myhrrhleine said:
I think most people are deluding themselves with respect to attributing “transparency” to an amplifier that produces perceptible levels of distortion.
The added harmonic and intermodulation components may very well give the impression of greater detail, but this detail is added, not “uncovered”, “unmasked” or “unveiled” from the recording and an amplifier that does such is definitely not transparent (or the better of the two by the subjective opinion of everybody).
Some people like this added "detail" and some even have quite a fondness of open loop SET amps that product double digit THD and IMD figures. Some people also like to eat Vegemite.
these are just armchair engineer "guestimates" but some details of the circuit look a little short of optimum for the topology to me
diff pair bias current seems a little hot, while the buffer to the VAS is perhaps underbiased?
C10 should be to the neg rail not gnd
C6,7 should be on the other side of R15, connected to gnd - the large body of the electro is an antenna and is connected directly to your input, adds parasitic C to gnd where you don't want it, could pick up other noise too
R19,20 value seems too large, more current in driver Qs would speed things up by reducing drive impedance
drive Q11,12 as mentioned above are way slow for the position in addition to being under biased, I'd look for 5-10x faster than output Q for the drivers
I'm unfamiliar with the Baxandall diode operation but 1n4148 are high speed switching diodes with large-ish series R - maybe not what’s wanted in this position either
sometimes a single digit pF size feedback "lead" cap is helpful - needs tuning for best pulse response - temporarily jumper around R10 to push the input fast enough
diff pair bias current seems a little hot, while the buffer to the VAS is perhaps underbiased?
C10 should be to the neg rail not gnd
C6,7 should be on the other side of R15, connected to gnd - the large body of the electro is an antenna and is connected directly to your input, adds parasitic C to gnd where you don't want it, could pick up other noise too
R19,20 value seems too large, more current in driver Qs would speed things up by reducing drive impedance
drive Q11,12 as mentioned above are way slow for the position in addition to being under biased, I'd look for 5-10x faster than output Q for the drivers
I'm unfamiliar with the Baxandall diode operation but 1n4148 are high speed switching diodes with large-ish series R - maybe not what’s wanted in this position either
sometimes a single digit pF size feedback "lead" cap is helpful - needs tuning for best pulse response - temporarily jumper around R10 to push the input fast enough
Going back to the original post, I suspect there is some point below which most people can not detect a further reduction in THD. It probably varies by person and maybe something that is impacted by training. This needs to be considered.
Also I suspect the subjective experience of more or less "clarity" may be more related with S/M ratio and is perceptible as greater or lesser dynamic range. Ambient background noise in the listening environment could could play a part in the degree to which someone can detect differences.
Finally, THD is additive if the THD of the loudspeaker is 1% (a pretty good figure) then a difference in the amplifier of .05% vs .005% will be an overall difference of 1.05% vs 1.005%; are approximately 1%. THD in the preamp and source part of the chain may further obscure differences.
Therer is a question of which harmonics are lower -- the higher order the harmonics, the more offensive. However, most of the time as you look at the spectra the hight of the harmonic decreases as the order increases. Note the word "most".
Also I suspect the subjective experience of more or less "clarity" may be more related with S/M ratio and is perceptible as greater or lesser dynamic range. Ambient background noise in the listening environment could could play a part in the degree to which someone can detect differences.
Finally, THD is additive if the THD of the loudspeaker is 1% (a pretty good figure) then a difference in the amplifier of .05% vs .005% will be an overall difference of 1.05% vs 1.005%; are approximately 1%. THD in the preamp and source part of the chain may further obscure differences.
Therer is a question of which harmonics are lower -- the higher order the harmonics, the more offensive. However, most of the time as you look at the spectra the hight of the harmonic decreases as the order increases. Note the word "most".
nitrate said:
I know what that sounds like, I've heard it often enough myself. That's the result of excessive gNFB. Horsing up the gNFB can get you some impressive THD numbers, but you lose detail in the process.
Hi Nitrate, Thanks for posting the circuit. Can I ask what it is you don't like about the sound. I find this very interesting as it mirrors my own findings over the years. All these designs are just so similar and are based on the same topology. It's that same "family" sound again, very clear very detailed but totally lacking in emotion. Are you talking about the subtleties here, or is there some fundamental problem. PCB layout is just as important as the design itself particularly correct grounding and the feedback take off and return points. As I say, it mirrors my own findings and in the end I opted for a totally different topology. I agree with with the others over the TIP41/2, much to slow and poor beta.
Regards Karl
Regards Karl
sam9 said:
Not necessarily. There is at least one outfit that uses compementarity in amp and speaker THD to get some kind of 'cancellation' (he calls it 'compensation')
Look at the audiopax website.
Jan Didden
Hands off my TIP's LOL,
I use tip41/2 for their linearity and the fact i have lots of them lying around. They make excellent sounding output devices on thier own too. Try the if you don't belive me 🙂
Anyway forget the published design, its what mechanism is causing my amp to sound like it has a lack of detail that interests me. I think IMD has a lot to do with it. Sombody mentioned earlier that you can check for this by mixing two sines in the amp. Can anybody go into more detail on this as i have a good function genarator at home that i can use to genarate complex signals with.
Regards
Leigh
I use tip41/2 for their linearity and the fact i have lots of them lying around. They make excellent sounding output devices on thier own too. Try the if you don't belive me 🙂
Anyway forget the published design, its what mechanism is causing my amp to sound like it has a lack of detail that interests me. I think IMD has a lot to do with it. Sombody mentioned earlier that you can check for this by mixing two sines in the amp. Can anybody go into more detail on this as i have a good function genarator at home that i can use to genarate complex signals with.
Regards
Leigh
nitrate said:
If you amp is built properly and has very low measured THD throughout the audio band, it will likely have very low IMD as well, so you are likely chasing a red herring.
In your opening post you wrote:
nitrate said:
How high is "relatively high"? If you have grown used to listening to a coloured amplifier, then listening to the same recorded material via an amplifier that doesn't add perceptible levels of distortion may very well give the impression that the amplifier is lacking something over the other, even so it is working perfectly well.
My car audio system has a frequency response and audio quality no where near as good as my home CD player with headphones.
But when I bring a CD in from the car after it's been there for a month or so and listen to it through the headphones inside, the tunes never sound as good (lacks something woud be a good description) as they did in the car, until I've grown accustomed again to the sound through the headphones.
When I take the material back to the car after this, first impressions are that it seems to lack something over the home system and headphones.............
RMAA has IM measurments as a standard feature. I aksed earlier what version you have?
Some version have a fixed setting I think but some earlier version offered you to set any tones and levels you like.
You just loop the signal as you use to.
I'm not sure you can use an external generator and analyse IM live.. however you can record a wav file of the intermodulated output of the amp with your soundcard and imprt that in a second into RMAA and voila you have a nice spectrum view.
Another alternative as Imentioned earlier is to download the excellent software ARTA which lets you analyse any signal directly iow. you can use your tone generator. OTOH ARTA has a built in generator that you can use for singel, dual and multi sine measurements.
Make sure you don't set the signal to high internal though as that may lead to clipping the DAC or ADC and give you false readings. Set the individual tone at -12dBFs and you'll be safe.
If you use the tone as Jan mentioned make sure you also go higher and use 9kz+10kHz and also 19kHz +20kHz.
/Peter
Some version have a fixed setting I think but some earlier version offered you to set any tones and levels you like.
You just loop the signal as you use to.
I'm not sure you can use an external generator and analyse IM live.. however you can record a wav file of the intermodulated output of the amp with your soundcard and imprt that in a second into RMAA and voila you have a nice spectrum view.
Another alternative as Imentioned earlier is to download the excellent software ARTA which lets you analyse any signal directly iow. you can use your tone generator. OTOH ARTA has a built in generator that you can use for singel, dual and multi sine measurements.
Make sure you don't set the signal to high internal though as that may lead to clipping the DAC or ADC and give you false readings. Set the individual tone at -12dBFs and you'll be safe.
If you use the tone as Jan mentioned make sure you also go higher and use 9kz+10kHz and also 19kHz +20kHz.
/Peter
Re: Re: Pros and Cons of THD measurement
Hi Fotios,
SMPTE IM definitely has value, and of course so does CCIF IM, especially when the latter is done with a spectrum analyzer to capture the odd-order products. I have long been an advocate of various kinds of IM distortion testing for the very same reason you gave: that the distortion products typically fall in-band.
In fact, you may want to read my paper "A fully in-band multitone test for transient intermodulation distortion" that is on my web site at www.cordellaudio.com. That MIM test used a combination of three tones (9.0, 10.05, and 20 kHz) to fold both even and odd IM products down to around 1 kHz. The instrumentation was fairly inexpensive.
One thing you have watch out for with SMPTE IM is that it is fairly insensitive to HF nonlinearity and moreover it can yield very low numbers for amplifiers that have a lot of negative feedback at the in-band frequency of 7 kHz. Of course, one could argue that the magnitude of IM products in-band is what we hear, and that the NFB is just doing a very good job in that case. The same can be said for my MIM test, where the distortion products are around 1 kHz and many amplifiers have a lot of distortion-reducing negative feedback in play at 1 kHz.
Cheers,
Bob
fotios said:
Hi Fotios,
SMPTE IM definitely has value, and of course so does CCIF IM, especially when the latter is done with a spectrum analyzer to capture the odd-order products. I have long been an advocate of various kinds of IM distortion testing for the very same reason you gave: that the distortion products typically fall in-band.
In fact, you may want to read my paper "A fully in-band multitone test for transient intermodulation distortion" that is on my web site at www.cordellaudio.com. That MIM test used a combination of three tones (9.0, 10.05, and 20 kHz) to fold both even and odd IM products down to around 1 kHz. The instrumentation was fairly inexpensive.
One thing you have watch out for with SMPTE IM is that it is fairly insensitive to HF nonlinearity and moreover it can yield very low numbers for amplifiers that have a lot of negative feedback at the in-band frequency of 7 kHz. Of course, one could argue that the magnitude of IM products in-band is what we hear, and that the NFB is just doing a very good job in that case. The same can be said for my MIM test, where the distortion products are around 1 kHz and many amplifiers have a lot of distortion-reducing negative feedback in play at 1 kHz.
Cheers,
Bob
I'm not that good withe math but I get the idea. Since it should be easy to test the theory I would have been more impressed if there had been traditional THD vs. Freq plots.
One for a good low THD power amp plus the Audiopax speaker.
vs
One for and Audiopax SE amp plus the Audiopax speaker.
The former would show pretty much the THD of the speaker itself. It the theory works the second should show a lower THD indicating that the additional THD of the SE amp "compensated" away some of the inherent THD of the speaker.
One of my amp design (solid state) has 0.2% THD (20kHz) at full power (100W/4Ohm). At 1kHz it produces 0.05% THD at the same power level.
At low power levels the THD is 0.02% (1KHz).
Are these THD figures too much?
(Anyway it sounds OK.)
At low power levels the THD is 0.02% (1KHz).
Are these THD figures too much?
(Anyway it sounds OK.)
Andy L. Francis said:
Hi Andy
According to international standards for consumer electronics, the THD of your amplifier it is accepted by far, so you can sale it in masive quantities. The subject it is how much good heared in your ears.
Tell me something: how much it is the rise time of your amplifier by applying a 10KHz square wave in its input with an amplitude so that the output not exceeds the 20Vpp?
Fotios
There is also a lot of material on Gerrit Boers website, with independent measurements:
http://gboers.xs4all.nl/daisy/home/g3/139/g1/loudspeakers/systemdistortion.html
and following pages.
Jan
http://gboers.xs4all.nl/daisy/home/g3/139/g1/loudspeakers/systemdistortion.html
and following pages.
Jan
fotios said:
Hi Fotios,
Thanks for the reply!
I measured the rise time at 20Vpp output using 10kHz square-wave signal. I got 15uS.
Andy L. Francis said:
Hey Andy,
Are you sure that you made rightly the measurement? Because 15ìs it is a very long rise time. Typical values don't exceed usually the 2,5ìs yet for an amplitude of 80Vpp in output. Try again the measurement without any load in output. I quote bellow a typical screensot from a DSO so you understand better what i say. Click on the link :
Fotios
- Status
- Not open for further replies.