 18:03: Cursors and additional THD measurements can be applied for further analysis. - 18:09: Discusses two different channels with a 6dB difference in THD (Total Harmonic Distortion) between them. - 18:20: Mentions that channel 3 performs slightly better but attributes it to process variation. - 18:33: Introduces cursors for measurements, criticizes their small font size. - 18:56: Describes how the 'Peak' function works to catch the highest value. - 19:14: Notes a 7.5dB difference in third harmonic between channels, emphasizes quick measurement capabilities. - 19:25: Clarifies that measurements continue even if the channel display is turned off. - 19:38: Talks about built-in visualizers like the THD bar graph, criticizes UI limitations for not showing numbers for both channels. - 20:18: Introduces another visualizer called "residual display" for examining non-linearity. *Testing and Comparative Analysis (**21:25** - **39:30**)* - 21:25: Describes dual-tone generation at 1.1 kHz to study intermodulation terms. - 22:00: Comments on the system's linearity. - 22:53: Shifts to Stanford Research Systems (SRS) DS360 Ultra Low Distortion Generator for performance verification. - 24:57: Utilizes Blackman Harris window to reduce leakage, results closely match previous THD measurements. - 25:32: Begins characterization of SRS SR560 Low-Noise Preamplifier, lists its features including filters and gain modes. - 27:06: Shows the output of SR560 with no input, mentions power line noise interference. - 27:23: Set SR560 gain to 0 dB; expecting 1-to-1 output to input ratio. - 27:35: With 1 kHz at -20 dB Volt input, output measures at about 100 mV RMS, confirming near-zero gain. - 27:54: THD (Total Harmonic Distortion) measures at -90 dB, dominated by second harmonic. - 28:14: Increasing amplitude to 0 dB Volt; THD worsens to -70 dB due to input/output nonlinearity. - 28:43: Frequency response measured using exponential chirp; SR560 shows flat frequency response from DC to 20 kHz. - 29:58: Filters tested; 1 kHz low-pass with 6 dB/octave drop confirmed. - 30:27: High-pass and band-pass filters also confirmed. - 31:22: Increasing gain to 20 dB, 40 dB; signal and noise both amplify as expected. - 32:19: Residual display reveals input/output nonlinearities at very small signals. - 33:45: Automated THD vs. input/output power test initiated; noise dominates initially. - 34:08: THD worsens beyond a certain input level due to harmonic distortion. - 34:56: Comparison between High Dynamic Range and Low Noise modes; trade-offs noted. - 35:45: Setup to test a two-stage audio amplifier; board expected to have 40 dB of gain. - 36:02: Amplifier circuit explained; class-A with active load and local feedback. - 36:34: Test with 1 kHz input; near-zero THD, signal amplifies by 40 dB as expected. - 36:54: Examining harmonic structure; mostly third harmonic, neglible second harmonic. - 37:44: Power spectrum analyzed; noise floor consistent with expectations. - 38:23: Multi-tone input used to check intermodulation. - 39:30: Checks final response; low levels of intermodulation, even-ordered harmonics well below the noise floor. *Conclusions and Recommendations (**39:55** - **45:08**)* - 39:55: Summarizes findings, praises device's high performance, broad capabilities, and versatility. - 40:28: Mentions minor concerns, mostly related to GUI. - 40:54: Recommends QA403 for its performance and pricing. - 41:30: Anticipates future software improvements. - 41:55: Final thoughts; QA403 a compelling solution for high-performance audio analysis. Disclaimer: I utilized GPT-4 to condense the video transcript into a summary. The original transcript exceeded GPT-4's context length limitations, so I divided it into five segments. For each segment, I employed Prompt 1 to generate timestamped bullet lists. I then used Prompt 2 to organize these lists into sections with titles. The summary was manually formatted using TH-cam comment markup. Prompt 1: "Generate a bullet list summary, including starting timestamp for each point." Prompt 2: "Organize the bullet list into sections. Retain timestamps and use this format for titles: *title*"
At that price point it actually makes me wonder about getting to >=100 dB THD by applying the digital pre-distortion techniques from microwave systems to this. Just not sure if a 2-channel sense will do as the extreme wide band of audio and physical desirable speaker shapes make directional couplers seem like a bad idea. And compensating with two spaced sense ports (microphones in very-low-THD architecture with characterized frequency response) per speaker to differentiate forward and reverse wave would double the number of sense channels needed. Worst-case, attaching a sensor to the membrane and calibrating the air interface non-linearity aspects separately from the closed-loop digital pre-distortion, should still work and only need one sensor each. Or it might even be that for speakers that have an extremely soft internal suspension, or even no spring to push the membrane back to center (relying on PA DC bias regulation to keep it from mechanically clipping), measuring the drive current would suffice.
Didn’t know this product existed at that price point! But only working on Windows is a no go. Also needs a way to send PCM to a DAC instead of assuming only USB. I was in the process of designing something like this but now I may need to rethink 🤔
Hmm... I deal mostly with balanced sources and sinks... do i need something else as well than only this analyzer to deal with them? Or can that be ran in balanced mode and not only single ended?
If they replaced the BNC connectors with XLR and made it appear as a USB soundcard instead, they could sell it for ten times as much to audiophiles, and it'd still be better than most interfaces at that price.
That is what THD is. It is a relative measurement. In the case of the digitizer, this parameter defines pretty much the worst case scenario because there are passive attenuators in the front-end.
Awesome video as usual. It’s funny that you decided to review an audio analyzer a few days after I decided to delve deeper into audio amplifier design. I was watching your video on class-D amplifiers a couple of days ago, which is also superb! You mentioned at the start of the video that this is a 24-bit analyzer, but the ADC seems to be 32-bit. Is that because there’s is some degradation in performance on other parts of the system, or is this supposed to be rated as a 32-bit analyzer? Thanks again for the amazing video!
I can try to answer. If you assume a +1V range, then 1/2^24 is a 50nV step! 1/2^32 is a 0.2nV step. These voltages are so small that just temperature differences across an IC can lead to massive swings in value compared to the least significant bit. So while the ADC will give you 32bits, preserving information from far away from the ADC into the ADC chip (or even within the ADC) is extremely challenging.
*I decided to go into a lot of details when examining this instrument, both in the teardown as well as the experiments. Please let me know if this level of details is too much!*
Just started watching, however, you always provide the right amount of detail. Bring it on!
The detail level is fine.
Not too much at all. Well done as usual!
6 stars for you, you deserve 1M sub
I want more details ❤️, thank you for your effort 🌹🌹
Nice review! Modern electronics witchcraft. Puts my 1980s HP 9803A audio analyzer to shame. And I can’t believe how affordable it is! This must set some new record in the bang for the buck category. Or utter silence for the buck in this case…
Awesome instrument, worthy challenger to ancient HP 35670A's. Thank you for detailed review!
I just got one of these just to measure path delay in AES67 audio networks. I was very happy to see your feature - I know it's from the same batch i got mine from! A++ work sir, your tutelage is amazing.
Brilliant review, it persuaded me this would be a great addition to my audio testgear and I'm loving it...
Thank you for reviewing this excellent test instrument. Some useful references to high specification parts inside.
Excellent video, thanks a lot.
I am really curious what ADC that is. You didn't mention a part number. :(
It's never too much!
The GUI looks from Windows 3.1
Can one use this as a preamplifier for lock in?
At Saelig we are selling these like hotcakes!
Yeah I ordered one today and they notified me it has shipped... same day.
That instrument is impressive. While watching I expected a >=$2k price point. At that price I’ll order one straight away. Amazing.
Thats exactly what I thought ! did you buy it please?
It is but you can get better bang for buck with some pro audio USB cards and some attenuators...
@@ZegaracRobert you can’t get nearly that level of performance with a sound card.
I know, but some cards gets pretty close...and when on budget -100-120db works fine...@@AndySpicer
@@ZegaracRobert i became curious, could you name an example please?
Not the kind of instrument I'd use, but I'm very surprised at how good it is and how little it cost.
When at the end of the video you showed the price I was very confused. For a niche instrument like this with this amazing quality and software I expected a price on the order of multiple thousands of dollars.
The guys who made this should be proud!
Very thorough, informative walkthrough. Thank you! I do have a question about this unit that doesn’t seem to be dwelt with much either here or on the QA website. There is a generator built in but, there aren’t any specs on the QA website or in this video. It’s nice to have a nice low noise input to the analyzer but, it’s even more essential to have a low noise and distortion signal generator. QA always seems to use this (?) generator with a notch filter or show the analyzer capabilities with another, external oscillator. You show the two sections in loop back but, you don’t really qualify the parameters that matter. Please advise.
At a $599 price point, this is a no brainer for anyone testing audio equipment, DACs or speakers. The top of the line is the Audio Precision APX555 which runs around $30k new.
Many thanks for this video!
It's a pity, that you obviously can't synchronize to external frequencies.
Is it possible to precisely control / trigger on the phase between internal oscillator and ADC acquisition?
You might compare this impressive low THD performance to your HP3458A, which has an ultra-linear and low THD ADC as well.
In 1990, I built an FFT analyzer with this instrument.
The 195Hz ac signal was in the µV .. mV range, amplified by a low THD, low noise x 100 pre-amplifier (better than the SR560) and the front-end of a LI, using a very weak LPF @ 50kHz, i.e. using a rectangular window. The 100kHz sampling frequency was phase and frequency locked, so it was possible to make a complex FFT on the magnetic response of High Tc Superconductors (X' + iX'' and higher Harmonics) with high accuracy regarding amplitude and phase (+/-50ns), and no frequency leakage.
The NIST and PTB still use a similar method to precisely measure audio ac signals, by extracting the internal 10MHz ADC clock of the 3458A, and phase-synchronizing that to the input signal.
accuracies
No phase measurements, I guess the dac/adc do not use zero phase halfbands. I am also concerned about what is going on above 20KHz. A lot consumer audio interfaces have a +36dB bump in noise above 20K (when running at 96/192K). My guess this is how they do asrc to sync the clock domains, by using clock jitter. This sort of out of band noise can affect a measurement of an opamp. You may want to measure the noise floor of the dac using a different instrument that displays to 96KHz.
watched this and bought one! much better than fiddling about with soundcards and random software
Just watched this awesome review and am currently thinking the same !
When on budget, it works, when you got that much to spend it is worth it!
And now i get the very unwelcome large customs import bill, that wasn't mentioned in checkout :
Oh Blimey.@@Sigmatechnica
The Wandel & Goltermann Audio Analyser NFA-1 was the gold standard in the eighties and nineties of the last century. It comprised of 16 bit DACs and ADCs and a bunch of lumped element filters. It's price was that time >50.000 DM (around 25000 $ @1990).
Now this for 700 €. Tempus fugit.
The standard instrument for this application is the Audio Precision APx555 which is around $30,000 list price.
Would be great to see a comparison. An immediate disadvantage for me is that the QA unit has a much lower input voltage range, whereas the AP inputs are rated up to 230V peak. This is useful as you can directly measure the outputs of high power audio amplifiers like a differential probe.
Only an external LPF is required for PWM amplifiers.
Yep definitely beats AP, but AP is reference, like fluke in DMM's...
@@ZegaracRobert ? In what way does the QA403 "beat" APx555?
Be specific, please. I am curious. Thanks!
Very nice demonstration. That is an impressive piece of gear.
The most interesting part about the design here, is the isolation. I have never seen that in audio devices. Great device btw!
Finally a piece of test gear I have at home ;) Since the SR560 also makes an appearance here: I built a small front-end clone PCB of that (due to it being covered in The Art of Electronics), same topology, better noise performance, all on a credit-card sized PCB. Fun little project.
What chapter and page please if you have a moment. :)
do you post your design somewhere? would be interesting to see it.
@@andymouse 8.6.4
@@kafinmufid994 TH-cam has been deleting my comments with links recently, unless they were other YT videos. I used JFE2140 instead of the original JFET and reduced the resistor values as already suggested in TAOE. Changed the Vds that is forced across the JFET pair to 5.1V instead of 6.2V, will eventuall go even lower (maybe try a 3.3V Zener), also slightly lowered the bias current to between 2-3mA per JFET. Everything else is pretty close to the TAOE/original Schematics floating around on the web (I was experimenting with drain resistors and JFET stage gain but was lacking some footprints for compensation components). I used 1206 (imperial) resistor footprints throughout the signal path and fitted Vishay MMA0204 MiniMELFs due to their comparably low excess current noise (see "Resistor Current Noise Measurements" by Frank Seifert). I also replicated the front-end circuit on a single layer home-etched board with TH parts and LSK389 for the input JFET, works nicely too but the single layer construction comes with its drawbacks regarding EMI etc.
@@kafinmufid994 I was going to make a PCB revision 2, fixing all the errors of revision one (and adding relays for the selective input grounding, input reversal and AC/DC coupling on the PCB) but got it got bumped down the priority list. As it is the board is not really optimized for hand assembly, cleaning or BOM cost. I want to change the miniture trimpots and latching relays to TH to allow for an ultrasonic bath after full SMD assembly for example. And some other stuff...
Really great hardware, but the user interface software has a lot space to be improved.
Really considering to have one
I own the Q401and its fantastic!
Wow that is really nice dynamic range! Amazing how far things have come the past 20 years
Nowadays they all offer higher sampling rates and more bits, but I can't see any significant improvements in the last 10 years. But to be honest, there are usually much more important parts in the audio chain that need to be improved, so here is no real need, except for the marketing team.
In terms of audio this is possible since at least 2005? In a loop test AD(CS5381)DSP(ADAU1452)DA(CS4398) (A-weighted) I have seen 0,000627% THD+N (-104dB) at 2V RMS (differential). The OpAmps used in this application are OPA1612 as well, they are available since 2009? AD/DSP/DA run synchronous clocked by a CCHD-957.
@@fio6098 agreed. I'm not that into audio anymore, but I remember when the Wolfson WM8741 came out in 2007 with 128 dB SNR and slightly worse THD+N than the Cirrus Logic. It's a trade-off of course, but the SNR has been up there for a while.
Thanks for the thorough and detailed video. Having designed a device using this DAC before, I found it especially interesting seeing how a ‘pro’ approached a similar problem! Also it’s nice to have a break from spectrum analysers 😂
This is quite a device. I've spent some hours with Audio Precison 525 and d-scope products, and while they have their place, this looks useful. Basic io done very well. Thnaks for sharing and for your thorough review.
I didn't know you also know how to work with DC.
Very interesting piece of tool. Have to buy one.
For those wondering, the input stage is: TI OPA1612 (5:45) then TI OPA1632 (7:11) and finally the ESS ADC ES9821Q (2:37)
I wanted to purchase the QA403 for a long time but then the component shortage came and the unit became unavailable for a long time. Then in EU (UK - well, almost EU…) Elektor started selling the unit a few months back. I bought one for my audio projects….yet maybe not the same as an AP but for a fraction of the price it is a really useful tool.
Finally the QA404 should have come out this year but this never materialised…
These reviews are awesome never to much detail just perfect ! I am very interested in audio but have a long way to go before I have a workable skillset but I love learning about it. I have been using high end opamp's to build 1kHz sine waves stabilized by bulbs and based around the Wein Bridged oscillators as I am trying to build a practical board as part of my test equipment. Do you think this is instrument is a waste of money for a guy who's just starting out or is it perfect because I can grow into it please ? Anybody else please you thoughts are most welcome...cheers
Great specs for such an affordable ( € 575,- ) instrument..
Nice hardware. Too bad they do not provide their application on more professional operating systems. I am done wasting time fighting with Windows for lab work.
Certainly a shame it doesn't have standard digital IO too, and as you mentioned the clocking is a limitation as well, otherwise a very nice device.
I wonder if with a firmware update the QA403 could be used as an Audio VNA. There is a lot of audio snake oil that needs a VNA graph to debunk with data, and on the flip side wonderful and interesting responses in things like tube-amps that also would show up. Like "warmth" only appearing when you do broadband chords that won't appear with pure tones.
what exactly do you mean by VNA?
You can already put in either a single tone sweep or white noise, for example, and look at how the response changes.
@@tommihommi1 vector network analyser. Looking at the signal reflections between amps and speakers. There are some interesting interactions at the system level.
aren't the frequencies at work way, way too low compared to the size of devices for this to be of any use
@@tommihommi1 No, VNA is about sensing current and voltage on a port separately, instead of only measuring one and inferring the other by multiplying/dividing with the nominal impedance.
E.g. speakers are typically highly inductive, and often speaker effects/distortions could be reduced if the amplifier could be more of a current source than a voltage source. But even that is very idealized, of course. Hence the desire for VNA tactics.
That's a valid point. Lot's of test on ASR are just useless as they don't show whole truth what's going on with signal. I'd be very interested in such use case as I always believe in audio hearing is believing, not just limited measurements as indication of DUT sonic quality. How is that tube devices sounds so good given higher THD etc versus uninvolving, clinical sound of newest silicon DACs, AMPs etc with 0.0000x% THDs. Also, such Audio VNA would be great to measure difference that different cables (signal, power) brings and are easy to spot (to hear) with high resolution audio setups but are tricky to measure using standard tests. If anyone knows of such device and test designed to show such differences please let me know.
To get rid of 60hz spike just connect USB GND to signal in GND of the interface. It is isolated system meaning there is no path fro 60hz signal to go, but as soon as you connect it to USB ground it will remove it.
Might I add that a competing product, also using the ESS9822, but without a generator, is available for approximately half the price. It’s called the E1DA Cosmos ADC.
I was just gonna mention this as well. There is also a Cosmos DAC. That way 3rd party software can be used. Obviously the Q403 is a more complete package incl software.
Both way more affordable options compared to Audio Precision.
There is an isolated version by now (isolation after USB-I2C bridge, not before it). Cosmos DAC isn't released yet afaik
Yep, I use the E1DA Cosmos ADC for all my measurements -- phenomenal for the price. E1DA has a whole line of crazy high-performance AF measurement gear, but you don't get the software integration of the QA stuff. I just use REW, but having a dedicated software package would be quite nice.
@@philippwie3539 TOPPING E30 II or TOPPING E50 Balanced are also quite a decent alternatives.
A FABULOUS review, thorough, educative and very informative. Analogue audio and LF baseband measurement is alive and kicking... thank you!
€800, that is a competitive price, the only downside is, you have to use a screen, and where place is an issue, this can cause problems. Or removing other test equipment. Great video tho.
Hi thank you so much for your extremely valuable analysis
I missed the part number of the opamps used inside the unit Their performance looks state of the art ?
Unfortunately the best figures are around 6V output A little too much to drive a power amp Maybe with a voltage divider at the output ?
this thing is great
9:53 it would be interesting to know why they implemented the power supply like that with transformers instead of using off the shelf isolated regulators... price/performance most probably but still a good question
I’d like to know that myself
Great question! My guesses, they wanted to synchronise the switching with the ADC clocking, and have this fully software defined to allow optimisation? Many smps chips will allow external clocking or sync but not full control of the feedback loop
@@ralfstocker7742 OK thanks.
So apparently they are driving a center tapped transformer to get split supplies and regulating them with LDOs. They claim really good noise performance but remain somewhat reserved in terms of cost efficiency.
14:33 to lower the DANL by another ~10dB, QuantAsylum should add a button to mathematically subtract the no-signal noise floor from the measurement, as is commonly available on RF spectrum analyzers from Keysight, Rohde & Schwarz, and others. Or use a min peak detector with enough sweeps to push the noise floor down, as a debugging tool to identify spurs hidden within the noise.
The residual noise visualizer from 20:14 is of course another way to learn more about the distortion characteristics of the DUT, and would help with this type of analysis.
i was blown away seeing that price at the end
E1DA Cosmos is also a very nice set. It consists of three very cheap modules - ADC, Scaler and APU.
hi are you using it ? with which signal generator? the cheap ones are all so so
thank you
Can QA403 (mis)used as a simple audio interface ?
Hello bnc shorter plug for input, do they must be zero ohm or 75ohm?
You should do a circuit board analysis of the Kitsune KTE Holo May DAC, which is an r2r ladder DAC. I think it would impress you. It’s my favorite DAC.
Would be nice to do some R2R DACs comparison and measurements. r2r ladder is one thing, other one is good I/V stage to achieve good quality sound from DAC.
Yeah its an end game level DAC in terms of sound quality. I own one and I wouldn’t trade it for any DAC at any price. But from a circuit standpoint its a unicorn because it surpasses the limitations of the traditional resister ladder where it can rival DACs using FPGA processors like say a DCS Bartok, Mola Mola Tambaque, or Chord Dave. The circuit board layout is a work of art. Same goes for the Benchmark AHB2 power amp which rivals the THD of that Stanford Research amp this channel did last video. The Benchmark AHB2 has a -137db THD/N bandwidth. And the Holo May is a perfect match for that amp.
wow .impressive .Here I sit with an analog Lindos audio test set from the early 80s it can just manage -100 db measurements with careful grounding .designed for audio test gear .
opamp model 6:00 ?
you can find the real specialist in 22:47
Why not use a AD7760 with a 256-bit decimation FIR filter at 2.5 MSPS? Same specs for ADC and higher bandwidth
Because the ES9822 has better dynamic range and linearity, not to mention it is two channels.
Great description of this high-performance clean device. Thanks for posting.
Very informative. Thanks
Nice product, although I wonder if I would make my design so dependent on TI stuff..... Look how bad they have performed during corona time. Lots of companies were brought to the edge of existing due to this case, like ours.
I’ve had this instrument on my want list for awhile now, but I’ve been holding out for the last year or so for the QA404, but it keeps having setbacks which I’m guessing are due to accessibility to parts. I’m wondering if I should just bite the bullet and grab the QA403, though with my luck I’d buy it and then they’d announce the release of the 404.
Has anyone been in the same boat, waiting for the 404 and just decided to get the 403? Curious if I should keep waiting. At the moment I have no analyzer. Out of curiosity, would this work as an analyzer to where I could check the IF alignment on radios and such?
funny, I played with a old SR785 dynamic signal analyzer a couple weeks ago, which is commonly used for some similar measurements
Very nice video as always. This QA403 device is not something I am interested, but looking at their website I saw the QA351, that one picked up my attention. Thank for sharing.
*Summary*
*Introduction and Instrument Overview (**0:00** - **1:00**)*
- Introduction of Quant Asylum QA 403, a 24-bit audio analyzer with high performance capabilities.
- Notable features include over 120 DB of total harmonic distortion (THD), two-channel generators and analyzers, and USB connectivity.
- Discussion of the need for such an instrument in measuring ultra-low distortion generators.
*Internal Architecture and Specifications (**1:06** - **8:49**)*
- *Detailed Teardown and Architecture Overview (**1:06** - **2:15**)*
- Internal inspection reveals a single PCB with digital components on the left and analog circuitry on the right.
- Discussion about how the design maintains dynamic range and linearity in the analog circuits.
- *ADC & DAC Specifications (**2:38** - **4:16**)*
- Both ADC and DAC are 32-bit architectures by ESS with high dynamic range and low THD.
- ADC operates at 192 kHz sample rate with potential for future firmware updates to double that.
- *Front-end Design and Components (**4:29** - **5:36**)*
- Initial op-amps are highly linear and drive the input to the ADC.
- Mechanical relay and solid-state MUX are used for signal attenuation.
- *Signal Chain and Op-amp Specifications (**5:49** - **6:40**)*
- Detailed look at the various op-amps used, their THD performance, and how they're placed in the circuit to maximize performance.
- *Output Stage and DACs (**7:55** - **8:49**)*
- Similar to ADCs, the DACs are high-performance, followed by a series of op-amps.
- Output stage also uses relays for attenuation.
*Board Features and Capabilities (**9:08** - **10:57**)*
- 9:08: Deck runs coherently, loopback maintains frequency and samples line up.
- 9:19: Some correlation between channels.
- 9:25: Processor interfaces with USB.
- 9:30: PCB cut separates two halves for galvanic isolation to avoid ground loops.
- 9:46: Expansion port with capabilities like I2S.
- 9:53: Transformers pass power to the left circuit side.
- 10:04: Optoisolators send control signals back and forth.
- 10:16: Loop-back test used for performance verification.
- 10:27: Two channels allow double measurement; single-ended test reveals harmonic distortions.
- 10:57: GUI lightweight and regularly updated, includes comprehensive help file and Wiki page.
*Software Interface and User Experience (**11:26** - **20:18**)*
- 11:26: Despite being from a small company, attention to detail in GUI is high.
- 11:44: Reports USB voltage and current, REST server for API runs in background.
- 12:04: Sample rate and FFT size adjustable; impacts frame rate.
- 12:29: Noise figure influenced by attenuation.
- 12:47: Markers can be placed for measurements, some limitations exist.
- 13:15: Signals very small, around -139 dBV.
- 13:26: ADCs and DACs have separate controls in GUI.
- 13:38: Multiple window types for FFT available.
- 13:56: Dynamic range optimized by adjusting attenuation.
- 14:25: Output from generator as expected; harmonics buried in noise floor.
- 15:04: Compromise needed between noise and nonlinearity.
- 15:30: Averaging reveals harmonics.
- 16:16: Extensive built-in measurements, including THD.
- 16:55: THD measurements are excellent; some variations between channels.
- 17:26: Nonlinearity coming from DAC, not ADC; due to semiconductor variations.
- 18:03: Cursors and additional THD measurements can be applied for further analysis.
- 18:09: Discusses two different channels with a 6dB difference in THD (Total Harmonic Distortion) between them.
- 18:20: Mentions that channel 3 performs slightly better but attributes it to process variation.
- 18:33: Introduces cursors for measurements, criticizes their small font size.
- 18:56: Describes how the 'Peak' function works to catch the highest value.
- 19:14: Notes a 7.5dB difference in third harmonic between channels, emphasizes quick measurement capabilities.
- 19:25: Clarifies that measurements continue even if the channel display is turned off.
- 19:38: Talks about built-in visualizers like the THD bar graph, criticizes UI limitations for not showing numbers for both channels.
- 20:18: Introduces another visualizer called "residual display" for examining non-linearity.
*Testing and Comparative Analysis (**21:25** - **39:30**)*
- 21:25: Describes dual-tone generation at 1.1 kHz to study intermodulation terms.
- 22:00: Comments on the system's linearity.
- 22:53: Shifts to Stanford Research Systems (SRS) DS360 Ultra Low Distortion Generator for performance verification.
- 24:57: Utilizes Blackman Harris window to reduce leakage, results closely match previous THD measurements.
- 25:32: Begins characterization of SRS SR560 Low-Noise Preamplifier, lists its features including filters and gain modes.
- 27:06: Shows the output of SR560 with no input, mentions power line noise interference.
- 27:23: Set SR560 gain to 0 dB; expecting 1-to-1 output to input ratio.
- 27:35: With 1 kHz at -20 dB Volt input, output measures at about 100 mV RMS, confirming near-zero gain.
- 27:54: THD (Total Harmonic Distortion) measures at -90 dB, dominated by second harmonic.
- 28:14: Increasing amplitude to 0 dB Volt; THD worsens to -70 dB due to input/output nonlinearity.
- 28:43: Frequency response measured using exponential chirp; SR560 shows flat frequency response from DC to 20 kHz.
- 29:58: Filters tested; 1 kHz low-pass with 6 dB/octave drop confirmed.
- 30:27: High-pass and band-pass filters also confirmed.
- 31:22: Increasing gain to 20 dB, 40 dB; signal and noise both amplify as expected.
- 32:19: Residual display reveals input/output nonlinearities at very small signals.
- 33:45: Automated THD vs. input/output power test initiated; noise dominates initially.
- 34:08: THD worsens beyond a certain input level due to harmonic distortion.
- 34:56: Comparison between High Dynamic Range and Low Noise modes; trade-offs noted.
- 35:45: Setup to test a two-stage audio amplifier; board expected to have 40 dB of gain.
- 36:02: Amplifier circuit explained; class-A with active load and local feedback.
- 36:34: Test with 1 kHz input; near-zero THD, signal amplifies by 40 dB as expected.
- 36:54: Examining harmonic structure; mostly third harmonic, neglible second harmonic.
- 37:44: Power spectrum analyzed; noise floor consistent with expectations.
- 38:23: Multi-tone input used to check intermodulation.
- 39:30: Checks final response; low levels of intermodulation, even-ordered harmonics well below the noise floor.
*Conclusions and Recommendations (**39:55** - **45:08**)*
- 39:55: Summarizes findings, praises device's high performance, broad capabilities, and versatility.
- 40:28: Mentions minor concerns, mostly related to GUI.
- 40:54: Recommends QA403 for its performance and pricing.
- 41:30: Anticipates future software improvements.
- 41:55: Final thoughts; QA403 a compelling solution for high-performance audio analysis.
Disclaimer:
I utilized GPT-4 to condense the video transcript into a summary. The original transcript exceeded GPT-4's context length limitations, so I divided it into five segments. For each segment, I employed Prompt 1 to generate timestamped bullet lists. I then used Prompt 2 to organize these lists into sections with titles. The summary was manually formatted using TH-cam comment markup.
Prompt 1: "Generate a bullet list summary, including starting timestamp for each point."
Prompt 2: "Organize the bullet list into sections. Retain timestamps and use this format for titles: *title*"
Great piece of HW but also SW is a really great ! I like it so much. Thank you!
Excellent review. Great detail! I could not determine the part number of the first stage input op amp. Is it the same as the 2nd?
what sort of USB current is required with all outputs at full send / hefty loading?
Great review, loved the review and the instrument. Thanks
thats a good machine. I might buy it
At that price point it actually makes me wonder about getting to >=100 dB THD by applying the digital pre-distortion techniques from microwave systems to this. Just not sure if a 2-channel sense will do as the extreme wide band of audio and physical desirable speaker shapes make directional couplers seem like a bad idea. And compensating with two spaced sense ports (microphones in very-low-THD architecture with characterized frequency response) per speaker to differentiate forward and reverse wave would double the number of sense channels needed.
Worst-case, attaching a sensor to the membrane and calibrating the air interface non-linearity aspects separately from the closed-loop digital pre-distortion, should still work and only need one sensor each.
Or it might even be that for speakers that have an extremely soft internal suspension, or even no spring to push the membrane back to center (relying on PA DC bias regulation to keep it from mechanically clipping), measuring the drive current would suffice.
The ESS DAC and ADC chips have distortion compensation registers, but I think only for 2nd and 3rd harmonic or something.
Good balance of build details and applied use!
Didn’t know this product existed at that price point! But only working on Windows is a no go. Also needs a way to send PCM to a DAC instead of assuming only USB. I was in the process of designing something like this but now I may need to rethink 🤔
I think the software can run under Mono on Linux (don't have a reference, but I did read it somewhere today!).
Hmm... I deal mostly with balanced sources and sinks... do i need something else as well than only this analyzer to deal with them?
Or can that be ran in balanced mode and not only single ended?
I believe this is a balanced device. Well, you would need an adapter from say XLR to 2x BNC. or am I wrong?
Wow, AF on TSP!
Very good video, but you speak little to fast. 😁
If they replaced the BNC connectors with XLR and made it appear as a USB soundcard instead, they could sell it for ten times as much to audiophiles, and it'd still be better than most interfaces at that price.
Probably doable with some minor SW/HW modifications. You have a nice idea with using it as audio interface instead of DAQ :)
120 dB is a ratio. It means nothing without the reference.
That is what THD is. It is a relative measurement. In the case of the digitizer, this parameter defines pretty much the worst case scenario because there are passive attenuators in the front-end.
If you have the spectrum plot at hands it is quite meaningful, e.g. for judging noise floor which is the relation of signal/noise...
Awesome video as usual. It’s funny that you decided to review an audio analyzer a few days after I decided to delve deeper into audio amplifier design. I was watching your video on class-D amplifiers a couple of days ago, which is also superb!
You mentioned at the start of the video that this is a 24-bit analyzer, but the ADC seems to be 32-bit. Is that because there’s is some degradation in performance on other parts of the system, or is this supposed to be rated as a 32-bit analyzer?
Thanks again for the amazing video!
I can try to answer. If you assume a +1V range, then 1/2^24 is a 50nV step! 1/2^32 is a 0.2nV step. These voltages are so small that just temperature differences across an IC can lead to massive swings in value compared to the least significant bit. So while the ADC will give you 32bits, preserving information from far away from the ADC into the ADC chip (or even within the ADC) is extremely challenging.
Your channel is such a great way to unwind and relax.
More details are the salt of a video!