Up until now I use as little feedback as possible in my headphone amplifier designs (made 2 until now. The last one a 2 stage amp).
Did made a lot effort to get the THD below 1% but my second design is suffering around 3% THD at an output level of 700mVRMS into 300 Ohm load.
What is your target THD for your designs?
Did made a lot effort to get the THD below 1% but my second design is suffering around 3% THD at an output level of 700mVRMS into 300 Ohm load.
What is your target THD for your designs?
Any common amp whitout NFB will have about 5% THD but in a way that is sympathetic to our ears. NFB not only reduces gain and THD, more important in this case is that it reduces the equivalent output resistance of the amplifier, making it a best amplifier under load. Also, small amounts of PCF (Positive current feedback) can enhance this aspect.
My target is always less than 1% THD.
While it's true that usually, datasheets give THD in the 5-10% range, it's what you get when you max it out. At reduced level, the THD goes down almost linearly; in other words, just design in a lot of headroom and you're cool: easy to achieve in your case of about 1V peak and 50 mA peak.
Of course, a lot depends on the topology also: output transformer or direct drive, etc.
While it's true that usually, datasheets give THD in the 5-10% range, it's what you get when you max it out. At reduced level, the THD goes down almost linearly; in other words, just design in a lot of headroom and you're cool: easy to achieve in your case of about 1V peak and 50 mA peak.
Of course, a lot depends on the topology also: output transformer or direct drive, etc.
Myself, I don't worry about it. (THD)
We don't actually hear total harmonic distortion, but rather the few harmonics near the fundamentals of the program source (e.g. "music"), in a nonlinear way. The upper harmonics (as found thru synthetic-distortion-by-infinite-impulse-response digital remastering) turn out to be nearly inaudible, except when amplifying bass band material.
My design goals are to use highly linear components; I don't extol the virtues of 'magical golden selection' super-expensive tubes, capacitors, resistors or even transformers and chokes. No garbage… of course, but no silver foil 24 karat gold lead stuff. I tend also to use very-local negative feedback to tame the V³⁄₂ valve response realities.
This is to say, NOT using cathode resistor bypass capacitors either on triode or pentode voltage amplification stages. Indeed … except for S₁ (which I prefer fairly high gain from), subsequent stages are tuned for lower gain. I'd rather have more stages of lower gain than fewer-of-higher. Yep, opinion. But the results speak for themselves…
When driving real acoustic loads (AKA speakers!), I do like to add about 10 dB of GNFB; this substantially increases the damping factor DF of the output, which drives more-modern, more-modestly-efficient multi-cone speakers more accurately. The result of course is a drop in output power over the no-GNFB design, but since I've typically got more stages of VAS between, … they can be tweaked 'up' a bit in gain to counter the GNFB.
Anyway.
Just reminiscing.
GoatGuy ✓
PS (edit) ... I also shoot for about 25W output for most power amplifiers for ordinary listening room use. Plenty of power to almost-shake-the-windows when needed, but when fully wide open on a high-dynamic-range source, offers exquisite reproduction at totally believable sound-pressure levels. Would 100W be better? Only rarely. So rarely that one can simply ... 'give in' and purchase a top-gun Class D amplifier to 'pull into service' when needed. Which is rare. Thankfully rare.
We don't actually hear total harmonic distortion, but rather the few harmonics near the fundamentals of the program source (e.g. "music"), in a nonlinear way. The upper harmonics (as found thru synthetic-distortion-by-infinite-impulse-response digital remastering) turn out to be nearly inaudible, except when amplifying bass band material.
My design goals are to use highly linear components; I don't extol the virtues of 'magical golden selection' super-expensive tubes, capacitors, resistors or even transformers and chokes. No garbage… of course, but no silver foil 24 karat gold lead stuff. I tend also to use very-local negative feedback to tame the V³⁄₂ valve response realities.
This is to say, NOT using cathode resistor bypass capacitors either on triode or pentode voltage amplification stages. Indeed … except for S₁ (which I prefer fairly high gain from), subsequent stages are tuned for lower gain. I'd rather have more stages of lower gain than fewer-of-higher. Yep, opinion. But the results speak for themselves…
When driving real acoustic loads (AKA speakers!), I do like to add about 10 dB of GNFB; this substantially increases the damping factor DF of the output, which drives more-modern, more-modestly-efficient multi-cone speakers more accurately. The result of course is a drop in output power over the no-GNFB design, but since I've typically got more stages of VAS between, … they can be tweaked 'up' a bit in gain to counter the GNFB.
Anyway.
Just reminiscing.
GoatGuy ✓
PS (edit) ... I also shoot for about 25W output for most power amplifiers for ordinary listening room use. Plenty of power to almost-shake-the-windows when needed, but when fully wide open on a high-dynamic-range source, offers exquisite reproduction at totally believable sound-pressure levels. Would 100W be better? Only rarely. So rarely that one can simply ... 'give in' and purchase a top-gun Class D amplifier to 'pull into service' when needed. Which is rare. Thankfully rare.
Last edited:
Why, why, why… that's totally respectable! What's an audio-nut to think?
I achieve similar results with 10 dB of GNFB, and get a much-more-solid damping factor to boot. Your results may vary.
One of the really big apostasy-verging-on-anathema methods I sometimes use is to deploy a high quality OpAmp to do the input-and-GNFB summing. They're just so good at that particular job. Also, it makes it quite easy to have a 4-way switch for feedback. None (but with switched in gain-compensation-reduction!), 'standard fixed' GNFB, OpAmp summed GNFB and variable GNFB with the troublesome potentiometers being part of the mix. We like rotary switches, yes we do. All those extra poles … allows for all sorts of nearly-magical circuit reconfiguration. Invest those design-seeds once, and they're like perennial flowers: keep coming back again, and again, and again.
Just Saying,
GoatGuy ✓
Lower THD means you're operating in a more linear range, which gives you hope you get less IMD and other nasty stuff. Within the context of the original Q, a headphone amp, it's easy enough to achieve; the line amp I'm currently working on should have less than .1% with just a bit of cathode degeneration. And if you have more than one stage, you may be able to achieve some distortion cancellation by messing with the operating points.
Guitar/pedals/electricbass/drum kit rock, no telling what the mastering guy heard. Most CD's are compressed too so it is all the same volume.
Strumma strumma acoustic guitar & vocal, you have to know the singer to detect differences from original. Top frequencies around 7000 hz.
Really difficult sources like top and bottom octave grand piano from a classical CD house, tinkly bells, other percussion with very high frequencies, 1% HD really shows up with good speakers. IM distortion sounds like the letter S with no pop filter. Which is why I test amps with tracks including the previous. The easy stuff doesn't count. I don't need a spectrum analyzer to detect bad reproduction. Warning my ears still respond to 14 khz. Most US males can't hear even 7000 hz due to fireworks, motors, power tools, too much R&R, other manly arts.
My ST70 with capacitors replaced to OEM specs and new tubes, 1% HD, is fuzzy on tracks containing instruments in paragraph 3. Upgrade of drivers from 7199 to 6SN7 purchased but I hate to tear up a classic. Looking at buying some clone transformers for another chassis.
.02% HD CS800s, much better sound on difficult sources.
I'm pleased to finally have some speakers that will actually sound pretty much like a Steinway grand. Note real wood piano grand has huge power peaks when the hammer hits, will show from ppp to fff (volume indications) whether your amp & speakers have the s/n, power and low distortion necessary. One reason I play the Sohmer piano so much myself, it still sounds better than my sound system. You think the peaks are stressful on a Steinway, listen to a Bosendorfer grand sometimes. Those have strings so tight & light they have to be replaced periodically. Most mastering techs will remove all such nasty dynamic range reality from a CD or LP.
Last edited:
I really try to keep it below 0.5%, l really like to see .25 or less. So far the worst I've had is about .45% throughout the spectrum (at any power level) in my hybrid 1625 amp, but that said I am beginning to wonder if some poorly matched power tubes is making that worse. Something to look into next time that amp is on the bench.
My goals are different from a lot of people, however. I like experimenting and seeing what I can achieve. For my discrete solid-state preamp, I eventually settled with about .008% THD+N throughout the spectrum, much of which is probably coming from the Jensen transformers on the ins and outs. Some of that is noise as well (poor shielding on the transformers).
On my tube preamp, however, I was far less stringent. It uses 12AU7s, no global feedback and a cathode follower to buffer the output. I get about .05% THD throughout the spectrum, and I consider that to be respectable. Given the less-than-ideal linearity of the 12AU7, I'm happy with that.
PRR's comment on the nature of the distortion is key. Some amps can have "Good" distortion figures and sound terrible.
I think a lot of the reason for the popularity of SE amplifiers is that the gapped transformer with DC flowing through it adds some considerable amount of distortion, particularly at the low end when the core nears saturation, and this distortion seems to be of the type that humans find to be pleasing. This is also probably the basis for a big part of the "Neve Sound" for those who know what I'm talking about.
My goals are different from a lot of people, however. I like experimenting and seeing what I can achieve. For my discrete solid-state preamp, I eventually settled with about .008% THD+N throughout the spectrum, much of which is probably coming from the Jensen transformers on the ins and outs. Some of that is noise as well (poor shielding on the transformers).
On my tube preamp, however, I was far less stringent. It uses 12AU7s, no global feedback and a cathode follower to buffer the output. I get about .05% THD throughout the spectrum, and I consider that to be respectable. Given the less-than-ideal linearity of the 12AU7, I'm happy with that.
PRR's comment on the nature of the distortion is key. Some amps can have "Good" distortion figures and sound terrible.
I think a lot of the reason for the popularity of SE amplifiers is that the gapped transformer with DC flowing through it adds some considerable amount of distortion, particularly at the low end when the core nears saturation, and this distortion seems to be of the type that humans find to be pleasing. This is also probably the basis for a big part of the "Neve Sound" for those who know what I'm talking about.
Adding a bit more to it. Perhaps measurement of the acoustic output would be beneficial.
Last edited:
Thank you all very much for your valuable responses!
In this respect I stumbled across this paper published in the September 1952 issue of Wireless World. I think I do not have to face copyrights anymore due to the long time ago it was published, but Moderator feel free to remove it if I am wrong. Please let me know this then.
In this respect I stumbled across this paper published in the September 1952 issue of Wireless World. I think I do not have to face copyrights anymore due to the long time ago it was published, but Moderator feel free to remove it if I am wrong. Please let me know this then.
An interesting topic of research is seeing if amp / speaker non-linearity cancellation can be shown to work with IMD as well as HD. My suspicion is it won't work when a speaker crosses through its resonant frequency as the phase change spoils cancellation.
It seems to me that trying to cancel speaker distortion by careful design of the speaker is a better approach.
It seems to me that trying to cancel speaker distortion by careful design of the speaker is a better approach.
Isn't it like using the MFB originally invented by Philips in the 70s of the 20th century?
I can do acoustic THD measurement at speaker and tune distortion of my amp best I can anytime, some cancellation still means less distortion, better than no cancellation. I let others do the "careful design of the speaker" approach.
MFB ≡ Motion Feed Back, Philips Co, 1970s.
Essentially a speaker with a built-in amplifier, and sensor on the woofer-cone; output from sensor phase adjusted and put into feedback network of the low-pass amplifier driving woofer. Resulted in much better bass response from surprisingly small speakers, at reasonable sound-pressure levels.
GoatGuy ✓
Essentially a speaker with a built-in amplifier, and sensor on the woofer-cone; output from sensor phase adjusted and put into feedback network of the low-pass amplifier driving woofer. Resulted in much better bass response from surprisingly small speakers, at reasonable sound-pressure levels.
GoatGuy ✓
I suspect the better speakers you use the more distortion will bother you, that and more complicated music. With 110db/w speakers the performance at < 1w is what really counts as on the peaks it's not just the system that'll be distorting but your ears! Noise is more important to me than getting the thd down to 3 decimal places. My PSE 6e5p manages 0.06% at 1w and if memory serves 2%at 4w so that'll do me.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.