Real Load THD - a somewhat philosophical question...

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Consider the two multi opamp buffers in the first image: they have - more or less - the same current output capability on low Z loads, but very different output impedances - the first one sits at about 2.5 ohm, while the second one is in the tens-of-milliohms range (up to reasonably high frequencies, at least). THD on resistive loads is the same too (~0.0008%-0.0009% @ 1Vrms out on 30 ohms), but when switching to * real * loads the high-Zout one shows a dramatical (~ 2 orders of magnitude at low frequencies) increase in THD - see images 2 & 3.

Well, if you drive a non-linear load through an * ideal * distortionless sinusoidal voltage source, the voltage on the load is undistorted even if the current flowing in it is far from being sinusoidal; if you switch to a * real * distortionless sinusoidal source (ie a distortionless source with finite Zout), the higher its output impedance, the more the voltage on the load is distorted * by the load itself *; sounds ok, doesn't? It seems only a matter of output impedance, and opamps, feedback and so on have no responsibility at all - the same holds for a no GFB ampifier I'm playing with: it is actually a discrete CF architecture which seems to work very well open-loop - I'm managing to build a HP amplifier out of it with the option to run it in both ways. Its open-loop output impedance is ~2.5 ohms, and when driving real loads it shows the same THD increase that I see with the opamp buffer; closing the loop lowers the Z out down to ~50milliohm, and there's no THD raise at all when switching from resistive load to real one.

Given that headphones and louspeakers * do are * nonlinear loads, my question is: the THD I see when driving real load through high Z buffers is simply the THD of the load itself? Or a low damping factor worsens overall LF THD too? I know the question may look pretty naive, but maybe I finally got why noone manages to perform real load tests on audio stuff...

Ciao,

L.
 

Attachments

  • OpAmpBuff.png
    OpAmpBuff.png
    61.2 KB · Views: 236
  • 4x5532highZoutTHD.png
    4x5532highZoutTHD.png
    37.5 KB · Views: 234
  • 4x5532lowZoutTHD.png
    4x5532lowZoutTHD.png
    34.8 KB · Views: 223
Last edited:
Hello,
I think that the THD seen when driving a real load through high Z buffers is likely to be the THD generated by the non linearity of the load itself.

It would be interesting to know what load you used to generate the plot.

Crossover inductors using a magnetic material aren't linear. I haven't seen a graph of inductance versus current for commercially available inductors making it difficult to select one for a particular power level. Possibly insufficiently rated inductors could be the reason behind stories like some loudspeakers are difficult to drive.
The capacitors are often electrolytic. Again non linear and it's probably not too difficult to exceed the current ratings at high levels. I guess that film capacitors are likely to be much less of a problem. I used polypropylene but they were expensive. Polyester are reputed to have some non linearity issues.
The actual transducers are also a non linear load. The inductance isn't linear. The cone movement hence back emf isn't linear either.
I guess that for the lowest acoustic THD the loudspeaker should be driven from a low impedance. But I guess that any electrical non linearity will be low compared to any mechanical / electromechanical non linearity.
I guess that any loudspeaker manufacturer would be happy with THD < -70 dB. Not many specify it.
I think that it's difficult / impossible to test an amplifier at the maximum power output using a real load because few loudspeakers would survive for long. Also as you imply the amplifier THD would be increased by the real load. If the loudspeaker survived the voice coil would heat up and the resistance increase changing the distortion measurement as well.
 
It would be interesting to know what load you used to generate the plot.

Hi PChi, the 'real load' was Denon AH-D310 headphones (the cheapest ones in the AH-D Denon's line), whose nominal impedance is 32 ohm. Given their sensitvity (105dB/mW), at 0dbV (~31mW) input the sound pressure is fair high and quite hard to sustain for a long time (with pure tones in the kHz range, at least) - the test was quite an extreme one. Reducing the drive lowers the THD, and the same holds if you try with higher Z headphones.

Ciao,

L.
 
Hi PChi, the 'real load' was Denon AH-D310 headphones (the cheapest ones in the AH-D Denon's line), whose nominal impedance is 32 ohm. Given their sensitvity (105dB/mW), at 0dbV (~31mW) input the sound pressure is fair high and quite hard to sustain for a long time (with pure tones in the kHz range, at least) - the test was quite an extreme one. Reducing the drive lowers the THD, and the same holds if you try with higher Z headphones.

Ciao,

L.

Hello L,
Thanks for the answer about the real load. At least that removes the crossover from the circuit. A quick look at 'Testing Loudspeakers' by Joseph D'Appolito section 7.3.4 describes drivers as weakly non linear devices. Section 2.6.1 mentions that parameters measured under large excursions will differ from their "small signal" value.
I have an ancient Hi-Fi Choice booklet on Cartridges and Headphones that doesn't appear to have any data on distortion and the 'Loudspeaker and Headphone Handbook' also ignores the issue!
 
Looking at your schematic, it seems the main difference between your two circuits is that the "low output Z" circuit has the last op amp directly powering the load, while the others are connected via their own 10 ohm resistor, and then the combined signal goes through another 10 ohm resistor. Is that a mistake, or intended?

-Charlie
 
Last edited:
Looking at your schematic, it seems the main difference between your two circuits is that the "low output Z" circuit has the last op amp directly powering the load, while the others are connected via their own 10 ohm resistor, and then the combined signal goes through another 10 ohm resistor. Is that a mistake, or intended?

-Charlie

Hi Charlie, you're right - the schematic of the second buffer is wrong: its output actually is the upper common node of the 10 ohm resistors; I'm sorry, but drawed it in a few minutes, and something went wrong with the cut'n'paste as usual...

Anyway, nothing really new - this load sharing technique is described in an old NS Linear Brief written by Bob Pease - just a clever method to get more power out of opamps without increasing output impedance (with the parallel buffer a-la Doug Self you need far more opamps, or pretty low balancing resistors, to keep output Z low) - actually it works very well, although with some opamps it may need a damping RC network in order to prevent instability (it seems unnecessary for the 5532s).

L.
 
(...) A quick look at 'Testing Loudspeakers' by Joseph D'Appolito section 7.3.4 describes drivers as weakly non linear devices. Section 2.6.1 mentions that parameters measured under large excursions will differ from their "small signal" value.
I have an ancient Hi-Fi Choice booklet on Cartridges and Headphones that doesn't appear to have any data on distortion and the 'Loudspeaker and Headphone Handbook' also ignores the issue!

It seems it's a very neglected topic - I've spent a few hours googlin', but found nothing really interesting or useful (except a couple of AES papers that aren't freely downloadable...:().

Ciao,

L.
 
I agree that Loudspeaker distortion both acoustic and load impedance is neglected but very important.

Douglas Self mentions in passing the effect of loudspeaker load non linearity in Chapter 8 of the Audio Power Amplifier Design Handbook.

I have only briefly measured acoustic harmonic distortion of poor quality speakers used in Car Handsfree units. It was possible to measure many 10s of percent at certain frequncies.

High Performance Loudspeakers by Martin Colloms includes a copy of a graph for a LS5/5 loudspeaker showing acoustic harmonic distortion levels of around -60 dB at 1 kHz and considerably more at lower frequencies. It also has a graph for a low distortion Yamaha Monitor Loudspeaker.

I guess that the measurements are difficult and don't look good so aren't available. The only hope is that an independent reviewer takes a scientific approach to testing loudspeakers which would encourage an improvement in THD performance both acoustic and load with better linearity.
 
Benchmark Media's John Siau discussed this a while back.

It is interesting that an output impedance of just 2.5 ohms (DF of ~12..13, which would normally be considered ample for headphones) would provoke such a marked increase in distortion.

What one should do is compare this to acoustic distortion. If this is much bigger than on the electrical side for each harmonic (at least 10 dB), further discussion is probably moot. Then you're basically seeing the driver's reciprocity in action, i.e. it feeds a distorted signal back which partly drops over output resistance.

Ultimately the decisive question should be whether output impedance has any effect on distortion on the acoustic side. (Which, as you'll hopefully agree, is where it ultimately counts.) If it doesn't, one can concentrate on its well-documented linear effects. Those are likely to be kicking in well before any degradation in nonlinear distortion are to be observed, though I guess the details would depend on the headphones in question (e.g. stiffly suspended AKGs vs. Sennheisers on the other end of the spectrum).
 
Last edited:
The subject is not distortion produced by mechanical systems such as loudspeakers when driven by a distortion - free current which yields the mechanical drive force B*l*I , but reactive distortion produced by amplifier. Insofar thd figures for a real load are very far from telling the truth about the audio qualities.
Much recent research has been published on this subject . That eventually led to amp topologies without a global NFB. Good test magazines always publish reactive distortion in test reports.
Psychologically - that is what matters - the total figure of thd matters far less than spikes of thd at particular frequencies due to minimum impedance with maximum phase shift of passive networks of passive speakers.
Thus, an amp having a thd of 1% without reactive distortion, and a harmonics spectrum
independent of frequency at complex loads, sounds audibly better than a super duper
low thd with a harmonics spectrum which depends on the complex load, i.e. varies with frequency.
This is why a triode amp with 5% thd sounds audibly better than a super duper MOSFET with 0.01% thd. Triode amps don't produce any reactive distortion. The human auditive perception is extemely sensitive to changes in spectrum of harmonics because human voice expresses "between words" either affection or rejection with even very small changes in harmonics spectrum.
 
hahfran said:
Triode amps don't produce any reactive distortion.
Actually, the opposite is true. Do some reading on the extra distortion produced by elliptical load lines. Any decent valve book will cover this topic. Heavy NFB, correctly applied, will reduce an amplifier's output impedance and potentially makes it less load-dependent. If triode amps sound better, it isn't for this reason.
 
and triodes have lower output impedance in the 1st place from internal negative feedback

in fact negative feedback is the only practical means of achieving low output impedance - it may be local or internal but all of our "low output impedance" subcirucits employ negative feedback

"the other" option is to load with a resistor of lower value than the load, wasting many time the power vs that delivered to the load
 
Hi sgrossklass;


Thank you for the reference - Mr. Siau is clearly trying to promote his low Z-out amplifiers, but it seems the results he shows are in pretty good agreement with what I've seen.

Ultimately the decisive question should be whether output impedance has any effect on distortion on the acoustic side. (Which, as you'll hopefully agree, is where it ultimately counts.)

Yeah - You've got the point: this is exactly what I'm wondering about. I've got a couple of ideas on how to set up a very qualitive test to investigate correlations between acoustic THD and - let's say - electrical THD at the amplifier's output, but had no time to try them. Anyway seems to be quite a difficult test to do without dedicated instruments...

L.
 
The subject is not distortion produced by mechanical systems such as loudspeakers when driven by a distortion - free current which yields the mechanical drive force B*l*I , but reactive distortion produced by amplifier. Insofar thd figures for a real load are very far from telling the truth about the audio qualities.
Much recent research has been published on this subject . That eventually led to amp topologies without a global NFB. Good test magazines always publish reactive distortion in test reports.

Hi hahfran, I'm afraid I can't agree with you - GNFB or THD on reactive loads are definitely not an issue in the present case. The output impedance of the ** zero-GNFB ** amplifier I'm working on is about 2.5 ohms, and when driving real loads it shows a fair high THD raise especially at LF; as already pointed out in my first post, closing the loop lowers the output impedance down to 50 milliohms, and THD drops to almost unmeasurable levels on whatever load (~0.0008% @ 1Vrms out, 100Hz on ~30 ohms, on both purely resistive load and real load). Besides, the impedance curve of the Denon employed in the test shows it is a very light load, with an almost resistive behaviour up to a few kHz (see attached graph).

As far as the opamp buffers is concerned, it's quite hard to talk about overloading when those little beasts can drive a ~30 ohms load to 0dBv with 0.0008% THD (see images in the first message), or a ~100 ohms load to 10dBv with 0.0005% THD (look at the attached spectrum).

Ciao,

L.
 

Attachments

  • Denon AH-D310.png
    Denon AH-D310.png
    37.2 KB · Views: 79
  • 10dBV-100ohm.png
    10dBV-100ohm.png
    35.4 KB · Views: 83
Yeah - You've got the point: this is exactly what I'm wondering about. I've got a couple of ideas on how to set up a very qualitive test to investigate correlations between acoustic THD and - let's say - electrical THD at the amplifier's output, but had no time to try them. Anyway seems to be quite a difficult test to do without dedicated instruments...

In the case of loudspeakers (I don't know about headphones) current drive (i.e. high output impedance) produces lower acoustic distortion at low frequencies. This must correspond to higher measured voltage THD at the terminals. Bruno Putzeys says he uses the effect to lower distortion on his Grimm active loudspeaker.

Plenty of interesting discussion on the topic in this thread: http://repforums.prosoundweb.com/index.php?topic=33953.0
 
Last edited:
A few tests on the amplifier I'm working on/still playing with. The first graph is a comparison between THD on resistive and real load in closed loop configuration @ 0dBv out on ~30 ohm (gain ~ 12dB, BW ~ 300kHz) - THD is < 0.001% in both cases, and the spectra are almost undistinguishable. The 2nd spectrum is acquired with the amplifier in open loop configuration (@ ~ 10dB gain, ~300kHz BW) driving a ~ 30 ohm resistive load; same setup for the last one, but now the amplifier is driving a real load (the mentioned Denon AH-D310). As you can see, THD raise when driving real loads is quite impressive (although THD itself is still relatively low). Once more, NFB is definitely * not * an issue, while output impedance * does matter * quite a lot.

Ciao,

L.
 

Attachments

  • CMCLAMP - load comparison.png
    CMCLAMP - load comparison.png
    34.2 KB · Views: 52
  • CMOLAMP - resistive load.png
    CMOLAMP - resistive load.png
    31 KB · Views: 50
  • CMOLAMP - real load.png
    CMOLAMP - real load.png
    31.3 KB · Views: 52
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.