E1DA 9038D Review (portable DAC & Amp)

[staticV3]

So you want me to put a filter on the output of a class of DACs to make them measure better but not do it for others which don't emit too much ultrasonic noise?

If there is excess ultrasonic noise, then the device may fail regulatory testing as well.

So no, I can't make a special case for testing this device. Plenty of other devices generate out of band noise which impact measurements. In this case, that impact is nothing compared to 0 dBFS saturation for which we still don't have an answer. I was hoping by now we would have an investigation of that and hopefully with a firmware update, would be behind us. Instead, we seemed to be wrapped around axil on minutia and theory with no consequence at all.

No, it's fine. I didn't ask you to make an exception and neither do I want you to.
I'm just trying to understand which parts of your results show the DUT and which show the lab equipment and its limits.

 

[NM156]

not sure what about you said .5mV, THD+N vs Power sweep I didn't make for that sample, only standard tests were performed(DR, [email protected]). I make such tests for each unit and calibrate THD compensation registers as well.

@IVX @amirm
OK, let me elaborate on that.

FFT: channel 2 (blue) is visibly higher than channel 1 (red) without additional or higher harmonics => ch2 seems to have higher noise than ch1
SINAD: ch1 = 111.481dB, ch2 = 110.255dB => 1.2dB imbalance (due to noise?)
Dynamic range: ch1 = 117.165dB, ch2 = 115.583dB => 1.6dB imbalance (due to noise?)
Dynamic range at 50mV: ch1 = 82.230dB, ch2 = 80.765dB => 1.5dB imbalance (due to noise?)
SINAD vs. output voltage sweep: ch2 (now red) is visibly lower than ch1 (now blue) up to clipping (due to noise?)
Intermodulation distortion sweep: no imbalance
Linearity sweep: no imbalance
Jitter: unknown since only one channel is measured
Multitone: unknown since only one channel is measured
THD+N vs. output power sweep with 300 Ohm load: ch2 is visibly higher than ch1 up to clipping and there is a very visible "hump" below 0.5 mW which should never ever be there (an anomaly)

Ch2 level = 2.67V, ch1 level = 2.665V => ch2 only 0.016dB higher than ch1 => no explanation for the imbalances (and especially not for the "hump")

Maybe (just maybe) Amir's sample you sent him is broken?
Or did Amir somehow make a measurement mistake (which I doubt very much)?
Otherwise: why the imbalances (more noise) and why the anomaly?

I'm very interested in this device but the imbalances (and especially the "hump") keep me from buying it.

 

[NM156]

I'm just trying to understand which parts of your results show the DUT and which show the lab equipment and its limits.

Yes, that's exactly what I would like to understand, too.

 

[NM156]

@NM156 there's your answer. Channel imbalance especially in THD+N vs output power is apparently caused by the measuring equipment.

@staticV3 What is your opinion on the "hump"?

 

[staticV3]

For people thinking of using this with smartphone it's worth noting two really important drawbacks not touched on in the review:

Power consumption is HUGE for a dongle. At idle it consumes 95 mA, and in use hundreds. This *will* have a very obvious effect on battery life. Most dongles designed for use with phones achieve standby power consumption a third or half that, and correspondingly lower consumption in use.

Noise from LTE: the dongle becomes noisy with 4G & 5G.

This looks like a great product for use with a laptop or PC but impractical or disappointing for pairing with a phone.

Here's a detailed power consumption chart:

Measuring current draw over 200mA would only be possible when playing DSD256 or something while completely maxing out the headphone Amp (i.e.constant, full-scale sine wave) under high load (<32Ω) and yet you said you measured 400-500mA? Are you sure about that?

 

[staticV3]

@staticV3 What is your opinion on the "hump"?

It doesn't look good.
What IVX said seems plausible, but really, I'm the wrong person to ask here. Sorry.

 

[IVX]

Veri, not so simple yet, needs to take into account the tolerances of reference voltage regulators(2pcs), they affect the final gain/clipping conditions.
Take a look at DF cobalt, it has THD+N .01% in the best case, the same as my bad-printed unit at 0dbfs

 

[IVX]

NM156, I can't say why, it wasn't my measurement. I did measure the sample before sending that(no THD+N vs level sweep yet), and I sent my results to Amir:
#9038D ES9038Q2M based 3.5mm jack unbalanced DAC+HPA(Susumu 0.1% resistors and 4000uF USB 5V rail-filter) 200mW@40ohm@THD+N = 1%, THD+N L/R 0.000125% at -0.5dbfs no-load or 0.00017% at -2db 32ohm load, Dynamic Range AES17 125/125.4db(A-weighted),
As you can see, this sample a bit better vs the bad-printed one that I measured today.

 

[capslock]

Well, i got one and also using the pro app. No hiss with my sensitive IEMs, sounds great and lots of functionality. The tube emulation is fun.

How are people setting the volumes? Cant work out if its better to lower the hardware volume and turn up my android phone volume or vice versa?

Not sure how the android stack does it's math, but if it is 16 bits, then doing it in android is not good, doing it inside the dongle will be infintely better, no matter whether it is digital or analog attenuation.

The description of the 9038SG3 says it has a new USB audio bridge with HW control. What does this mean, digital attenuation in the bridge or an analog attenuator after the DAC? Will the android volume control tell the USB bridge to attenuate?


Back to the question of output levels and residual ultrasonic noise:
- I think setting attenuation to deal with worst case load impedance and supply voltage is a good idea. Most of us don't listen at full volume with a headphone, and analog input stages might not like too hot a signal anyway.
- Amir, I think any preamp or amp with an analog input stage will have a low pass before the first amplifier. If the AP doesn't filter until after the first amplifier IC, a small amount of HF might get rectified, especially if is an older design that uses older op amps that were designed before cell phones and SMPS were ubiquitous. So it might be a good idea to use an input filter on the AP on anything you measure simply because the enviroment that it is operated in has changed. HF may cause rectified signals at the - 115 dB level long before it becomes a regulatory issue.

 

[staticV3]

Not sure how the android stack does it's math, but if it is 16 bits, then doing it in android is not good, doing it inside the dongle will be infintely better, no matter whether it is digital or analog attenuation.

The description of the 9038SG3 says it has a new USB audio bridge with HW control. What does this mean, digital attenuation in the bridge or an analog attenuator after the DAC? Will the android volume control tell the USB bridge to attenuate?


Back to the question of output levels and residual ultrasonic noise:
- I think setting attenuation to deal with worst case load impedance and supply voltage is a good idea. Most of us don't listen at full volume with a headphone, and analog input stages might not like too hot a signal anyway.
- Amir, I think any preamp or amp with an analog input stage will have a low pass before the first amplifier. If the AP doesn't filter until after the first amplifier IC, a small amount of HF might get rectified, especially if is an older design that uses older op amps that were designed before cell phones and SMPS were ubiquitous. So it might be a good idea to use an input filter on the AP on anything you measure simply because the enviroment that it is operated in has changed. HF may cause rectified signals at the - 115 dB level long before it becomes a regulatory issue.

Hardware volume inside the 9038SG3 is done digitally. After all, there's just no way to accommodate multiple gain stages inside the tiny enclosure and switches or pots aren't quite IVX's style afaik.
When you plug any DAC with UAC2-controlled hardware volume into an Android device, then the OS will reset its hardware volume to 100%, however, you can configure that 100% using the Tweak9038 App's maximum volume limiter.
Beyond that one action, Android has no control over a UAC2-enabled DAC's hardware volume.
If you want to control it from your phone, then you need to use USB Audio Player Pro, Hiby Music Player, or Tweak9038. All of which will bypass the Android Audio environment completely and will control the DAC directly.

 

[Jimbob54]

Well, i got one and also using the pro app. No hiss with my sensitive IEMs, sounds great and lots of functionality. The tube emulation is fun.

How are people setting the volumes? Cant work out if its better to lower the hardware volume and turn up my android phone volume or vice versa?

I would recommend getting USB Audio Player Pro for Android- especially if you are Tidal or Qobuz or local files user (doesnt support Spotify). There is a small fee for the full version but it integrates (all) USB DACs brilliantly- you can set it to control the hardware vol on the device with small increments using the phone buttons - or use software volume with the same buttons. Avoids the horrid Android default resampling of everything too.

Personally, I cant tell any difference using hardware or software control. I think in your set up I would set the hardware vol to the absolute max your ears can take with software vol maxed, then back the software down to comfortable and you then have a bit of headroom in the phone controls

 

[Tks]

Not sure how the android stack does it's math, but if it is 16 bits, then doing it in android is not good, doing it inside the dongle will be infintely better, no matter whether it is digital or analog attenuation.

The description of the 9038SG3 says it has a new USB audio bridge with HW control. What does this mean, digital attenuation in the bridge or an analog attenuator after the DAC? Will the android volume control tell the USB bridge to attenuate?


Back to the question of output levels and residual ultrasonic noise:
- I think setting attenuation to deal with worst case load impedance and supply voltage is a good idea. Most of us don't listen at full volume with a headphone, and analog input stages might not like too hot a signal anyway.
- Amir, I think any preamp or amp with an analog input stage will have a low pass before the first amplifier. If the AP doesn't filter until after the first amplifier IC, a small amount of HF might get rectified, especially if is an older design that uses older op amps that were designed before cell phones and SMPS were ubiquitous. So it might be a good idea to use an input filter on the AP on anything you measure simply because the enviroment that it is operated in has changed. HF may cause rectified signals at the - 115 dB level long before it becomes a regulatory issue.

Yeah I agree with this, I never really run any DAC+AMP all-in-one device at those levels. Setting a limit out the box (for anyone who isn't a tinkerer that buys the dongle for the insanely high power output).

 

[IVX]

capslock, any USB host that supports USB UAC2(W10, Android, Mac os, Linux, iOS) will send UAC2 volume commands when you move the volume slider on your GUI. USB bridge will receive these commands and send them next to the ES9038Q2M. ESS doesn't tell us how they control the internal DACs volume but I'm sure it is a trivial 32-bit divider.

 

[BDWoody]

Defined as the weight of a 357 bullet.
View attachment 119211

.38 Special would have worked with the name. Maybe he should offer a # 90357M version as a high power option.

 

[bobbooo]

@NM156 there's your answer. Channel imbalance especially in THD+N vs output power is apparently caused by the measuring equipment.

Again, looks like the Qudelix 5K measurements might show a similar issue.

 

[NM156]

It doesn't look good.
What IVX said seems plausible, but really, I'm the wrong person to ask here. Sorry.

Is it plausible that the SMPS causes such an anomaly on only one channel?
I'm not an audio engineer, but frankly, well, really?

 

[NM156]

NM156, I can't say why, it wasn't my measurement. I did measure the sample before sending that(no THD+N vs level sweep yet), and I sent my results to Amir:
#9038D ES9038Q2M based 3.5mm jack unbalanced DAC+HPA(Susumu 0.1% resistors and 4000uF USB 5V rail-filter) 200mW@40ohm@THD+N = 1%, THD+N L/R 0.000125% at -0.5dbfs no-load or 0.00017% at -2db 32ohm load, Dynamic Range AES17 125/125.4db(A-weighted),
As you can see, this sample a bit better vs the bad-printed one that I measured today.

@IVX OK, what if you make a THD+N vs. output power sweep at 300 Ohm on both channels with your AP?
It should not show that anomaly on any channel. Otherwise your device may have a design flaw (this is not against you personally!).

 

[NM156]

If there is excess ultrasonic noise, then the device may fail regulatory testing as well.

@amirm I think this is an important point.
Could you please measure for ultrasonic noise on both channels? I think your AP can measure up to 1MHz (at least), right?

 

[staticV3]

@IVX OK, what if you make a THD+N vs. output power sweep at 300 Ohm on both channels with your AP?
It should not show that anomaly on any channel. Otherwise your device may have a design flaw (this is not against you personally!).

he already did, here: https://www.audiosciencereview.com/...iew-portable-dac-amp.21556/page-8#post-724238
It's 300 Ohm, so not stictly "no load", but close enough I'd say.

 

[staticV3]

@IVX if you have the time, could you please do a THD+N vs output level sweep "the wrong way", so without the low pass filter, just to see if that will result in similar measurements as Amir got?