This discussion started from one at Head-fi.com called "Why Your Awesome IEMs Sound Bad from your iPod and what you can do about it." It initially involved a discussion concering lowering a file's levels using Audacity so that the EQ stages of the iPod wouldn't overload. That led to several pages of discussion, where eventually I cited dbPowerAMP as a way of automating such a task, using the "Volume Quieten" DSP, but then led to markopolo's suggestion that I investigate replaygain/mp3gain/aacgain as an even more efficient method, undo-able, and working on the internals of the mp3/aac file so it is decoded by the player at lower digital levels, pre-amplification. Following is a copy of my post verbatim. http://www.head-fi.org/forums/showpost.php...mp;postcount=77

I am wishing for an accuracy check concerning what I have posted, please, as I know there are many very knowledgeable programmers and electronics-engineering savvy people here. Is what I am stating accurate?

Thank you - I hope this leads to a useful FAQ item helping people gain-stage optimize their portable players, PC soundcards and laptops and simulataneously load-optimize them for low impedance headphones through a simple and easy file manipulation. [Edit:] This query is to benefit people who do wish to use their iPod's (or other source's) built-in EQ and their existing iPod (or other source's) headphones-out amplification.

Terry

[Head-fi.com original post]:
mp3Gain and AACGain appear to affect pre-EQ stages - Success

OK, according to this thread at Hydrogenaudio, http://www.hydrogenaudio.org/forums/...topic=24527&hl, mp3Gain and AACGain (both being implementations of the ReplayGain algorithms) affect volume settings of the actual few-millisecond-long individual blocks of the mp3 file, so it is an internal adjustment to the mp3 (or aac) file itself, and so, said volume adjustments are decoded directly by the player when reading the file, before entering any amplifier stages at all. The upshot is that using this program to reduce the overall volume sends a lowered signal to the fixed-preamp EQ in the iPod, and eliminates the problem of that EQ's output then overloading the input stages of the final amplifiers downstream.

The tags these programs write are apparently only for undo purposes.

That's right - the changes you make are entirely undo-able!

There are some problems that are raised in the above-mentioned post concerning these "undo" tags causing older versions of foobar2000 as a media player to ignore the id3 tags in favor of the APE format tags that hold the undo information, but workarounds are listed in the above post for people using foobar2000 who have "lost" the id3 tag info. (again, not really lost, only ignored...) These problems no longer seem to exist in current editions of foobar.

I used a setting of 83 dB that allowed me to pump both my Nano's and laptop's amp to 8-ish or 9-ish and maximize their power loading on my low impedance headphones. It is making these smallish amps work fine with Shure E4c and Ultrasone Proline 750s, as well as with my "portable" phones like the Portapros and Sennheiser PCX-250s. (Coincidentally, that was the program author's original default, though that has been raised to 89 dB in current editions. For power-loading/impedance matching purposes, 83 should work better.)

This should result in a very clean and distortion-free output from even these little 10 mW headphone amps. The added benefit is that this blows Apple's Sound Check out of the water. Be certain to turn Sound Check off on your iPod or in iTunes for playing files you've converted using either mp3Gain or AACGain.

BTW - get mp3Gain from here (This is all you need if you only intend to work with mp3 formats)

Then get AACGain from here near the bottom of the page (this is a direct link to that file)

If you intend to use the AACGain with this frontend: Go to your Program Files folder and inside the mp3Gain folder, change the name of mp3gain.exe to mp3gainOrig.exe to keep it for using with mp3s at another time, and then change the name of AACGain.exe temporarily to mp3gain.exe instead so the front end will find it.

The help files are very useful that come with the program. Experiment and tell us what results you got from it. Thank you, Markopolo for drawing my attention to these programs.

I hope this is of huge benefit to everyone who used to hate the iPod's unamped EQ sound! It definitely has improved matters, plus added the fringe benefits of levels controlling my tunes.

Terry

You can still technically undo the volume changes even without the APEv2 tag.

Given that most IEMs have insane sensitivity anyway, why not just add a 16 ohm resistor in series? You volume would get ever so slightly lower, but its not like you really care, since that the average pair of 16 ohm IEMs can easily cause life long hearing loss when the ipod's volume is maxed.

At least to me it seems like solving a hardware problem with a hardware solution makes the most sense.

This technique appears to give you more headroom so that the iPod's EQ (which apparently only adds, not subtracts) doesn't lead to clipping.
I don't understand how this technique does anything to address the fact that RMAA test after RMAA test shows the iPods (sans Shuffle) having a devil of a time driving sub 32 ohm loads w/o distortion.


This will help greatly with the impedance issue, but will do nothing to solve the clipping caused by usage of the EQ.



(I'm assuming there is no debate on the iPod's flaws when it comes to 16 ohm loads, but I will post links if that is what it takes to avoid a TOS violation)

This may be off-topic, but: are there passive impedance-matching devices for headphones, that work with low-level signals? I've seen some that work with the speaker outputs of power amps, but nothing in the portable domain.

I suppose some of the portable active headphone-amplifiers might allow one to work around this issue, if they have a better impedance match. But they all require power, of course.

-brendan

Depending on headphones that will alter the frequency response because the headphone impedance may not remain same for all of the audio band. With Shure E2C there was no problem in adding series resistance, but with Etymotic ER6i higher frequencies were emphasized relative to mid and low frequencies.

Also this solution doesn't help with the software EQ clipping.



That's what I was wondering too.

This technique appears to give you more headroom so that the iPod's EQ (which apparently only adds, not subtracts) doesn't lead to clipping.
I don't understand how this technique does anything to address the fact that RMAA test after RMAA test shows the iPods (sans Shuffle) having a devil of a time driving sub 32 ohm loads w/o distortion.

Well, this doesn't try to address this issue, really, but allowing the final gain stages to run more full-out does allow more voltage to run, dropping across the impedance available for any given frequency to supply more current, if I remember correctly from my sound-reinforcement days. But I've remembered things backwards before, and might yet again...

Given that most IEMs have insane sensitivity anyway, why not just add a 16 ohm resistor in series? You volume would get ever so slightly lower, but its not like you really care, since that the average pair of 16 ohm IEMs can easily cause life long hearing loss when the ipod's volume is maxed.



That is a popular solution with the E4c crowd, since it comes supplied with an in-line volume attenuator. But the whole idea of impedance is that it changes resistance in line with the frequency. A fixed resistance would add coloration, and if that's what you want, OK. But that is not what I was seeking to do here.

True, but theres already a strong bass roll off anyway with 16 ohm headphones that the resistor would fix, so even if you headphones have a very high reactance, I doubt its going to be much if any worse then what you'd get without it. Possibly much better.



I assumed that with the leveled out frequency response, you wouldn't need EQ. If you're going to use it regardless, I guess you're probably going to be unhappy because of how poorly it works in the Apple firmware. Ever looked into buying an X5 and using the more powerful Rockbox EQ? It would solve all your issues at once.

I'm not clear where the 16-ohm headphones thing came into this... The poster was mentioning a 16 ohm resistive load be dropped across the output, but that was for any headphones having various impedances.

Some headphones I'm familiar with:
Shure E4c - 29 ohms @ 1kHz
Ultrasone Proline 750 40 ohms @ 1 kHz
Sony "Altus" MDR-D777LP 24 ohms @ 1kHz
Koss Portapro 60 ohms @ 1 kHz
Sennhieser PXC-250 noise cancelling 300 ohms (have built-in, inline amplifier as part of circuitry)

This query is to benefit people who do wish to use their iPod's (or other source's) EQ and do not wish to use outboard amplification, so though a useful solution to problems, an X5 and Rockbox (which is incompatible with iPod Nano 2nd generation still) unfortunately does not address the needs being addressed here. Thanks for the info, though!

Terry

A (very safe) assumption made from the title of the referenced post:
"Why Your Awesome IEMs Sound Bad from your iPod and what you can do about it."

I originally said "sub 32" and Mike said "16 ohm IEM". My argument was simply that the clipping caused by the iPod EQ was not the only reason "Your Awesome IEMs Sound Bad" and I didn't see any mention to the impedance issue at all in the Head-Fi discussion. (Sometimes I wonder if Head-Fi groupthink precludes RMAA testing.)




Of all the headphones you list, only the Shure E4Cs are IEMs, and they are still below the 32ohm point at which the iPod output stage starts to behave.
Most IEMs on the market are 16ohm. That is where the number came from.



P.S.
I don't understand this quote:

I don't understand how adding a constant resistance adds color. As far as I understand (and feel free to correct me if I am wrong), all you are doing is shifting the impedance graph up. You should be equally attenuating all frequencies, thus preserving the original character of the 'phones, no? Coloration would come from a change in the delta ohms, which a constant doesn't do. Correct?

Aha! I've muddied things up myself, then. Sorry.

Yes, the original topic focused upon IEM's, but actually was the original poster's response to the problem of iPod EQ overloading. The entire discussion spun off of that effort alone, and so was actually more to do with gain-staging generally and not much to to with IEM's per se as it panned out.

I myself was not aware of the RMAA testing reports nor realized it brought up issues concerning low impedance headphones (sub 32 ohms.) Very interesting! Yes, then, a 16 ohm load might work for those cases very well placed in parallel, if I recall. [Edit - nope - series. Parallel can only reduce impedance.]

As far as it adding coloration, I do not know - I have been lulled into thinking it did add coloration by testimonials from people employing their volume attenuators to round out their sound somehow. I do not understand the mechanism by which that would be the case - it may be the attenuators apply the load in series, literarally in-line with the load, and a resistance in series with to an inductance is simple-additive, right? I don't remember!

I always assumed as you do that a constant would apply a constant effect. Please restore my dusty memory on this issue!

The primary purpose of using mp3gain or aacgain on the files was to be able to gain-adjust the files pre-EQ, and these programs seem to change the internal instructions by which the mp3 or aac file is decoded into volume levels, making use of these programs seem an ideal solution for this purpose. The EQ distortion problem is eliminated by lowering the input coming into it. This is the original and primary issue of interest: whether this approach to inherent gain-staging problems is a useful one.

I added the note of lowering as much as to 83 as a target because of the ability it then gave me to open the final amplifier stages up, seeking a further benefit from that event. I do not know if "opening the valves" can exert a positive effect counterbalancing whatever it is that is causing the sub-32 ohm distortions problem, though I can postulate that it might make matters worse if the cause of the distortion is excess current being returned to the rails somehow. (Though I had assumed the opposite would be true, I can see where that assumption might fall flat. I am assuming also that the iPods and other similar DAPs are using op-amp feedback configurations - there are other things I am assuming for this "load warrior" thinking of mine to be applicable.)

Thank you for your clarification about these things - I really would like now to also be better educated concerning the use of the volume attenuator as adding any further benefit beyond that of volume attenuation!

Terry

I too would like clarification on this issue, because I think you are making this much more complex than it appears to me.
I believe you need to forget everything about the analog stage of the chain for now. I think the entire iPod EQ issue comes down to digital clipping. If you take an audio file created from a CD with no volume headroom (as many new ones are) and apply 5 or even 10dB of gain with a digital EQ, you will clip. By applying MP3gain you are reducing the volume of the track before it gets to the EQ, thus creating the headroom needed to safely boost w/o clipping. Only the 5th generation iPods have a hardware EQ, the prior generations had a software one (thus in the digital domain), and it appears that Apple did not apply a negative pregain to the audio before boosting with the EQ.

With Rockbox you can easily reproduce the Apple firmware behavior, OR you can apply a negative pregain, and avoid the clipping / distortion.

If someone can think of a test to prove the EQ problem is (or is not) simply digital clipping, I will be more than happy to perform said test and post the results.

RMAA test showing that Rockbox has no magic powers compared to the Apple FW: (thus RB can be used as an accurate test of iPod behavior)
Here.

RMAA test illustrating the iPod issue with sub 32 ohm loads - this one on a Nano 2nd gen, others are very similar.
Here.

Holy Moly! That is a disaster area at 16 ohms in every way possible! What a bass rolloff! That THD! Gadzooks! No wonder people using their E4c's feel the bass is a little weak and want to EQ it back! Worse for lower ohm impedance IEMs (I'd love to know whose models are at 16 ohms!)

Yes, you are probably right that I am jumping the gun in trying to get wrapped up in the analogue stages, which only affect the final output as far as I can guess, all before being in the digital realm. The mp3gain/aacgain trick does seem to work to perform the fix in the digital EQ domain and stop the clipping. So, using that, people can EQ some of their bass rolloff back into the picture! But I am just amazed to see the problem is that BAD! I'll pass this graph along to the head-fi-ers.

I never expected the phones out jack to be so identical to the line-out either, Rockbox or Apple! Seems that if we can load the built-in amp adequately, there may be much less need for external amps to show an improvement - a marked improvement at least. GIGO applies even to line outs going into nice external amps, right? (I still want a Xin mini V5!)

Terry

A made up example:

Impedance: 1 ohm at 1KHz, 2 ohm at 10KHz.

Let's add series resistance 2 ohm.

Apply 1V @ 1KHz -> headphones get 1/3 volts, 2/3 volts to resistor.
Apply 1V @ 10KHz -> headphones get 1/2 volts, 1/2 volts to resistor.

Take resistor out and headphones get same voltage across audio spectrum.

Regarding this and other RMAA tests in general, care should be taken when choosing testing parameters. The noise performance hints that Nano was at full volume. With a low impedance load and high voltage the headphone amplifier may simply be running out of current. Distortion will quickly rise in such conditions. I don't believe this particular RMAA run is representative of normal use. It only shows that 1G Nano's performance isn't quite as bad as 2G Nano's in unrealistic and extreme situations.

I can do just that. Give me a few hours.
EDIT - A few more.

Good. When you turn the volume down to normal levels, don't be alarmed by the noise level. A 'poor' figure may not actually be that bad. All that matters is the level of hiss audible with the headphones of choice.

Here is another mention of the series resistance affecting sound quality that I found at head-fi.com

http://www.head-fi.org/forums/showpost.php...amp;postcount=6

"A unique quality of single driver balanced armature canal phones is that adding resistance to the circuit increases the level of treble response. The easiest way to do this is to plug in a Radio Shack in-line headphone volume control, which is basically a variable resistor. Lowering the volume on the RS control decreases volume, but not in equal proportion across the spectrum. If you compensate by increasing the volume on the source, you will notice an increase in level of the treble range. Advanced DIY types can use this technique to determine how much resistance suits their taste, then fabricate a higher quality fixed resistance. For us lazy types, we can just dial in the RS control and get on with the day. - gerG"

I am trying this approach right now, and it is allowing me to run my E4c's flat, with no EQ, and plenty adequate treble, using the attenuator that came with these. Interesting. And it must be helping the bass response curve not to roll off at the same time - two for the price of one fix! So simple at that. Had I not seen the graphs, I would have been dubious that I was only "hearing things". But this does point to the addition of a resistor leading quite definitely to a sonic signature change which is quite audible.

According to bangraman in another post at head-fi http://www.head-fi.org/forums/showpost.php...mp;postcount=19
the ideal resistance he came up with was 46 ohms (well, two of them, one for left and one for right channel in series).
"The resistor value is 46 ohms. Tested with an E2c load, what it does is (bad) introduce a ~0.5db cut in the iPod's response across a wide range from 20khz to 500hz. This gives it less 'air'. However, at 50hz, the regular E2c is showing a >4.5db falloff, yet [using the 46 ohm resistors] is showing a 1db falloff."

And this page: http://phonephile.blogspot.com/2005/06/ipo...s-fall-off.html discusses how these RMAA tests are accomplished and discusses the rolloff issue. Very interesting!

Terry

Much mention has been made of using series resistors to resolve the impedance issue with IEMs and iPods. That solution works to reduce the bass rolloff, but also affects the highs somewhat, as the resister changes the impedance differently at different frequencies, plus it interacts with the output capacitors making it act like the resistor in a tone control circuit, right?

Doesn't a parallel resistor (probably a fairly high resistance) increase the impedance evenly no matter what the frequency? (I'm pretty sure I remember that a parallel resistance increases an impedance evenly... but do not remember the formula. Is it as simple as straight addition, or is it like the R1*R2/R1-R2 parallel resistance formula?) What would be good values for the parallel resistance if that is a workable solution? [Edit - nope - see below.]


And would it still interact with output capacitors creating a tone control effect?

The series solution works adequately, but I couldn't help but feel these questions nagging at me. My memory is distant of the particulars behind this circuitry stuff.

Terry

[Edit]
I think I've answered my own question:
* Parallel AC circuits exhibit the same fundamental properties as parallel DC circuits: voltage is uniform throughout the circuit, branch currents add to form the total current, and impedances diminish (through the reciprocal formula) to form the total impedance.

http://www.allaboutcircuits.com/vol_2/chpt_3/4.html

So, it wouldn't work - a parallel resistor could only diminish the inductance, not add to it, if I'm reading this correctly...