Build thread for Diyinhk ES9018 DAC on Ebay

Status
Not open for further replies.
Hi,

The xo is okay but will restrict the performance; the lme49710 are fine in the buffer; but unfortunately, they don't sound good in the I/V so just about anything will be better. What can you get for reasonable money that is fet input ?

Opa134 are good, opa132 are the upgrade from 134, opa1641 are smooth and warm, opa827 are good and opa627 and ada4627 sound best imo. lt1028 are too bassy, ne5534 are a little dull. Opa602 and ad8597 are okay. Probably ad8610 will sound good (limited to +/-12V and now replaced by ada4627). Any others you can get ?

cheers
 
Last edited:
KK have you tried AD8065 for I/V? - they're very cheap on Taobao (1.9rmb I paid). I found them a little coloured at the top end as input buffer on my amp, not particularly unpleasant. Haven't tried in I/V as prefer something more neutral but as you're a fan of FETs I wondered if you'd listened? Dynamic specs do look good for I/V.

On eBay, its about $5 a pop for AD8065AR.
 
Hi,

I haven't tried them - they're very fast but they didn't make my wishlist 'cos of noise. I used the parametric searches and set noise limit to 5nV/Hz, already quite noisy. The only ones left on my ad wishlist are ada4898; not fet but worth a try.

BTW, I'm not fan of any particular design - it's just the strong output from this DAC seems to want something that can behave very linearly and fets seem to have stolen the show for this.
 
Last edited:
Thanks.

I don't really understand why noise is so important with DACs - if I'm building a phono preamp then sure, noise is potentially a problem because the signal is at most 5mV say with a MM cartridge. But with a DAC the signal's around 2V, so that's about 50dB higher. By my handwaving reckoning if 0.3nV/rtHz is the bees knees for MC phono then adding 50dB to this gets us close to 100nV/rtHz. Sure, we'd like to do a bit better than phono with digital but do we really need that much less noise?

Yes I agree with you 100% about linearity being king and FETs being better at HF linearity. However in my experience there's a negative correlation between linearity and noise - by sacrificing some noise we can gain linearity. This seems to be because when an LTP is involved (as it is with most opamps) degeneration buys us more linearity at the expense of noise. For DAC duty, I want maximum linearity hence higher degeneration - this comes with higher noise.
 
Last edited:
I agree to some extent. I'm using class-T amps, bi-amping, and relative to classA, and this DAC, they're noisy as hell. But I seem to get a better soundstage with lower noise - more detail and better placed. Christopher Cross, Sailing, has some nice chimes on it that flow across the room better. Stuff like that....

I'm think about replacing the I/V op amps with a very simple CFP - two transistors that make a unity gain stage. What do you think ? I'm sure someone must already have done this and a quick search gives me this :

http://www.diyaudio.com/forums/solid-state/2298-fet-compound-pair.html#post18038
 
You mean doing I/V passively then buffering it? My money's still on LC filtering before any active stages - once you've tried that, you won't go back. The soundstage improvement was the biggest shock when I first heard it :) Once there's a passive filter there, the active stage becomes much less critical.
 
Hi,

I wasn't thinking passive - I would want to avoid coils unless they are tiny air-core jobs. But just for arguments' sake, I could use the 4 air-core output coils (10uH) I have from an old Charlize TA2020 amp that burned, and add some caps to make a filter ?

I was thinking two fets connected as a complementary feedback pair, or Szilaki, with a constant current. Here's a description :

http://www.epanorama.net/sff/Misc/Amplifiers/Complementary_Feedback_Pair.pdf

Anyway, I'm not going to do it for this DAC. I'm very pleased with the OPA627 and ADA4627 so I'm going to leave this DAC alone and "work" on the two others - AK4399 and dual AD1955.

cheers
 
10uH is a bit on the low side as regards the (too high) corner frequency it will create, but its better than nothing yep.

What's the idea behind using two FETs in a CFP? Because you don't think a single FET is linear enough? I don't have any experience to speak of with JFETs so couldn't comment. As regards bipolars in this configuration I'm not sure that the linearity is gained in the area where we need it - high frequencies and low level signals. AP distortion plots tend to measure lower freq linearity and high levels.
 
Last edited:
As I understand it, CFP with fet should have lower output impedance than single fet so it'll drive the bjt buffer better.

I have some Mundorf silver/oil caps also sitting around: 2x 6.8uF and 2x 8.2uF - big expensive caps that were doing duty as tweeter protection caps when I first built this hifi. I could give that a try... maybe cut too much... reduce the coil size..
 
Last edited:

TNT

Member
Joined 2003
Paid Member
You mean doing I/V passively then buffering it? My money's still on LC filtering before any active stages - once you've tried that, you won't go back. The soundstage improvement was the biggest shock when I first heard it :) Once there's a passive filter there, the active stage becomes much less critical.

I would be very interested in seeing a bid for an diff. output stage for 9012/18 from you Abrax as your ides are different from what is usually advertised - if you don't mind?

/
 
Last edited:
Hello. Recently finished my diyinhk board.
Power supply: 2 x LT1763 for 3.3V VA, VD and onboard ADP151 for 1.2V VD. I/V conversion with AD797 (powered by 2 x LM317). I use balanced output with analog filter from reference board ESS schematics.
Connected to a PC with CM6631 USB-I2S also diyinhk. x49 clock signal taken from this board too.
Noticed big impact of VA & VD regulator's output baypass capacitors on harmonic distortion.
Measured with RMAA & EMU0202
(white, blue green - VD is only bypassed by 10uF ceramic, VA - as described on graph;
magenta - VD bypassed with 3300uF)

Some close-ups:
An externally hosted image should be here but it was not working when we last tested it.


and the whole picture:
An externally hosted image should be here but it was not working when we last tested it.


The differences look drastic on the graphs, but sonically I couldn't hear them at all :)
It would be great to see some other people's measurements of this board and generally of ES9018.
 
I'm using multiple resistors because my DAC is a DAC array - individual DACs have their contributions summed together in the I/V resistors. The total (sum) of the I/V resistors is about 50R. The gain stage is AD605 though I'm currently looking at AD8129 in this role for the next design. The LPF I'm using is a 7th order elliptic using ferrite cored inductors and NP0 caps.
 
Did some more testing of power supply and output stage.
- Is I/V stage necessary, and how does it improve sound?
- Is the voltage regulator noise a really big issue?
- Do I really need separate power supplies for 3.3V VD and VA?
I tried to answer these questions myself. So I powered the whole circuit from single 3.3v supply based on LM317 (ADP151 for 1.2V). Took signal directly from ES9018 outputs, so it works in "voltage mode". Here is the graph comparing such setup with "fully equipped" circuit with I/V output stage (AD797) and separate low noise regulators (LT1763)

An externally hosted image should be here but it was not working when we last tested it.


An externally hosted image should be here but it was not working when we last tested it.


As you can see, there is not much difference. Single supply causes some garbage to appear on the high frequency side, but the spikes are very low. Usage of LM317 risen the noise a tiny bit, also at upper freq. There is significant change in the level of 3rd harmonics. The change in first is moderate and may be due to change in input gain of EMU0202, because level of ES9018 Vout is lower then Iout, and I had to compensate.

Conclusion: simply make one 3.3V voltage reg based on anything, hook ES9018 output directly (or thru some caps) to your amp and enjoy the music. This DAC is excellent. Other DAC's require much more effort from diyer to perform so well.
 
hehe, your emu does not have the capability to measure the difference effectively, since the THD performance of ESS far outstrips ADC performance of the EMU. connecting directly, DNR will be similar enough, but THD and dynamic performance will be considerably worse. none of the regulators you tried have truly low noise for AVCC, none are suitable for getting best performance of ES9018.

your conclusion is based on false premise, not the best vantage point to be giving advice.

this has all been undertaken and measured with AP2, the differences are in the range of 15-20dB for direct vs a well made IV stage with decent power supplies for AVCC
 
Last edited:
hehe, your emu does not have the capability to measure the difference effectively, since the THD performance of ESS far outstrips ADC performance of the EMU.
Yes, you are right, but in my experiment I tried to make ESS perform worse, not better. And it happend, but not to such extent as i thought. I can easily observe, that 2nd and 3rd harmonics I measured are almost the EMU adc distortion, but disapperance of higher order harmonics, when I applied bigger regulator bypass caps was not my measurement equipment fault. Also the noise floor of the EMU is still lower than ES9018:

An externally hosted image should be here but it was not working when we last tested it.


connecting directly, DNR will be similar enough, but THD and dynamic performance will be considerably worse. none of the regulators you tried have truly low noise for AVCC, none are suitable for getting best performance of ES9018.
LM317 has around 300nV/sqrtHz noise vs 20nV/sqrtHz noise of LT1763. Difference in noise level of ES9018 powered from this two regulators was marginal (2dB in upper frq region). So I doubt, that building sophisticated AVCC power supply for this diyinhk implementation of ES9018 is worth the effort (anyway I'm going to do this and prove my point of view).

your conclusion is based on false premise, not the best vantage point to be giving advice.
Maybe I was unclear, so let me say it again: If you are going to listen to music, not to build a piece of laboratory equipment, than even the humble way of using diyinhk ES9018 board would let you enjoy it without compromising the sound. Distortions always stay below 0,01%

this has all been undertaken and measured with AP2, the differences are in the range of 15-20dB for direct vs a well made IV stage with decent power supplies for AVCC
Did they measured diyinhk board, conected to a PC with diyinhk USB converter and decent power supplies were diy made? What Audio Precision unit are you referring to? ATS-2? I would like to see these results.
 
Status
Not open for further replies.