[Iron_Dreamer]

This mod I will describe to your is the result of a recent revelation in the RME modding thread. It only costs a few bucks, needed to get a pair of high-quality capacitors, and a few minutes of your time. Better yet just about anyone that can use a soldering iron (and maybe a few that never have) can complete this mod, because it is so easy to do.

Bypassing the entire analog output stage as is done here gives the card an amazing sound quality increase for such an easy tweak. The former harshness in the high-end is gone, the highs are also more extended than before, detail is up across the board, macro and micro dynamics are vastly improved, and bass impact is way up as well. Basically this makes everything about this card better, and corrects many of its' flaws in stock form.

It does decrease the output level by about 6dB, so if you are near the upper limit of the volume control of your amp with you current cans at a normal level (with all the volume controls in the RME driver maxed), it might not be loud enough for you once the mod is done.

Hopefully I will have pics soon, to make my instructions that much more clear.

Note: this mod only works if you have a headphone or speaker amp attached to your RME, you should not plug headphones directly into the card in this configuration. I take no responsibility for those who might misconstrue my instructions, or improperly complete these mods.

Step 1: Order 2 electrolytic capacitors (at least 5V, 10uF, preferably non-polar) from anywhere you prefer. I used Blackgates Pk series 50V 22uF, from www.partsconnexion.com .

Step 2: Shut down computer, remove the RME card from your system.

Step 3: Remove the stock output capacitors from the top of the card, by grabbing with a pair of needle nose pliers, and twisting gently back and forth until they pop off.

Step 4: Hold the card with the outputs facing left. On the underside of the card, locate the 8 legs that used to be attached to the capacitors you just removed. You will be soldering the second and seventh of those legs to your new capacitors (counting from left to right).

Step 5: Align one of your capacitors between the second leg discussed in the last step, and the lower of the two solder points at the underside of the 1/4" output jack, and cut the capacitor legs to an appropriate length. Repeat for the upper solder point of the jack, and the 7th leg. Be sure that the capacitor legs don't touch any other parts of the card.

Step 6: Solder down the capacitors as described above, again making sure not to make contact with any other parts of the card, save your intended solder points.

Step 7: Inspect to make sure both ends of each are well-joined.

Step 8: Put the card back into your computer, start it up, hook up you headphone or speaker amp, and test to make sure that you have done it properly.

Step 9: Enjoy! The sound only improves as your new caps burn in.

 

[Iron_Dreamer]

And BTW, if you don't think you could do this yourself, I am open to doing this mod for fellow head-fiers for parts cost+return shipping.

 

[Wodgy]

Just one extra detail that's worth mentioning: if you're using polar capacitors, you want to get the orientation right. Connect the positive leads to the capacitor legs you removed, and the negative leads to the output jack.

 

[Edwood]

This begs the question, if this sounds so good, why in the world did RME include an analog stage to begin with?

Was it boost the output volume?

Well, since it seems to lower the output volume, this volume loss would not be good news for powering my K1000's. I guess we'll see.

Pics to follow.


-Ed

 

[Wodgy]

Quote:

Originally posted by Edwood This begs the question, if this sounds so good, why in the world did RME include an analog stage to begin with?

I gave three reasons for this in the RME modding thread. Interestingly, one of the three reasons turned out to be unexpectedly untrue -- the op-amp summing stage introduces more noise and lowers the dynamic range when one would expect the opposite to be true.

 

[Iron_Dreamer]

Quote:

Originally posted by Wodgy I gave three reasons for this in the RME modding thread. Interestingly, one of the three reasons turned out to be unexpectedly untrue -- the op-amp summing stage introduces more noise and lowers the dynamic range when one would expect the opposite to be true.

My listening experience is definitely consistent with the RMAA test confirming that result. The background is definitely blacker, notes jsut seem to fade away into oblivion.

 

[Edwood]

Well, I guess you never know until you try.

Good call. Looks like we may have a winner.

So I guess this means I'll be making a trip up to your place again, Peter?

Heheh, this time I'm bringing a lot less equipment. And we'll probably not have the whole day to play. My better half is starting to get annoyed by this hobby.

-Ed

 

[Glassman]

it's simple - low pass filters, either analog or digital, introduces pre-ringing effect and time smearing.. this is the magic behind non-oversampling filterless DACs.. there is more or less serious side effect - significant high frequency mess after 20kHz.. the question is if it does matter or it doesn't.. with today's delta-sigma oversampling DACs, this high freq mess is well above 20kHz so it probably doesn't really matter.. however, there are probably some requirements or limitations on how much of this mess could be present on the output if you want to get a sticker with 'approval'

the method described here by Iron has some serious issues.. driving interconnect cable directly from DACs high impedance voltage outputs is definitely bad condition.. also, only positive outs are being used.. it would be a lot better if balanced outputs were used, this would work even without any capacitors..

in case you'll have some time for your solder, try removing the LPF components so that the opamp will only sum, amplify and buffer the DAC's outputs, ok?

 

[Iron_Dreamer]

Quote:

Originally posted by Glassman the method described here by Iron has some serious issues.. driving interconnect cable directly from DACs high impedance voltage outputs is definitely bad condition..

What is bad about it? I have 1.5M cables, and it doesn't seem to be a problem. Perhaps much longer cables would have issues.

Quote:

Originally posted by Glassman in case you'll have some time for your solder, try removing the LPF components so that the opamp will only sum, amplify and buffer the DAC's outputs, ok?

Which components would I remove to do that? All in the signal path between the original cap positions and the opamps?

 

[Wodgy]

Quote:

Originally posted by Glassman it's simple - low pass filters, either analog or digital, introduces pre-ringing effect and time smearing.. this is the magic behind non-oversampling filterless DACs.. there is more or less serious side effect - significant high frequency mess after 20kHz.. the question is if it does matter or it doesn't.. with today's delta-sigma oversampling DACs, this high freq mess is well above 20kHz so it probably doesn't really matter.. however, there are probably some requirements or limitations on how much of this mess could be present on the output if you want to get a sticker with 'approval'

A -120dB stopband is probably good enough for most regulatory agencies.

Quote:

the method described here by Iron has some serious issues.. driving interconnect cable directly from DACs high impedance voltage outputs is definitely bad condition.. also, only positive outs are being used.. it would be a lot better if balanced outputs were used, this would work even without any capacitors..

As long as you're using reasonably short cables and an amp with a high input impedance, this is not a problem. Iron_Dreamer did the RMAA tests that prove it. As for the balanced outputs, it's clear that the analog stage's power supply or lack of shielding meant that the benefits were theoretical, rather than real. Iron's RMAA tests again show this.

Quote:

in case you'll have some time for your solder, try removing the LPF components so that the opamp will only sum, amplify and buffer the DAC's outputs, ok?

Not a bad idea, and worth trying, but it's not a major change in the overall topology. You've still got tons of feedback to kill the microdetails. But someone should try it.

 

[Glassman]

Quote:

Originally posted by Iron_Dreamer What is bad about it? I have 1.5M cables, and it doesn't seem to be a problem. Perhaps much longer cables would have issues.

I don't claim it sounds bad, but it can sound a lot better.. if we want to get rid of high negative feedback, you can try open loop buffer like OPA634 or HA3-5002..

Quote:

Which components would I remove to do that? All in the signal path between the original cap positions and the opamps? [/B]

I think removing all ceramic caps before (maybe also after) the opamp would do the job..

 

[Wodgy]

Quote:

Originally posted by Glassman I don't claim it sounds bad, but it can sound a lot better.. if we want to get rid of high negative feedback, you can try open loop buffer like OPA634 or HA3-5002..

Open loop buffers don't sound very good unless they're in an op-amp's feedback loop to correct their nonlinearities. Also, I don't see how adding unneeded components to the signal path would make it sound "a lot better."

 

[Glassman]

Quote:

Originally posted by Wodgy A -120dB stopband is probably good enough for most regulatory agencies.

yeah, but that's the stop band attenuation, after that point it climbes high very fast.. look at some typical sigma-delta modulator output spectra

Quote:

As long as you're using reasonably short cables and an amp with a high input impedance, this is not a problem. Iron_Dreamer did the RMAA tests that prove it. As for the balanced outputs, it's clear that the analog stage's power supply or lack of shielding meant that the benefits were theoretical, rather than real. Iron's RMAA tests again show this.

this is a problem no matter how long cables you use.. cables should be driven from near zero output impedance sources.. this is too far from reality in this case..

Quote:

Not a bad idea, and worth trying, but it's not a major change in the overall topology. You've still got tons of feedback to kill the microdetails. But someone should try it. [/B]

opamps are not THAT bad, when good ones are used, biased and with no LPF on them I'm pretty sure it would sound better.. the safest way is to use monolithic open loop buffer..

 

[Iron_Dreamer]

Quote:

Originally posted by Glassman I don't claim it sounds bad, but it can sound a lot better.. if we want to get rid of high negative feedback, you can try open loop buffer like OPA634 or HA3-5002..

Could you explain negative feedback to me? Also, how would a buffer affect the sound as compared to an opamp?

I think I may try removing all those other components, since I know how good it can sound without any of the stage, so screwing it up (potentially) is of no concern to me anymore.

 

[Wodgy]

Quote:

Originally posted by Glassman yeah, but that's the stop band attenuation, after that point it climbes high very fast.. look at some typical sigma-delta modulator output spectra

Glassman, you have a very poor understanding about this stuff. The AD1852 uses an 8x oversampling digital filter. The ultrasonic noise floor is flat to -120dB all the way up to 176.4 kHz, at which point you get noise from the sigma-delta modulator. This is so far from the audio band that it really doesn't matter.