• 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!

E1DA Cosmos ADC teardown

IVX

Major Contributor
Joined
Mar 3, 2019
Messages
1,436
Likes
2,828
Location
South of China, SHZ area, - Слава Україні
Well, I guess that swapping the polarised caps with non-polarised type, like anything between 6.8...10uF MKP will dramatically increase the max. DC input voltage range, right? However, size and price would increase a lot, so not sure it worth it, unless it's really-really required.
220uF gave -3db at 4-5Hz, 10uF will set that to the 90Hz if that's fine, you can try.
 
OP
trl

trl

Major Contributor
Forum Donor
King of Mods
Joined
Feb 28, 2018
Messages
1,986
Likes
2,575
Location
Iasi, RO
Thanks for the reply, you said on the other page "220uF/10V" while I thought these caps are 10uF.
 

capslock

Senior Member
Joined
Jun 19, 2020
Messages
364
Likes
184
This kind of lays to rest the notion that polarized caps are not suitable for audio, doesn't it? Or can we think of test that looks at the distortion introduced by these caps at say 20 Hz? I realize this is not relevant because we just cannot hear distortion very well at low frequency.
 
OP
trl

trl

Major Contributor
Forum Donor
King of Mods
Joined
Feb 28, 2018
Messages
1,986
Likes
2,575
Location
Iasi, RO
There are no distortions created by these caps, you're probably referring to freq. response (HPF), which in this case is "220uF gave -3db at 4-5Hz", per Ivan's wrote above.
 

capslock

Senior Member
Joined
Jun 19, 2020
Messages
364
Likes
184
Oh, I'm referring to distortion. I remember reading an article (maybe by Scott Wurcer) looking at cap distortion in filters. Main takeaway for me was that distortions are negligible when there is essentially no voltage across the cap, i.e. well into its passband, but became significant especially for electrolytic caps near corner frequency or in stop band.

For passive loudspeaker crossovers this spells trouble. For simple DC blocking filters, this is another reason to put the corner frequency as low as possible.

I think recent testing of blocking filters here has shown that electrolytics do not cause any distortion down to -140 dB or beyond in the passband. But what about near corner frequency, e.g. at 20 Hz?

I would also be wary of slower chemical reactions in electrolytics. What if a sudden burst signal triggers a change and a slow recovery for the subsequent couple of seconds? On the other hand, replacing 220 µF electrolytics with 10 µF film for a 90 Hz corner frequency might be ill advised because it would open the door for dielectric effects in the film caps in the range up to maybe 500 Hz.
 
Last edited:
OP
trl

trl

Major Contributor
Forum Donor
King of Mods
Joined
Feb 28, 2018
Messages
1,986
Likes
2,575
Location
Iasi, RO
THDN-Ratio.jpg

Source: https://www.l7audiolab.com/f/e1da-cosmos-adc/
I invite @WolfX-700 and @IVX to comment the above graph, but I would say that input caps are not the cause for the increased THD+N on the low end.
 

axismundi

Member
Joined
Dec 2, 2019
Messages
16
Likes
9
I invite @WolfX-700 and @IVX to comment the above graph, but I would say that input caps are not the cause for the increased THD+N on the low end.
I confirm, it is not due to input caps (sadly). I bypassed them and got the same performance as with them in the circuit.

One potential culprit for the reduced THD+N performance at low frequencies may be the VREF circuit (see this post for example https://audiosciencereview.com/foru...est-spec-adc-chip-currently.13469/post-935248 ). However, based on some limited brute force experiments I've done (basically adding capacitance to VREF pin) I saw no improvement. I'm still not ruling this out, but if there is an issue with VREF circuit, it is more subtle (too high output impedance for the regulator?...) and will require more tinkering for when I'll have more time.

Also, I have played with the THD and noise compensation parameters at lower frequencies, but to no avail - I was not able to get better performance.

One slight improvement came after updating the firmware for the version with less DC offset (https://audiosciencereview.com/forum/index.php?threads/e1da-cosmos-adc.27038/post-962020), bizarrely enough.

All in all, this is the biggest complaint I have for an otherwise excellent item, it's really a pity. Low input impedance is not great either, but at least this was clearly disclosed from the beginning and can be overcome quite easily by adding a buffer.
 

capslock

Senior Member
Joined
Jun 19, 2020
Messages
364
Likes
184
Interestingly, the AK5394A (of RTX6001 and emu 1616m fame) likes a big cap to buffer its voltage reference. There is some talk at diyaudio about improving emu performance by using a low esr 1000 µF cap.
 
OP
trl

trl

Major Contributor
Forum Donor
King of Mods
Joined
Feb 28, 2018
Messages
1,986
Likes
2,575
Location
Iasi, RO
@IVX might know better if adding caps across VRef will improve THD+N between 20...100 Hz. Thanks!
 

noel_fs

Active Member
Joined
Dec 12, 2018
Messages
271
Likes
182
Yes. This would require a line-level input, and could work with a standard setting for level (providing a bit of headroom for later normalization).

Rick "who currently uses a Benchmark ADC for that purpose, but it's a bit overkill just for that" Denney
do not normalize rip vinyl rips please, completely ruins them.
 

rdenney

Major Contributor
Forum Donor
Joined
Dec 30, 2020
Messages
2,317
Likes
4,089
do not normalize rip vinyl rips please, completely ruins them.
Huh?

All normalizing does is shift the entire file so that the peaks are some defined distance from 0dB.

The dynamics are not changed at all. I can detect no difference when comparing them other than having to adjust the level.

I normalize whole albums at one time, not track by track.

The dynamic range of the digital recording is at least 20dB greater than an LP. That headroom means not clipping it compressing the recording even by accident, which really does ruin the recording.

Rick “who compared carefully” Denney
 

noel_fs

Active Member
Joined
Dec 12, 2018
Messages
271
Likes
182
Oh you meant volume normalizing. I guess thats fine but still not a big fan of it, depending on some things i think it can slightly compress dynamics in the treble range. Atleast from my exprience with foobars replaygain, which is basically what you are talking about im asssuming (EBU R128). Anyway, those peaks are only an approximation, analogically it can peak differently. Also the replaygain itself can can remove clipping effectively compressing it, its something really small not very noticeable but still.
 

ald0s

New Member
Joined
Apr 11, 2022
Messages
2
Likes
0
Can this device be used (sensibly) as an ADC frontend for mic preamps? I've been searching high and low for a portable ADC to connect my Useful Instruments Hornet to, and I don't really want to get something like a UCX II just for this.
 

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,736
Likes
11,045
Location
Prague
I invite @WolfX-700 and @IVX to comment the above graph, but I would say that input caps are not the cause for the increased THD+N on the low end.

A pity that the tester did not use finer step in the stepped sine frequency test. Not enough measuring points and not enough low frequency resolution there, so it is difficult to comment.
 
Top Bottom