ES9018K2M, ES9028Q2M, 9038Q2M DSD/I2S DAC HATs for Raspberry Pi

Ian

Just found this thread

I have been using LL1545A for a quite sometime now and I can confirm that 1:4 Amorphous core with a Sabre loaded with an impedance slightly less than the output impedance of the DAC sounds the best. I have tried tubes, opamps, direct out etc.

The one thing that I found better than the cap connected to the ground to sink asymmetrical noise, is a cap across the legs.

For my Buffalo II dad (ES9018) - I have an effective impedance of 303 ohms across the legs shunted by a silver mica 47nf creating a 3db at 20Khz.

So in my implementation I chose to not use the central tap the transformer at all. I have a single 390 ohm wire wound resistor across the legs (303 ohms with the reflected secondary impedance from the 22k preamp pot i.e. reflected is 22k/16 with a 1:4 gain)

Best probably is to have both caps. The cap in the center like you have plus a silver mica or a really good shunt poly cap across the legs resulting in a 20K-22k 3db pt (depending on your tweeter). I haven't tried yet.

If you have time - you should try it shunting the legs too. That made the most difference to the depth and emotional content of the music. Much more than the cap at the central tap.
 
Ian

Just found this thread

I have been using LL1545A for a quite sometime now and I can confirm that 1:4 Amorphous core with a Sabre loaded with an impedance slightly less than the output impedance of the DAC sounds the best. I have tried tubes, opamps, direct out etc.

The one thing that I found better than the cap connected to the ground to sink asymmetrical noise, is a cap across the legs.

For my Buffalo II dad (ES9018) - I have an effective impedance of 303 ohms across the legs shunted by a silver mica 47nf creating a 3db at 20Khz.

So in my implementation I chose to not use the central tap the transformer at all. I have a single 390 ohm wire wound resistor across the legs (303 ohms with the reflected secondary impedance from the 22k preamp pot i.e. reflected is 22k/16 with a 1:4 gain)

Best probably is to have both caps. The cap in the center like you have plus a silver mica or a really good shunt poly cap across the legs resulting in a 20K-22k 3db pt (depending on your tweeter). I haven't tried yet.

If you have time - you should try it shunting the legs too. That made the most difference to the depth and emotional content of the music. Much more than the cap at the central tap.

Thanks madds1,

That's great. Do you have some details with schematic?

Regards,
Ian
 
At work so I just drew it up to give you an idea.
 

Attachments

  • 116CD4A9-E5FC-40BF-9BB9-2E0961FBD476.jpeg
    116CD4A9-E5FC-40BF-9BB9-2E0961FBD476.jpeg
    647.4 KB · Views: 1,237
The drawing and the my post above say LL1545A. Thats a mu-metal transformer. The one I like better than that one is 1544A which is amorphous. That is what I am using. Apologies for the confusion

As an aside - mu-metal core sounds thicker and has more body than the amorphous cores. Slightly less details
 
At work so I just drew it up to give you an idea.

Thanks madds1, you approach looks nice. What's that 47n cap for? Dose it improve sound quality? what kind of cap did you use?

Both your LL1544A and my LL1674 are amorphous, right?

For ES9018 DAC stereo configuration, the internal resistance is 195 ohm. while ES9028Q2M is 403 ohm, ES9038Q2M and ES9018K2M are around 800ohm. So, the 390 ohm resistor need to be increased I think?

BTW, what kind of resistor did you use? Does it affect sound quality much in your test?

Regards,
Ian
 
Thanks madds1, you approach looks nice. What's that 47n cap for? Dose it improve sound quality? what kind of cap did you use?

Both your LL1544A and my LL1674 are amorphous, right?

For ES9018 DAC stereo configuration, the internal resistance is 195 ohm. while ES9028Q2M is 403 ohm, ES9038Q2M and ES9018K2M are around 800ohm. So, the 390 ohm resistor need to be increased I think?

BTW, what kind of resistor did you use? Does it affect sound quality much in your test?

Regards,
Ian

The 47nf cap creates a corner 3db point of 23K without the preamp secondary impedance reflected to the primaries and 20K with it reflected. But that is not the reason why I have the cap. I discovered the effect of a cap across the legs on sigma delta DACs nearly 3 years ago by accident. I reported in the these posts

http://www.diyaudio.com/forums/digi...native-post-dac-filtering-11.html#post4445352

http://www.diyaudio.com/forums/digi...native-post-dac-filtering-11.html#post4445577

That thread was started by a member who also noticed the same effect. He unfortunately got flamed for reporting it without adequate explanation. The effect is that is creates is a layering of instruments that makes listening to music - emotional.

The oldest reference I can find to this effect is here

http://www.raylectronics.nl/pdfs/Mauros_CD6367_tweak.pdf

Regarding the transformers - yes LL1544A is amorphous too. Nearly $85 in the US vs LL1674 which is $130

The resistor value to be chosen depends on your taste. Put simply - the deeper you put the DAC into current mode (by lowering the resistor) - the THD improves. At 0 ohms - the THD is the best.

But in all honestly - I can't hear THD in my system below -105db.

But I what I can hear is that the slam/attack/dynamics are lost deeper in the current mode.

The optimal value for the resistor is 4/5 times the output impedance of the DAC for my system. So 640 ohms for ES9028Q2M.

I have a modified IRAUDAMP7S driving a highly modded Magnepan 1.7. Preamp is a poorly biased JFET that creates a 2nd order harmonic at nearly -80db. I have been told 2nd harmonics change the tone but I perceive it more as body to the music. LPs naturally had this and Tapes had tons of 3rd harmonics

Sorry for the long reply. Hope this helps
 
Last edited:
The 47nf cap creates a corner 3db point of 23K without the preamp secondary impedance reflected to the primaries and 20K with it reflected. But that is not the reason why I have the cap. I discovered the effect of a cap across the legs on sigma delta DACs nearly 3 years ago by accident. I reported in the these posts

http://www.diyaudio.com/forums/digi...native-post-dac-filtering-11.html#post4445352

http://www.diyaudio.com/forums/digi...native-post-dac-filtering-11.html#post4445577

That thread was started by a member who also noticed the same effect. He unfortunately got flamed for reporting it without adequate explanation. The effect is that is creates is a layering of instruments that makes listening to music - emotional.

The oldest reference I can find to this effect is here

http://www.raylectronics.nl/pdfs/Mauros_CD6367_tweak.pdf

Regarding the transformers - yes LL1544A is amorphous too. Nearly $85 in the US vs LL1674 which is $130

The resistor value to be chosen depends on your taste. Put simply - the deeper you put the DAC into current mode (by lowering the resistor) - the THD improves. At 0 ohms - the THD is the best.

But in all honestly - I can't hear THD in my system below -105db.

But I what I can hear is that the slam/attack/dynamics are lost deeper in the current mode.

The optimal value for the resistor is 4/5 times the output impedance of the DAC for my system. So 640 ohms for ES9028Q2M.

I have a modified IRAUDAMP7S driving a highly modded Magnepan 1.7. Preamp is a poorly biased JFET that creates a 2nd order harmonic at nearly -80db. I have been told 2nd harmonics change the tone but I perceive it more as body to the music. LPs naturally had this and Tapes had tons of 3rd harmonics

Sorry for the long reply. Hope this helps

Hi madds1,

That's very helpful. Thank you so much.

It seem's I should buy a pair of LL1544A to give it a try. I mainly use balanced output, is there any problem?

I still have some PCB available for the transformer I/V. I'd be very happy to share with you if you want to try something on it. DAC will be available later. Please let me know.

Regards,
Ian
 
Just wondering if the cap to ground really necessary?
I have a LL1636 , tried 2 x 10ohm to ground on primary. But output is too low. I will try a higher resistor value later.

Hi ChuckT,

The capacitor is for the visual ground. Theoretically you can remove it. The only thing I'm bit worry about is that in the real world the two windings can not be 100% balanced. So the center may have some floating voltage. Please
see madds1's implement for reference:

http://www.diyaudio.com/forums/pc-b...i2s-dac-hats-raspberry-pi-15.html#post5369598

Please let me know your result if you have any.

Regards,
Ian
 
Last edited:
Hi ChuckT,

The capacitor is for the visual ground. Theoretically you can remove it. The only thing I'm bit worry about is that in the real world the two windings can not be 100% balanced. So the center may have some floating voltage. Please
see madds1's implement for reference:

http://www.diyaudio.com/forums/pc-b...i2s-dac-hats-raspberry-pi-15.html#post5369598

Please let me know your result if you have any.

Regards,
Ian

I think Madds1 circuit is voltage mode? While with low value resistor to ground will run dac in current mode.
 
Yes, I use a capacitor directly after the DAC as well.

I don't believe that it is doing anything magic, but rather it acts as a functional low-pass filter to remove the very high-bandwidth transition artefacts generated at sample boundaries.

These transition are likely to extend into the 10s of MHz region, or even higher, and may well cause problems with op-amps in the I/V or any subsequent gain stages.

The need for such a filter was explained at least 20 years ago by the famous Dr Hawksford:

Dr MOJ Hawksford -Transimpedance Amplifiers

See section 5, page 21

On account of the bandwidth, the capacitor used in this position should be a surface mount type. The leads and general construction of through-hole caps will have enough inductance to impair the filtering at such high frequencies.

22-100nF are easily available in C0G 1206 ceramic caps, and these would seem to be first choice, although the polyphenylene sulphide [PPS] dielectric of the Panasonic ECHU (surface mount) series are also worth trying. Both of these should be soldered with extreme care, if using an iron - the damage caused by overheating is not easily visible.
 
Are there any cheaper alternatives on the market for ll1544A...from Neutrik, Monacor, ebay... Amorphus, Mumetal ...?
Madds1...i agree with your conclussions...i also found that es9018 is playing more analogue with (Rasmussen) filter cap...and that some second harmonics have influence on sound perspective...
With AK dacs it is not necessary because they are more damped inside dac Ic and noise is muuuch smaller on output of the AK vs ES dac. (and are voltage not current dacs) I think that too much digital noise coming in first filter opamp kill the dynamics, freedom, natural big soundstage...
 
Last edited:
Hello Ian, hello guys! I am very happy to know about Ian's new project. And it is about DAC this time! Cheers!
Regarding transformers available for ES90xx DACs outputs... For now it is totally finished the job of matching nano-crystalline-core transformers with ES90xx outputs. Schematic is similiar to madds1' (ratio is 1:3.5), but due to better ratio of inductance to coil-resistance in comparison with specs of any other available transfomers on the market, I get 2.6VRMS without any resistor across the primary coil. Moreover, the cost is much lower than that of brands.
Of course I will try yours DAC implementation on my table later, so I am in the list :)
 
Hi madds1,

That's very helpful. Thank you so much.

It seem's I should buy a pair of LL1544A to give it a try. I mainly use balanced output, is there any problem?

I still have some PCB available for the transformer I/V. I'd be very happy to share with you if you want to try something on it. DAC will be available later. Please let me know.

Regards,
Ian

PMed you separately.

You should be able to use any transformer with shielding pins for balanced output. Ensure that that the ground of the DAC board is connected to the shield which should then be connected to the chassis ground of the preamp

Cheers
 
Yes, I use a capacitor directly after the DAC as well.

I don't believe that it is doing anything magic, but rather it acts as a functional low-pass filter to remove the very high-bandwidth transition artefacts generated at sample boundaries.

These transition are likely to extend into the 10s of MHz region, or even higher, and may well cause problems with op-amps in the I/V or any subsequent gain stages.

The need for such a filter was explained at least 20 years ago by the famous Dr Hawksford:

Dr MOJ Hawksford -Transimpedance Amplifiers

See section 5, page 21

On account of the bandwidth, the capacitor used in this position should be a surface mount type. The leads and general construction of through-hole caps will have enough inductance to impair the filtering at such high frequencies.

22-100nF are easily available in C0G 1206 ceramic caps, and these would seem to be first choice, although the polyphenylene sulphide [PPS] dielectric of the Panasonic ECHU (surface mount) series are also worth trying. Both of these should be soldered with extreme care, if using an iron - the damage caused by overheating is not easily visible.

Hi Rod

I thought so too initially that there was the HF explanation but there is some other mystery to be solved as well. Specifically - if you ground the caps, the effect is gone. In other words if you replace the 47nf across the pins with 94nf to the ground from each leg - the layering of instruments and the emotional connection is lost.

Also - if you vary the 3db corner - i.e. move it to say 30Khz - the effect is gone. It only seems to near the 20-22Khz region. I settled on 19.9 corner for my system. See this post. Other has have the same experience as well

http://www.diyaudio.com/forums/digi...iscussions-upgrading-mods-76.html#post3777347

Joe tried to seek an explanation of this phenomena this thread but it didn't go anywhere

DAC Filtering - the "Rasmussen Effect"

He again tried nearly a year later and failed again

Practical Implementations of Alternative Post-DAC Filtering

Whatever the phenomena is - the music is more enjoyable :)

Regards
 
Are there any cheaper alternatives on the market for ll1544A...from Neutrik, Monacor, ebay... Amorphus, Mumetal ...?
Madds1...i agree with your conclussions...i also found that es9018 is playing more analogue with (Rasmussen) filter cap...and that some second harmonics have influence on sound perspective...
With AK dacs it is not necessary because they are more damped inside dac Ic and noise is muuuch smaller on output of the AK vs ES dac. (and are voltage not current dacs) I think that too much digital noise coming in first filter opamp kill the dynamics, freedom, natural big soundstage...

Completely agree. It is not required (and does not work as effectively) with AK DACs

So regarding transformers. Three things matter. Core, Winding and Structure. The core material determines 2nd and 3rd harmonics ratios relative to the amplitude, primarily because of the hysteresis curve. I used amorphous core since it had the least amount of artifacts at the 1V RMS. If your preamp and amp are low distortion then you can experiment with EDCOR (Steel Laminate) or Jensen (best mu-metal out there). Haven't tried anything else

In my system - the only place where color and warmth are introduced is at the preamp stage where I deliberately introduce in phase 2nd and 4th harmonics.The reason I do this is because of the open baffle nature of my Magneplanars. I prefer that it have asymmetric radiation. It also helps with improving dynamics since amplification follows a square law.

You can use the transformer to generate in phase 2nd and 3rd harmonics (symmetrical hysteresis curves generate both) and it may suit some systems

So core choice is very dependent on you overall system

Windings - these mostly influence low level detail retrieval. Lundahls, Jensens, Sowters etc are tightly wound. I would stick with Lundhals (price performance)

Structure primarily determines magnetic leakage or interference. Most audio transformers have a shield

In short - hard to find a transformer that can beat LL1545A or LL1544A for the price.

If you like the transformer effect in you system - use LL1545A (mu-metal) or any other brand. If the transformer is simply to sum the diff output - use amorphous core (affordable) or FineMET (ridiculously expensive - see)

If you don't care for summing the diff output - simply put a cap across the outputs. That is what Barbara does

http://web12491.serv4.spacequadrat.de/html/DAC_Nitro_DAB_V10.pdf


Regards
 
Hi Rod

I thought so too initially that there was the HF explanation but there is some other mystery to be solved as well. Specifically - if you ground the caps, the effect is gone. In other words if you replace the 47nf across the pins with 94nf to the ground from each leg - the layering of instruments and the emotional connection is lost.

Also - if you vary the 3db corner - i.e. move it to say 30Khz - the effect is gone. It only seems to near the 20-22Khz region. I settled on 19.9 corner for my system. See this post. Other has have the same experience as well

http://www.diyaudio.com/forums/digi...iscussions-upgrading-mods-76.html#post3777347

Joe tried to seek an explanation of this phenomena this thread but it didn't go anywhere

DAC Filtering - the "Rasmussen Effect"

He again tried nearly a year later and failed again

Practical Implementations of Alternative Post-DAC Filtering

Whatever the phenomena is - the music is more enjoyable :)

Regards

Yes, I did see that thread and have thought about it from time to time.

But one thing can be cleared up immediately: if you connect the capacitors to ground, I would expect a degradation of the sound.

Why? because connecting them to ground will cause the noise-currents to circulate into ground and contaminate some of the most sensitive nodes in the whole playback chain.
The bandwidth of the noise can reach into the VHF region, and it will not follow the paths that you intended!

The best to deal with this noise is to connect the cap across the DAC-out phase and anti-phase, which gives the best chance of a short and direct path for the currents.

Beyond that, anything that increases enjoyment should naturally be put to use ;)
 
You are absolutely correct. I don't know how that didn't occur to me at all. No wonder Barbara in the Nitro DAB uses hand wound (5 turns around a pencil) inductor to connect the digital and analog grounds!

All this chat has made me wonder if I should simply uses Barbara's DAB design (skip transformers). The only reason I using transformers is to get back the 2V RMS, sum the outputs and finally be able to use diff or single ended in the future for experiments

But over the years I have settled on single ended. The noise floor in my system at listening levels is easily at -100db. Time to try retiring the transformers :)
 
Yes, I did see that thread and have thought about it from time to time.

But one thing can be cleared up immediately: if you connect the capacitors to ground, I would expect a degradation of the sound.

Why? because connecting them to ground will cause the noise-currents to circulate into ground and contaminate some of the most sensitive nodes in the whole playback chain.
The bandwidth of the noise can reach into the VHF region, and it will not follow the paths that you intended!

The best to deal with this noise is to connect the cap across the DAC-out phase and anti-phase, which gives the best chance of a short and direct path for the currents.

Beyond that, anything that increases enjoyment should naturally be put to use ;)

Make sense :)
Ian