| percy |
When I measure the frequency response on my PC sound card using RMAA in loopback mode (line in=>line out) it measures ruler flat from 20hz-20Khz. But as soon as I put a load on it - like a headphone(33ohm) - the low freqency response starts rolling off as early as 70-80hz.
What could be the potential causes for this ? Could it be the output current capability of the output opamp ? If the opamp is capable of the output current that what could be the next possible cause ?
The output opamp is a TLV2465C http://focus.ti.com/docs/prod/folde...nt/tlv2465.html |
|
|
| neutron7 |
| does it roll off at all volume levels? or does it get worse as you increase it |
|
|
| sangram |
| quote: | Originally posted by percy
What could be the potential causes for this ? |
Usually it would be the size of the output capacitor causing the rolloff.
I can't think of anything else actually. |
|
|
| tiltedhalo |
| Depending on many factors, frequency impedance, power and induction of the head phones and all the other variables. You get factor loading effect short of like a engine on a dyno 300hp at 5200RPM, but load the dyno and poof change of HP and torque due to RPM and load, same type of physics applys to electronics. |
|
|
| Ouroboros |
| As your sound card will be using a single +3.3v or possibly +5V rail, it will use an output coupling capacitor to block the dc bias on the output from getting to the headphone output. With low-impedance 32R phones, you'll need at least 470uF to get good LF response. The output cap on most sound cards is much lower than this, resulting in a first-order roll-off of the bass. |
|
|
| AndrewT |
Hi,
is the sound card intended for driving a 32ohm headphone load?
An output of 1Vrms (about31mW) is 44mA peak current. |
|
|
| sreten |
| quote: | Originally posted by percy
Could it be the output current capability of the output opamp ?
If the opamp is capable of the output current that what could
be the next possible cause ? |
Hi,
No, that would cause massive amounts of distortion.
| quote: | Originally posted by sangram
Usually it would be the size of the output capacitor
causing the rolloff. I can't think of anything else actually. |
Seconded.
:)/sreten. |
|
|
| percy |
Thank you all. Your responses deserve some more information from me. Here are some pictures of the frequency response and distortion also if that helps give a better idea.
The freq response drops equally across the entire range - with an added rolloff in the low frequency range.
The output caps are all 100uf.
I dont know the recommended output impedance of the soundcard's output but the manual does say I can connect heaphones (and there is a headphone option in the windows/soundcard driver's mixer also which I dont think is doing anything). Oh btw its a Turtle Beach Santa Cruz card.
Its almost like a classic 1st order rolloff dont you think ?
Left (white) channel is the one connected to the headphone. Right (green) is looped directly to give you can idea of both the base soundcard and under load together. |
|
|
| percy |
Here's some voltage measurements I took at a fixed frequency (80hz) across different resistor loads to give you an idea of the output impedance -
Open ------ .455Vac
200ohm --- .426Vac
33ohm ---- .315Vac (the actual headphone)
10ohm ---- .177Vac |
|
|
| Ouroboros |
| 100uF into a 32 Ohm load will give a -3dB point at 50Hz. |
|
|
| percy |
| would appreciate if anyone can also share some thoughts on the increased distortion and imd under load. |
|
|
| AndrewT |
Hi Percy,
those voltages into the various loads are telling you that the output stage is NOT designed to drive either 10r or 32r.
I would think from the Vdrop into 200r that the stage is designed to drive 300ohm headphones taking account of the reactive load they present. |
|
|
| lineup |
| quote: | Originally posted by Ouroboros
100uF into a 32 Ohm load will give a -3dB point at 50Hz. |
You should change output cap to 470uF (10Hz rolloff) or even 1000uF (5 Hz rolloff).
Use best quality electrolytic (see below) cap you can find.
You can use two 220uF or 470uF in parallel.
Paralleling will add the capacitance:
Two 220uF = one 440uF.
100uF parallel with 220uF = 320uF.
OSCON Caps would be suitable. They are non-liquid with longer lifetime.
They have better audio qualities, than liquid electrolyt caps.
OSCON are only available for lower voltages. ~10-20 Volt.
As your circuit is low volt, you will be able to find and use 10 Volt OSCON capacitors. |
|
|
| sreten |
| quote: | Originally posted by percy
The freq response drops equally across the entire range
- with an added rolloff in the low frequency range.
|
Hi,
the 3db loss indicates output impedance ~ = headphone impedance.
:)/sreten. |
|
|
| percy |
ok. sorry for the stupid question but -
What element of an amplifier's design determines the output impedance of the amplifier ? or in other words which factors determine the optimum load for the amplifier ?
When I try to think through it, somehow I just end up visualizing that it depends on the current delivering capacity of the output stage(?) But then again the opamp has decent output current rating and I dont think is being exceeded... |
|
|
| EWorkshop1708 |
First of all, the Frequency response...........
It's the output caps, like everyone is saying. Only reason the soundcard uses those caps, is because the output of a single supply amp (single 12V supply in computer) has DC on the output of the amp, and that has to be filtered out with capacitors. However the size of the capacitor determines how much bass you get.
If you want to understand that more, take some caps, 1uf, 10uf, 100uf, and 1000uf, and put each one at a time in series with a 4 ohm speaker. You will notice the smaller caps allow mostly treble to flow while going to larger caps allow voices to play, then even larger caps allow bass. Now if you switch to an 8 ohm or higher speaker, you will also get some more low frequencies to play.
Same is happening with your headphones and soundcard. When you feed line in, you may be driving 10K to 22K ohms with those 100uf caps. Since that's hardly much of a load to drive, nearly all of the low frequencies go through. However, when driving a 32 OHM load, which is a HUGE difference, you lose some bass because your caps are not so big for a 32 ohm load, vs drving a load of several thousand ohms.
To keep it simple......
A smaller capacitor charges/discharges faster, so smaller caps filter out some low frequencies because they charge up before the low frequency wave is done. When a cap is fully charged, no current flows. With a large cap that takes a lot longer to charge, the wave is not slow enough to fully charge it, so most or all of it gets played. Also the resistance that the capacitor is feeding, determines how fast the cap charges and also determines the low frequency limit. Just like your 32 ohm headphones vs your soundcard input impedance.
Now if you are concerned with the output current, you can always connect a small amp to the soundcard. If you think the bass sounds distorted with headphones, that would indicate lack of current, but if it's clear sounding, then it's not a problem. Try raising the 100 uf to 470 or more and see if your bass sounds clear; if not, then make a headphone amp instead with better output current than the soundcard.
BTW output impedance/current drive capability is determined by the output transistors in the amp, it's bias current, and also by what output protection the amp has, and what mA it's limited at. |
|
|
| percy |
actually I already understood the low frequency rolloff and how the output caps are affecting it. Now I am trying to understand what contributes to the "overall" loss in level - across the rest of the frequency band - like sreten said - it gets shelved -3db down for a 33ohm load. Its about -1db down for a 200ohm load.
Again I understand its because the output stage/opamp is not capable of driving that high load(low impedance), in other words is not able to cope up with the extra current demand. But why ? I guess what I am trying to understand - and what my question should be is - What could cause output current limitation even if the device (the opamp in this case) has good current rating. And hence my comments in the previous post -
| quote: | What element of an amplifier's design determines the output impedance of the amplifier ? or in other words which factors determine the optimum load for the amplifier ?
When I try to think through it, somehow I just end up visualizing that it depends on the current delivering capacity of the output stage(?) But then again the opamp has decent output current rating and I dont think is being exceeded...
|
|
|
|
| AndrewT |
Hi Percy,
I do not know what device forms your output stage in the sound card.
I would be very surprised if the line out was designed for much under 1k0 load impedance. More likely 10k.
If a heaphone output is provided I would expect it to perform better than your voltage measurements seem to indicate. It is NOT suited to low impedance loads.
I still believe that your output stage is designed to drive 300ohms or higher and this is starting to get close to the line out spec. I mentioned above. |
|
|
| percy |
both lineout and headphone are the same jack.
Device is:-
|
|
|
| Eva |
Most old sound cards had four jacks: Speaker out, Line out, Line in, Mic in. The speaker output employed a separate amplifier IC and gentle output capacitors, it was indeed intended to drive 8 ohms and it worked nice as a quality headpone output.
However, nowadays computers and its peripherals are suffering from strong cost cutting and size reduction, so soundcards no longer have an amplified output. Instead they just have a single non-specific op-amp driven output jack where people connects everything. Obviously, gentle amounts of distortion are produced when this output is asked to drive headphones, but nobody seems to care. It's much easier to find someone complaining about that distortion because he has measured it than someone complaining because he perceives the sound distorted on the headphones (by the way, this speaks a lot about our ability to perceive massive distortion). |
|
|
| neutron7 |
re TLV2465
if you look at figure 34 on
http://focus.ti.com/lit/ds/symlink/tlv2465.pdf
they show the distortion for 250 ohm load. as the peak to peak signal reaches close to the railss it gets pretty high for even 250ohms. they probably were too emabarassed to even show 33 ohms :)
I am assuming the sound card is using 3.3v standard PCI voltage for supply. if its 5 volts you will be slightly better off.
even if you could get more power at low frequencies by using a bigger capacitor it would not be very pleasant! i guess the distortion would get close to 10%
i think you would be better offf to build one of the many mini headphone amps such as Cmoy, ad815, or one of the bazillion other variants.
I believe the reason they did it like that is not just to save money, but because these days most PC speakers are self amplified. the chips not too bad when its driving 10kohm |
|
|
| Eva |
| quote: | Originally posted by neutron7
re TLV2465
I believe the reason they did it like that is not just to save money, but because these days most PC speakers are self amplified. |
Well, but there is also a lot of people using headphones. So, where do I connect mine? These output jacks usually have a "headphone" icon near them, yet they are intented to drive not much less than 5Kohm and distort when any headphone is connected.
There would be an additional jack intended to drive headphones or speakers, as there was in almost all old ISA soundcards, if that kind of equipment wasn't nowadays suffering from so extensive cost cutting. Note that the PCI slot also provides 12V power, so including a built-in amplified output is a matter of adding some capacitors, a boom-box IC and another output jack. |
|
|
| neutron7 |
| quote: | Originally posted by Eva
Well, but there is also a lot of people using headphones. So, where do I connect mine? These output jacks usually have a "headphone" icon near them, yet they are intented to drive not much less than 5Kohm and distort when any headphone is connected. |
You are supposed to be a good little consumer sheep and buy DRM infected music from itunes and put it on your ipod :)
It would be an iteresting project to check what output chip various soundcards have so they can be reccomended. for headphones. |
|
|
| Eva |
Due to cost cutting, most consumer soundcards doesn't have any particular output chip, actually they are single-chip solutions where everything comes in a QFP SMD IC.
On the other hand, I own a SB AWE32 from 1994 where you can see everything in the huge PCB (analog mixer, DSP w/ADC+DAC, mic pre-amp with AGC, output amplifier IC, wavetable CPU w/DAC, wavetable ROM and two SIMM sockets for wavetable RAM where I have 4Mb :D ). I also own a Gravis Ultrasound from 1994 that follows the exact same topology (mixer, in/out amplifiers, ADC/DAC, CPU and 1Mb RAM).
Unlike modern clumsy PCI cards that are just made of a poor ADC/DAC fed synchronosuly from the PCI bus without any buffering, these old cards can play 32 samples simultaneously without using any CPU cycles, bus cycles or a single byte of main memmory. |
|
|
| AR2 |
While we are on the subject of impedances, could someone tell me what is the imput impedance of this circuit and what elements are determining it?
Would that be combination of R36 and R13 with C29 and C15 and for negative rail R37 and R153 with C30 and C84?
I think that impedance is hardest to grasp and understand, so please elaborate. |
|
|
| AR2 |
| Attachment didn't go. Another try |
|
|
| AndrewT |
Hi Ar2,
I am new to this imaginary gobble de gook.
Here goes; seen from x13/14:-inZ14 = 2K +1k18//4k99 = 2k95
but from x13/13:- Is 1k18 feeding a high impedance? I think yes. the inZ13 = 2k+4k99=6k99.
Then one has to add in the capacitor effect. I think the 390pF after 1k18 is swamped by the 2n2F before 1k18.
But 2n2F is isolated from the input by 2k in both cases. So impedance at DC is as above falling to 2k at infinite frequency. At 20kHz 2n2F has an effective impedance of 3k6 but it is not in phase with the resistance so simple // and series cannot be used. As the reactance gets smaller say above 50kHz the capacitor starts to dominate and then it's 2k in series with 2n2F but for how you write this and use it, you need an expert, not me.
How do you analyse the differential impedances and the common mode impedance? |
|
|
| AR2 |
Andrew thank you.
As I mention, understanding and callculating impedance is too complex for me.
"Here goes; seen from x13/14:-inZ14 = 2K +1k18//4k99 = 2k95"
What is the callculation here - are you deviding 2K +1K18 by 4K99? |
|
|
| Eva |
Common mode input impedance is 3.5k (2k+4.99k from each leg, then placed in paralell).
Differential mode input impedance is 4K (2k+2k from ecah leg in series).
This only applies to frequencies where the capacitors still show high impedances, though. |
|
|
| AndrewT |
Hi Ar2,
the // sign means in parallel with.
so 4k99//1k18 has a resistance = 1/[1/r1+1//r2] = 954ohms.
An alternative that many use is r1*r2/[r1+r2] = 954ohms.
now just add 2k0 + 954r = 2k954.
This only applies to the inverting input which has an effective zero volts (virtual earth) at it's input (pin2).
The input impedance at pin3 is the opamp impedance and this is usually over 100k sometimes much over.
BUT this amp is connected as a differential and the normal rules for virtual earth (pin2) and high impedance (pin3) inputs may no longer hold. The reason I doubt the standard rule is that the voltage difference at the input pins is the output voltage/opamp gain which defines the volts at pin3 as no higher than about 50uV, pretty near zero but changing with signal.
However no one has jumped in to correct me yet!! |
|
|
| AR2 |
Thank you Eva and AndrewT.
I am trying to understand how to callcullate impedance and your help is greatly appreciated. For the curious ones the circuit is part of output portion of Behringer DCX 2496 - right after DAC. I am sorry if I interupted Percy's tread. |
|
|
|
