Mine showed up today and the extra voltage is just what I wanted, thanks again for the review. Subjectively it sounds the same as the Apple dongle with about half the gain on the volume knob.
-
Welcome to ASR. 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!
(Unofficial) Review of JCALLY JM20 MAX: USB-C Headphone Dongle with High Unbalanced Output Power
- Joined
- May 15, 2019
- Messages
- 1,033
- Likes
- 4,331
I am curious - did you ever measure what 15% on Windows volume slider really means? For example in AIMP volume slider the 15% give around -16 dB (unless you activate the log option), which matches good enough. But with the volume slider on the right of the taskbar in Windows 11 set to 15% I get -43 dB (aka real 0.7%). IMHO any such % values indicate sloppy implementations and are quite useless.
- Thread Starter
- #143
I agree. I also learned that there's no fixed correspondence b/w Windows volume % and different USB audio devices, meaning the actual dB level of Windows volume % depends on each device. When I get a chance, I will measure the JM20 MAX's output levels corresponding to Windows % values.
And of course, I knew this when I wrote the post. There I simply meant the digital volume control was set to a very low level for some sensitive headphones. I did not mean % values literally.
Last edited:
Not really sure, but I believe Windows volume percentage ‘curve’ is based on: 100% -0 dB; 52% -10 dB (1/10 power, ~half perceived loudness); 27% -20 dB (1/100 power); 14% -30 dB (1/1000 power)
Of course, Windows has no idea of the actual volume—that’s up to the device implementation.
unpluggged
Major Contributor
- Joined
- Jan 5, 2023
- Messages
- 1,134
- Likes
- 1,420
You can select between percentage and dB display in Windows Sound control panel window (at least in WIndows 10).
According to the specification sheet on page 34, Class H is implemented as follows when switching to the next higher voltage supply:
The "high dv/dt transient" may clip the outputs at the current (lower) voltage supply, and my uneducated guess is that this "transitory clipping" may be what the "Russian website" (reference-audio-analyzer.pro) shows.
And there is no delay here. The aforementioned 5.5 second delay only occurs when switching to the next lower voltage supply (see page 35).
However, I could be wrong.
nick_l44.1
Member
- Joined
- Jan 11, 2024
- Messages
- 67
- Likes
- 57
I'm pretty sure this combination is not good for producing high quality sound. I have never come across a high-sensitivity headphone among the best quality ones. Active use of the DAC as a preamplifier will surely degrade its performance. So, using the regular amp gain switches is still much better.
You were better off using an APU for this test or using cross-correlations. ADC noise makes it impossible to directly measure accurately the noise of such quiet DUTs. On my prototype 9039s (it is slightly noisier than the production samples) the noise goes down to -125 dB, I gave you this measurement in your thread about SMSL DL200.
In itself your measurement is quite accurate, but not very representative. All recordings contain intrinsic noise, which is higher than quantization noise. If you look at the best microphones, their SNR only goes up to ~90 dBA. I suggest using this mix to measure the CS43131 DR: 1 kHz sine -60 dBFS + 40 kHz sine -1 dBFS + white noise -120 dB. If there are concerns that this noise has affected the result, there is an option to subtract the signal noise mathematically.
Here is the spectrum of the test signal in its pure form without conversions
It is impossible to use notch in the presence of 40 kHz -1 dBFS tone, so I put ADC into mono mode and activate cross-correlations. Checking the setup on a standard DR signal:
The resulting DR is 129.4 dB. So there are no issues with the setup.
Next I feed the mix described above:
Taking into account a slight signal suppression by the lower ADC sensitivity, the noise at the DAC output is -96 dB. The influence of -120 dBFS white noise added by me can't exceed 0.02 dB, I don't see any sense to take it into account further. The reason for such a noisy result is that the white noise added by me blocked the noise-shaping work approximately as it will happen when playing real recordings instead of synthetic test signals.
According to this test we get DR = 36.0 + 60 = 96 dB
or DR = 100.9 - 62.7 + 60 = 98.2 dBA.
Do you still think that the DR = 130 dBA obtained by AES17 rules is not a fake?
Last edited:
Quick question about gain staging the JM20 Max.
When I connect it to my phone, the USB Audio Player Pro app pops up and the hardware volume is at -15dB by default. Is that digital volume inside the CS43131 that's in line with the OS volume or is it separate analog attenuation?
I have the old Earstudio ES100 and that one has a separate analog attenuator. I was wondering how does the internal volume in the JM20 Max work.
When I connect it to my phone, the USB Audio Player Pro app pops up and the hardware volume is at -15dB by default. Is that digital volume inside the CS43131 that's in line with the OS volume or is it separate analog attenuation?
I have the old Earstudio ES100 and that one has a separate analog attenuator. I was wondering how does the internal volume in the JM20 Max work.
- Thread Starter
- #149
Of course, if you want the best possible low-level performance, you can have an alternative solution, as studied here by @Rja4000. But practically, we cannot deny the utility of digital volume control in applications like dongle-type devices.
I examined your suggested method on my bench extensively. Before presenting my results, here is one critical thing we want to consider. Your measurement does not represent a realistic situation given the fact that CS43131 is very sensitive to strong ultrasonic signals. The higher the signal frequency, the more sensitively it responds---see my measurements below showing something not obvious in your measurements. Yes, I agree that it may be viewed as a design flaw. But the question is, how much does it affect realistic performance? What music/audio content has -1 dB signal at 40,000 Hz?
Ok, here's a baseline FFT of a 1 kHz sine tone @ -60 dB into the JM20 MAX (cross correlation was used in REW):
DR = 127.1 dB
Now, added -120 dB white noise:
DR = 124.1 dB (i.e., the effect of adding -120 dB white noise is about 3 dB reduction in DR)
If the purpose of including a strong ultrasonic signal is to defeat the adaptive noise reduction in CS431xx, then a signal of -6 dB @ 22,000 Hz should suffice. Here is the FFT of -60 dB sine@1 kHz + -120 dB white noise + -6 dB@22 kHz sine:
DR = 114.2 dB
Surely, with no adaptive SNR enhancement in place, DR has decreased. But this decrease is NOT entirely due to the SNR enhancement being turned off. Even if the signal is at -10 dB @22,000 Hz, the SNR enhancement is off. See the FFT with a tone of -10 dB @ 22 kHz substituted:
DR = 116 dB
Only a 8 dB difference now from the case of the SNR enhancement being turned on.
With a -40 dB @ 22 kHz tone substituted, the performance goes back to:
DR = 124.1 dB, i.e., same as the results with no ultrasonic tone, because the SNR enhancement is turned on in this case.
In fact, if you want to make the performance really, really awful, you can substitute a -1 dB @ 50 kHz tone (instead of 40 kHz):
Abysmal DR = 77.6 dB!
So, which represents this dongle's realistic performance? People may judge on their own
I for one have no problem with this dongle in practical use cases.Anyway, this investigation of CS431xx gets really interesting! I am already eager to see Cirrus Logic's next iteration of this small form-factor high-performance DAC/amp combo chipset. Hopefully this ultrasonic anomaly is ironed out in their next iteration. I am sure it is possible if they want to address this issue.
Do you still think that the DR = 130 dBA obtained by AES17 rules is not a fake?
Have you read my clarification of our language use / conceptualization difference?
Quite a chunk of debate on this topic comes from the definition of Dynamic Range and its AES17 measurement procedure. Looking at this topic in better perspective now, I can say that the name we have given to this technique, "Dynamic Range Enhancement" is a misnomer, because DR is defined as the ratio of the loudest, clean signal to the noise floor of a device. If you want to be faithful to this definition, the result of DR measurement of CS431xx using the AES17 procedure (implemented in AP as well) is considered not valid because DRE in the chip tricks the method.
But as you guys agreed, the effect of DRE is real, and if implemented correctly with no audible, adverse side effect, should be beneficial.
So, from now on, I suggest calling this technique "Adaptive Signal-to-Noise Ratio Enhancement" or ASNRE.
So, going forward, there's no need to debate on this particular topic (i.e., whether the AES17 DR measurement of CS431xx is valid or not) since my focus is not on that.
Last edited:
- Thread Starter
- #150
Thanks for adding your input. I am (I'm sure also @nick_l44.1 is) aware of this information on the datasheet. But according to the test results described by the reference-audio-analyzer website, the crunchy behavior observed there does not seem to be due to this time-limited transition (in either direction). It's just illogical to explain it that way.
EDIT. But as I said here, this does NOT necessarily mean that the Class H mode is not responsible for that behavior. It simply does not seem to be due to this "transition" process as is described on the datasheet. As you know, a Class H amplifier by construction modulates the supply voltage constantly. Some particular signals can trick the logic behind it so clipping may occur.
Last edited:
- Thread Starter
- #151
There's no analog volume control supported in the CS43131. I am not a user of the UAPP but heard that UAPP can access a device's internal volume control. I do not know if there's any difference in noise performance b/w the OS volume attenuation and the DAC's digital attenuation. Maybe @CedarX can answer?
Last edited:
nick_l44.1
Member
- Joined
- Jan 11, 2024
- Messages
- 67
- Likes
- 57
I think it's too early to draw conclusions. I assumed that the white noise would block the shaping from working, but it looks like that's not the case. Even before your post I was unable to repeat the same experiment for the THD+N@1kHz test.
Oh, sorry! I was incorrect, you didn't say exactly that. Just in my opinion the results of DR AES17 and DRE performance are very closely related. The question was purely rhetorical. The most interesting task is to find a way to cheat noise-shaping.
- Thread Starter
- #153
Sure, no simple tests can uncover the entire picture of what is happening in CS431xx. I am beginning to think it is not worth doing this painful 'reverse engineering'---there's simply no obvious way to fully test it without the original designer's level of knowledge. There are potentially multiple factors intertwined: SNR Enhancement, Class H operation, and noise shaping. Only very few people like us must be interested. Most (even technical) people will just look at the standard set of tests (like what Amir does) and subjectively evaluate its sound quality.
I have no problem concurring with this part, which is obvious even in my measurements posted much earlier.
Last edited:
The developer of Neutron V1 dongle (based on ES9219) just shared some thoughts about these CS dacs "issues" here:
audiosciencereview.com
i honestly can't follow on some points due to lack of technical knowledge, still the discussion is intriguing, it would fit well into the CS43131 dedicated thread
audiosciencereview.com
Neutron HiFi DAC V1 Review
but if they've been measured here on ASR and they measure well then you're likely just imagining the differences Unless USB DAC is based on nowadays popular Cirrus Logic DAC chip (CS43198, CS43131,...) with enforced Dynamic Range Enhancement (DRE) technique (using ESS name for it here) which...
audiosciencereview.com
i honestly can't follow on some points due to lack of technical knowledge, still the discussion is intriguing, it would fit well into the CS43131 dedicated thread
A Thread Dedicated to Cirrus Logic CS43131
The USB-C Dongle Market Apparently CS43130 was a Wolfson chip. It was Hidizs / Tempotec / Meizu that first drew my attention to CS43131. Later in 2020 products with dual CS43131 such as iBasso DC03 began appearing. We’ve had these pages...
audiosciencereview.com
nick_l44.1
Member
- Joined
- Jan 11, 2024
- Messages
- 67
- Likes
- 57
Take a look.
It seems indeed high-level ultrasound blocks noise-shaping due to the risk of clipping at the tone frequency. The tone at 83 kHz has the greatest effect on Awt noise, although at this frequency the effect of the DAC LPF is already noticeable. The noise [20..20000 Hz] turns out to be a really huge 59 dB/61 dBA. So, noise-shaping lowers the noise in the audible region by about 55 dB or some other effect happens. From a technical point of view this is amazing, but from a quality conversion point of view it is very questionable.
Actually, the inner workings of the chip are not very interesting. Many patterns we have figured out. But it is a non-trivial task to estimate from the data obtained how acceptable such use of technologies is.
FFT of the same tone from 9039s
The noise [20..20000 Hz] is -126 dB/-128 dBA
- Thread Starter
- #156
More and more stores at AliExpress have stopped shipping products to the US.
As of now it appears that only one store sells the JM20 Max to the US for $40.
Not a good time to live in the US...
As of now it appears that only one store sells the JM20 Max to the US for $40.
Not a good time to live in the US...
dtaylo1066
Addicted to Fun and Learning
- Joined
- Jan 12, 2019
- Messages
- 796
- Likes
- 1,093
Just got mine via AliExpress. On Zero Reds from Samsung Galaxy Tab 8+ usb out it sounds wonderful. 2/3 on the volume bar is about all my ears can take. This thing has some power. Playing from Tidal. Have to load UAPP to get bit perfect and circumnavigate Andoid 48kHz sampling rate, though it is not a audible issue at present. On first quick listen it can certainly give my my Micro iDSD Black Label. For $25 that is astounding.
Yes, I bolted in an order right after the review was posted when pricing was still low. No issues. One week later might be a new world.
Yes, I bolted in an order right after the review was posted when pricing was still low. No issues. One week later might be a new world.
Thinking back about this, i just remembered the very bad findings by L7audiolab on 2 old Tempotec implementation of the CS43131, the Sonata E35 and E44Exactly. The pattern that looked like an early rise in THD+N is mainly due to the DRE being phased out gradually (when the output level increases):
E35
E44 even worse
Now i wonder, if this was also due to DRE, it seems a very bad implementation of it, then some registers on the chip must exist where you can manage DRE behavior, and here they messed them completely.
But I could be totally off path.
- Thread Starter
- #159
Without knowing the exact measurement settings or the Temptec's design surrounding the DAC chip, it is difficult to give a definite answer. But one thing clear is, in these stepped THD+N tests, the DRE or adaptive noise shaper in CS43131 cannot make THD+N increase without contribution from distortion. The worst case would be flat THD+N relative to fundamental tones. The rising THD+N must be due to rapidly rising THD (i.e., not due to the N part). I suspect there must be some design flaw in its power supply circuit.Thinking back about this, i just remembered the very bad findings by L7audiolab on 2 old Tempotec implementation of the CS43131, the Sonata E35 and E44
E35
E44 even worse
Now i wonder, if this was also due to DRE, it seems a very bad implementation of it, then some registers on the chip must exist where you can manage DRE behavior, and here they messed them completely.
But I could be totally off path.
Well, I guess it's the case, those dongle probably had flaws in powers supply or firmware, as said my technical knowledge is limited and my main goal is to put on the table as much as possibile additional information that could help drawing the picture
Similar threads
- Poll
- Poll
