feedback about gears. stop doing it wrong! impedance.

[castleofargh]

For years I've been mad when reading stuff like "this device is warm", or "those 2 amps really sound very different"... because I knew that half the time the guy was making a claim based on poor testing and even poorer understanding of what was going on.

For this demo/advice/critic, I used only 2 IEMs and the sources I could directly plug into my laptop so I could get done with it rapidly, but obviously this applies to DAPs(digital audio players) and anything with an amplifier output.



Here is an example with my laptop's headphone out, Odac/O2, Scarlett 2I2, and a portable DAC/amp, the UHA760. I measured them all at the same voltage straight into the input of the Scarlett(10kohm input). And as you can see even without clicking on the pic to expend it, they're all flat. I even pushed the evilness and used asio once, 44.1 one time, 48khz for others, different USB inputs(usb 3 or usb 2). Yet all we can really see is that one of them is going down just before 20khz( the one I set to 44.1khz). Else, super flat, super good!

You see how meaningless those unloaded specs can be when it comes to guessing how DAPs or amps will sound with your headphone. The reason is simple, an unloaded measurement shows how the device behaves when it's not plugged into a headphone... Yup it's that silly. Once you plug your headphone into a device, the signature measured and what you're hearing can very much be different(and will be).
Unloaded frequency response is given because it looks nice(marketing) and is easy to do, not because it's helpful for the consumer.
It really doesn't help the consumer. If anything it makes many people think that measurements are wrong and pushes them to distrust graphs even more. That's a really sad side effect of having irrelevant graphs and people who don't know what they're looking at.

"My DAP XXXX has a warm signature", "all amps sound different", and other such nonsense.

One obvious possibility is that the specific headphone I'm using to listen to different gears is in part causing the differences I'm hearing. This is really at the heart of this post. I won't touch all the placebo, bias, and psychological stuff now. Real sound differences heard correctly are today's only cause for misleading conclusions.

Someone will think:
"I'm using the same headphone, so any difference I'm hearing must come from the source". So if that DAP sounds warmer, then it's warmer. Which seems intuitive. But it is a fallacy. Audio gears aren't Lego! They're electrical circuits and the headphone becomes part of the total circuit. meaning the amp specs are now interacting with the headphone/IEM.

Here I'm showing the actual variations in signature as measured right out of a pair of JH13 IEMs with a microphone, all when using the different "super flat" sources from above:

Oh boy!

Already it's pretty clear that the unloaded graph from before was not helping in any way to guess the final signature I'm hearing out of my JH13. it's not that measurements don't "work". This one tells it like it is. It's just that if you measure the wrong stuff, that's what you get. Wrong stuff. ^_^

So what could be the cause of those changes in signature from device to device with the same IEM, given how they were apparently flat on their own?
I measured the UHA760 somewhere around 0.4ohm.
I get the O2 around 0.6ohm
The headphone out of the scarlett 2I2 is close to 10ohm
And my laptop's headphone out is in the area of 60ohm.
could there be a correlation between my JH13 sounding brighter and the source's impedance. Indeed there is:

This is the impedance curve of my JH13. it doesn't take an electrical engineer to notice the similarities between the frequency response variations I measured, and the shape of that impedance curve.
For those who don't know, headphones and IEMs aren't perfect resistors so they can often show such fluctuations in impedance depending on the signal we send into them. The wildest curves are usually the result of using crossovers with multidriver IEMs or speakers. But just one driver can also show fluctuations over frequencies.

Now let's go back to me making a review or a post on headfi and discussing the sound of gears. If I knew none of this, I would come and say that the UHA760 is a super warm device, and that my laptop sounds cold. Because of course it's very noticeable when I listen to music myself with the JH13.
But this is true only when I'm using the JH13. So anytime I'm making claims about the signature of the UHA760 without being specific about how it is the sound I heard while using JH13 IEMs, I'm misguiding the community with a false claim.
It's hard to blame the ignorant, one cannot know what he doesn't know. But some make it easy for me by acting like they're audio experts while actually being absolute Jon Snows.

The easy fix when you guys share your experience, is first to always mention all the elements in the playback chain you used to get your impressions. It's not hard and it changes everything. So please do it.

Second is to try your best to make it clear that they are your impressions, as in what you felt instead of what factually is about a device. That's a mighty important point too, because it makes the difference between sharing your honest impressions, or being wrong about the objective nature of a DAP or amplifier.
When writing a post, the line is very thin and I'm sure I've crossed it myself many times. But let us all try to be more honest by describing impressions, as impressions.
"I feel like", "what I'm hearing is", "it seemed to me that".... As long as you're sincere, you're telling the truth about youyr feelings, which are typiclaly the only thing you have to share.
But when writing, "the DAP is warm", "that amp has rolled off bass", and other objective claims presented out of your very specific listening context, is the device like that in general? Is it only on occasion when paired with peculiar gears? is it at all about the DAP or the amp? You probably didn't get the objective data to be conclusive about this, so why even take the risk? You could simply tell what you felt with a specific combo and leave anything beyond personal impression for others to figure out.



Here is a different example, in case the correlation with the IEM wasn't obvious enough. Now I'm using a pair of Shure SE215 with the same sources:

We go from a signature shift of almost 8dB with the JH13 between my laptop and the UHA760, to less than 0.5dB variations here while using the SE215 with the same sources under the same conditions. That just how much the load(IEM, headphone) can impact the results sometimes with how they interract with the amp section.

So imagine a second reviewer with his SE215 who's going to argue that all those sources sound pretty much the same. What happens when he meets me and my JH13? Noise between 2 wrong people is what usually happens. And if neither knows about all this, they can only rely on how they really felt a certain way and conclude that the other guy is wrong. When we have too little information, we still draw conclusions, it's in our nature. but of course those conclusions can be complete BS if we're missing out on relevant information.

And by now you should be able to guess why the SE215 doesn't show much variation. The impedance of the SE215 is almost completely flat:


So depending on the IEM or headphone used, different amps and DAPs might give you different impressions. Call that synergy if you don't care to understand the electrical implications, but remember that even when what you're feeling is correct, your anecdotal experience of a device probably does not define the device under all or even most circumstances. Remember it next time you want to claim that you know how some device sounds and how that other guy must be wrong because he experienced something different with a different playback chain.
Again here I'm not talking about personal taste, differences in our hearing abilities which are obviously relevant but a story for another time. I'm also neglecting all aspects of "hearing with your eyes"(the great many psychological biases inherent to human brains). I'm only talking about people who have the right impressions about the sound but come to false conclusions anyway, in this particular case, because of impedance and the habit to turn anecdotal experience into rules.

If I had something like a Shure SE846 with me, I could show the reverse effect compared to the JH13. With the 846 my laptop would now be the one sounding super bassy, and my UHA760 would have sounded the "coldest". because the general impadance curve of the the 846 tends to go down as frequencies rise, while for the JH13 it's the opposite.

From a practical angle we don't always have the impedance curve of the IEM/headphone we use, so that can make it difficult to predict the sound you'll get with a given source. But always keep in mind that if you use IEMs and headphones with an impedance much bigger than that of your amp or DAP, then the signature changes due to amps will stay small(less than 1dB if at any frequency the impedance ratio is at least 1/8). To make reviews it is a great idea to procure an IEM/headphone with a flat impedance curve and ideally not too low an impedance. That way you can expect most DAPs and amps to sound the way they should when used correctly(cf: impedance bridging). That's one easy way to improve the accuracy of a review, by testing the gear under nominal conditions. Having other IEMs with extreme electrical specs can be of value too, very sensitive to notice background noise, crazy low and wildly varying impedance to notice amps with high impedance even if you don't know how to measure it(like if the bass is a little weak on my JH13 I know the amp section is most likely above 1ohm. Extreme gears can be very helpful, just not as the main source of impression discussing normal use.

Also some device might have protection caps at the output, in that case, a low impedance IEM will roll off the low end, and the lower the IEM's impedance the bigger the roll off. That can be another way to get technical information from a subjective impression with the right IEM and some minimal understanding of what we're hearing and why.
Otherwise, low impedance amps and a good deal higher impedance IEM/headphone are what you should aim for. The "at least 1/8" damping ratio rule of thumb is your friend most of the time. And can really help a review hit closer to home.

Of course I focused this post on the amplifier section, but it's just as true when you decide to tell people how a multidriver IEM sounds but forget to mention that you're using some weirdo cable into the high impedance balanced output of some fancy DAP. The electrical interactions obviously go both ways.
And of course because I'm focusing on one frequent cause for signature change, doesn't mean there cannot be other causes. Nonetheless, following impedance bridging rules and having matched output levels tends to take care of a great deal of the actual sound differences that people notice. That's why I believe it's worth knowing it exists.







Tomorrow I'll post about how eating Pringles while reviewing gears was kind of an issue. (OK maybe I won't)










Addendum(stuff nobody cares about):
All those measurements were done rapidly by noob me. I used rather cheap measurement gears and didn't wait for the IEMs to settle in the coupler(the se215 had foam tips but I'm crazy I don't wait^_^). I moved the cables around a lot from device to device between measurements, so some stuff aren't as flat and clean as they should be. But I feel that the general points are visible enough to make those graphs relevant. If you don't think so, shoot me and tell me what I absolutely have to change.


I used:
Behringer ECM8000 microphone with a cheap coupler.
Scarlett 2I2 as ADC
The output of all gears were set to reach close to 90dB SPL out of the IEMs, and then fine tuned with digital attenuation(again, way faster).
No smoothing on FR graphs.
The impedance is measured with a worthless 100ohm resistor so I won't claim utmost accuracy down to 0.1ohm.

 

[headwhacker]

Excellent post @castleofargh. Was discussing how source OI can affect certain load a certain way but could not find a good source a newbie can easily understand. I think (hopefully) this post will help get them the sense. I agree that impressions should come with the details of the chain used. But sometimes I'm to lazy to mention them until somebody ask about it. perhaps I should start that practice as well.

 

[jeffhawke]

This post should be a mandatory read for anyone joining head-fi, and of course they should take a test on it before being approved

 

[castleofargh]

most people would have explained the same thing better with 1/3rd the words. I'm really bad at making clear short sentences that stay on topic. but thank you for the kind words.

 

[jeffhawke]

most people would have explained the same thing better with 1/3rd the words. I'm really bad at making clear short sentences that stay on topic. but thank you for the kind words.

Not so sure about that... eagerly waiting for the Pringles post

 

[jeffhawke]

On a more serious note, I find the topic of impedance matching fascinating (as an economist rather than a sound engineer everything in this realm is fascinating to me when I think I start to understand), so please post some more on this subject, if you please. Only people who have clear ideas can expose them clearly...

 

[headwhacker]

most people would have explained the same thing better with 1/3rd the words. I'm really bad at making clear short sentences that stay on topic. but thank you for the kind words.

Nah I could not have laid it better. Check the Hiby R6 thread. It’s almost the perfect DAP to me. Caveat is the OI is 10ohms. Still a few tries to dismiss the issue cause they are not affected. So this thread is perfectly in good timing for such discussions.

You save me the effort of googling for a good discussion/reference. The good ones I can find most likely get deleted from my post due to sources outside if this forum.

 

[71 dB]

My opinion which I am not forcing on anyone is this:

If you know the minimum and maximum impedances of your headphone to be Zmin and Zmax and you want the frequency response error to be less than x dB, the output impedance Rout of the amp shouldn't be more than

Rout < (Zmax * Zmin * (beta - 1)) / (Zmax - beta * Zmin)​

where beta = 10^(x/20).

Example: Maximum and minimum impedances of your headphone are 50 Ω and 30 Ω and you want frequency error be less than 0.5 dB. We get:

beta = 10^(0.5/20) = 1.06.
Rout < (50 * 30 * (1.06 - 1)) / (50 - 1.06 * 30) = 90 / 18.2 = 4.9 Ω.

Letting frequency response error to be 1 dB, we have

beta = 10^(1/20) = 1.12.
Rout < (50 * 30 * (1.12 - 1)) / (50 - 1.12 * 30) = 180 / 16.4 = 11 Ω.

There is the 1/8 -rule for output impedance, but it's not very accurate in every situation. Which impedance value of the above headphone should be used when calculating 1/8 rule? People check the specs and see that these are sold as 32 Ω cans, so they calculate 32/8 Ω = 4 Ω which in this case is very close to the 0.5 frequency response error criteria, little better in fact. Some headphones have so flat impedance curve, that any output impedance keeps the frequency response flat enough.

Excessive output impedance causes

- frequency response error
- lack of damping
- linear distortion

Of these three, frequency response error seems to be the most demanding. If a headphone has very little mechanical damping (to be sensitive), it needs electrical damping meaning low output impedance for the amp, but lack of mechanical damping also means strong resonance for the driver which creates strong maximum impedance on the electrical side. So, if you need electrical damping, you also need to reduce frequency response error. Linear distortion seems the least of the problems. Linear distortion here means that the control of the amp on the movements of the driver is compromised so that the movements of the driver don't follow the amp voltage as accurately as when the output impedance is zero.

I use Sennheiser HD-598. It is pretty "demanding" in respect of output impedance. Zmax = 275 Ω, Zmin = 60 Ω. Just 10 Ω of output impedance gives 1 dB boost around 100 Hz, the resonance frequency of the driver. The output impedancies of my DIY headphone adapters are 1 Ω and 2.2 Ω. Lately I have been testing increasing the output impedance a little to "modify" the sound. If I put a DIY "UPOC" crossfeeder crossfeed off in between my DIY headphone adapter and headphones the output impedance raises to about 10 Ω. The sound becomes a bit more bassy and also "more relaxed" which I don't fully understand yet. Going back to normal set up makes the sound ultra detailed in comparison. So, tinkering with the output impedance can be used to optimaze the sound in respect of accuracy of detail, degree of relaxation, frequency response and damping. Damping and degree of relaxation might be connected, but my understanding at this point is lacking. My headphone is overdamped up to 56 Ω output impedance. I have designed an output impedance selector (5, 10 and 20 Ω), but I haven't constructed it yet. I wish I could understand this better. I hate it when I don't understand something. I have my suspicions, but I am not sure about them at all.

 

[bartzky]

@71 dB I'm using the same formula for quite some time now, especially for output impedance recommendations in my reviews. I've made an excel sheet in which I can load my impedance measurements so the max impedance for delta 1 dB is calculated. Record holder till today is the Campfire Audio Andromeda with 0.5 ohm for delta 1 dB
I like to add that beta*(Zmin/Zmax) !< 1 to use this formula. If that's not the case, delta dB will never be reached.

 

[Redcarmoose]

Wow,
maybe it's obvious, but this should be directly under the Head-Fi header. First two things I purchased after reading Head-Fi in 2006 were completely miss matched. I purchased the AKG k701 headphones and a Woo 3 tube amp. In hindsight, I could have never found two devices at more extreme ends of fitting each other. I ended up getting other equipment to make stuff work and essentially two different systems.

But I think the consumer thinks it's like buying cloths or something when they first arrive. Then the reviews come from people with access to only one amp or one headphone and the fables get a going.

 

[headwhacker]

Some

Wow,
maybe it's obvious, but this should be directly under the Head-Fi header. First two things I purchased after reading Head-Fi in 2006 were completely miss matched. I purchased the AKG k701 headphones and a Woo 3 tube amp. In hindsight, I could have never found two devices at more extreme ends of fitting each other. I ended up getting other equipment to make stuff work and essentially two different systems.

But I think the consumer thinks it's like buying cloths or something when they first arrive. Then the reviews come from people with access to only one amp or one headphone and the fables get a going.

I think some call it synergy.

 

[castleofargh]

Nah I could not have laid it better. Check the Hiby R6 thread. It’s almost the perfect DAP to me. Caveat is the OI is 10ohms. Still a few tries to dismiss the issue cause they are not affected. So this thread is perfectly in good timing for such discussions.

You save me the effort of googling for a good discussion/reference. The good ones I can find most likely get deleted from my post due to sources outside if this forum.

on a few occasions where no other problem arises from fooling around with the damping ratio, we can use the impedance of a source as an EQ. a very specific and limited EQ but an EQ still. I've had a rare few IEMs where I personally preferred the signature from a source with higher impedance. so maybe that is what happens to some people and some DAPs? when it comes to preferred signature, it's different strokes for different folks, so you can't expect a systematic consensus on any of those stuff. still, my experience was that I rarely prefer high impedance source or low impedance source with added resistors in series. I liked it for the old westone4 because the default signature just lacked mids too much for my taste. and kind of the same idea for the pair of togo334 to change the bass a little. the famous ER4S with the extra resistors to "increase" the treble(really making everything else lower). and I would probably find 2 or 3 more where I liked extra impedance at the source(or in the cable), and that's about it.

My opinion which I am not forcing on anyone is this:

If you know the minimum and maximum impedances of your headphone to be Zmin and Zmax and you want the frequency response error to be less than x dB, the output impedance Rout of the amp shouldn't be more than

Rout < (Zmax * Zmin * (beta - 1)) / (Zmax - beta * Zmin)​

where beta = 10^(x/20).

Example: Maximum and minimum impedances of your headphone are 50 Ω and 30 Ω and you want frequency error be less than 0.5 dB. We get:

beta = 10^(0.5/20) = 1.06.
Rout < (50 * 30 * (1.06 - 1)) / (50 - 1.06 * 30) = 90 / 18.2 = 4.9 Ω.

Letting frequency response error to be 1 dB, we have

beta = 10^(1/20) = 1.12.
Rout < (50 * 30 * (1.12 - 1)) / (50 - 1.12 * 30) = 180 / 16.4 = 11 Ω.

There is the 1/8 -rule for output impedance, but it's not very accurate in every situation. Which impedance value of the above headphone should be used when calculating 1/8 rule? People check the specs and see that these are sold as 32 Ω cans, so they calculate 32/8 Ω = 4 Ω which in this case is very close to the 0.5 frequency response error criteria, little better in fact. Some headphones have so flat impedance curve, that any output impedance keeps the frequency response flat enough.

Excessive output impedance causes

- frequency response error
- lack of damping
- linear distortion

Of these three, frequency response error seems to be the most demanding. If a headphone has very little mechanical damping (to be sensitive), it needs electrical damping meaning low output impedance for the amp, but lack of mechanical damping also means strong resonance for the driver which creates strong maximum impedance on the electrical side. So, if you need electrical damping, you also need to reduce frequency response error. Linear distortion seems the least of the problems. Linear distortion here means that the control of the amp on the movements of the driver is compromised so that the movements of the driver don't follow the amp voltage as accurately as when the output impedance is zero.

I use Sennheiser HD-598. It is pretty "demanding" in respect of output impedance. Zmax = 275 Ω, Zmin = 60 Ω. Just 10 Ω of output impedance gives 1 dB boost around 100 Hz, the resonance frequency of the driver. The output impedancies of my DIY headphone adapters are 1 Ω and 2.2 Ω. Lately I have been testing increasing the output impedance a little to "modify" the sound. If I put a DIY "UPOC" crossfeeder crossfeed off in between my DIY headphone adapter and headphones the output impedance raises to about 10 Ω. The sound becomes a bit more bassy and also "more relaxed" which I don't fully understand yet. Going back to normal set up makes the sound ultra detailed in comparison. So, tinkering with the output impedance can be used to optimaze the sound in respect of accuracy of detail, degree of relaxation, frequency response and damping. Damping and degree of relaxation might be connected, but my understanding at this point is lacking. My headphone is overdamped up to 56 Ω output impedance. I have designed an output impedance selector (5, 10 and 20 Ω), but I haven't constructed it yet. I wish I could understand this better. I hate it when I don't understand something. I have my suspicions, but I am not sure about them at all.

yup with TOTL multidriver IEMs, knowing the lowest impedance comes down to having somebody online measuring it. because the manufacturer will at best give a value at 1khz and that's not always enough. some don't even seem to give that. I don't know if they aim for an average over a specific band or if they just suck at measuring the impedance value, but sometimes you wonder where their number is coming from^_^.
for headphones, between innerfidelity and a few other places online, chances are we can find the impedance curve. so it's a problem too, but internet really helps. more often than not in the last years I was able to know the lowest impedance of a headphone soon after it came out. for IEMs... well let's just say I ended up soldering a little something to measure it myself. so we can probably do better there.


about an actual sound for overdamping/underdamping, there seem to be a consensus on how headphones don't usually make much of an underdamped sound, at least not like some underdamped big fat speakers can. of course headphones rarely end up being 4ohm and modern amplifiers going above 100ohm are becoming less frequent nowadays. so there is that. maybe we notice bad damping less because it happens less?maybe because they're just so much lighter compared to some speakers and they're just naturally more mechanically damped? I'm not sure with all the headphone designs.
for IEMs, now some reach very low impedance and that's a problem IMO, but more for the lack of stability in FR and because some portable amp sections aren't good at dealing with something like a 5ohm load.

it's probably obvious for you, but I feel that we should keep high impedance amps and amps+added resistors before the IEM, as 2 separated entities. for the IEM and the resulting FR it can be similar indeed and serve as examples for FR changes, but for the amp section it's another story. it's "seeing" the small IEM load vs "seeing" a massive impedance of the IEM+ resistors. if we start looking into distortions and other fidelity variables, then making the distinction seems relevant IMO.

 

[Redcarmoose]

Some

I think some call it synergy.

No at times it’s much more, like putting bike wheels on the front of a car.

Bad synergy...........no, some stuff simply is not made for each other.

 

[castleofargh]

@71 dB I'm using the same formula for quite some time now, especially for output impedance recommendations in my reviews. I've made an excel sheet in which I can load my impedance measurements so the max impedance for delta 1 dB is calculated. Record holder till today is the Campfire Audio Andromeda with 0.5 ohm for delta 1 dB
I like to add that beta*(Zmin/Zmax) !< 1 to use this formula. If that's not the case, delta dB will never be reached.

^_^ I made this a few years back to review the togo! 334.

it was pretty bad(enjoy the totally non linear horizontal axis). plus I didn't actually know what impedance was used for Rin Choi's measurements. but it gave some vague idea about the changes in signature at least. I remember spending like 10mn on the formula making plenty of assumptions to keep it simple. and almost a full day googling to find out how to make stuff happen in excel.

 

[71 dB]

@71 dB I'm using the same formula for quite some time now, especially for output impedance recommendations in my reviews. I've made an excel sheet in which I can load my impedance measurements so the max impedance for delta 1 dB is calculated. Record holder till today is the Campfire Audio Andromeda with 0.5 ohm for delta 1 dB
I like to add that beta*(Zmin/Zmax) !< 1 to use this formula. If that's not the case, delta dB will never be reached.

Wow!
Of course, Zmax must be creater than beta*Zmin or negative Rout is needed to produce the frequency response error limit.