Bandwidth question?

I am interested in the qualitative difference in sound betw amplifiers that have -3db roll off at 100khz vs -3db at 300khz. Thru the amps I have tried, I suspect increased bandwidth has more openness and transparency and hence a better sense of space sharing. At the same time, it is easier to screw up the sound due to noise (from components/AC/RF) or improper cartridge loading. I am not very certain of the correlation and interested in what you guys think?

In reviewing the measurement sections of stereophile, many amps with -3db at 100khz demonstrate subtle rounding of the edges when reproducing 10khz square waves. I don't listen to square wave so I don't know what that translate into.

I realize that some amps (Spectral or Soulutions) has very high bandwidth (MegaHz) to implement negative feedabck. I am not refering to that.
Let's start with audible frequency bandwidth = 20Hz...20KHz.
Than take any full range speaker that would be less-likely designed to cover the whole audible bandwidth.
Than take onto consideration that vast majority of people wouldn't hear anything bellow 28Hz or above 16kHz.
Than let's see where 100 or 300KHz???

A square waves are usually considered as samples in digital carying or encoding frequencies that are absolutely unaudible just like DC. A sound is a continuous signal in general. As far as "rounding of the edges..." looks like heresy to me so don't listen and don't translate unlistenable.

Many amplifiers can go way beyond audible bandwidth, but last would less-likely dictate performance difference.
Some people hear more air with vinyl than CDs. CDs usually cut off at 20kHz due to filters, for the noise we may hear. Vinyl is capable of going to over 30kHz, if the cartridge and system allows it. I seem to hear more air with vinyl too. If I was given a hearing test, I'd probably not hear above 16kHz at the extreme. Maybe it is harder to identify a continuous steady test tone, than a variety of them, warbled tone, or music. I do hear some metal dome tweeters that resonate at 22kHz. This causes extreme an irritable fatigue that leaves me no choice of either shutting it off, or leaving the room quickly. So, I think we are aware of sounds up to this 22kHz (at least in my case) sound that may be there. When it comes to amps, I don't seem to hear any more air with an amp that may be down 3 decibels at 20kHz, in comparison to one that goes to over 100kHz.I have several audiophile friend that don't hear the upper end air from an amp that has a high bandwidth like that. Two are sensitive to that 22kHz noise like myself. Ultrasonic alarms also irritate us.
I don't seem to hear any more air with an amp that may be down 3 decibels at 20kHz
Let me clarify. I mean may be down 3 decibels at over 20kHz. In other words down 3 db at 22kHz for example.
I don't think that it has anything at all to do with sounds past the audible range. My belief is that it has everything to do with phase respose within the audible range. And ultra high freqency roll offs have measurable effects on the time response within the audible range. If sounds are not arriving in the proper time sequence, it may sound good, but that doesn't mean that it's right.
Bandwidth is directly related to other factors, including feedback.
Isn't there also a phase shift component here, too?
Glai 12-06-11:
In reviewing the measurement sections of stereophile, many amps with -3db at 100khz demonstrate subtle rounding of the edges when reproducing 10khz square waves. I don't listen to square wave so I don't know what that translate into.
An ideal 10kHz square wave, having infinitely fast edges (rise times and fall times), consists of a summation of sine wave components at frequencies of 10kHz and all of its odd-numbered harmonics (out to infinity), each of them having different weightings that decrease as their frequency increases. The first odd numbered harmonic of 10kHz, aside from the 10kHz "fundamental frequency" itself, is 30kHz, which we essentially can't hear. I say "essentially" because some seemingly credible studies have reported that subjects are able to sense the presence of frequencies possibly as high as a few ten's of kHz above 20kHz, when those frequencies are present in combination with lower frequencies.

My basic point being that since risetimes and falltimes inversely correlate with bandwidth, our hearing mechanisms, not to mention the speakers, the source material, and the source component, will round off high frequency square waves to a considerably greater degree than most amplifiers will. With the possible (and likely very subtle) qualification suggested by the studies I mentioned, even if the 10kHz square wave were rounded off to a pure sine wave, we would not be able to hear the difference, assuming that the round-off effect is not accompanied by other side-effects.

On the other hand, a reasonable degree of bandwidth overkill in an amplifier will prevent that component's bandwidth limitation from contributing significantly to the overall bandwidth limitation of the system + our ears. As well as eliminating or minimizing other effects that could be manifested below 20kHz, some of which the others have noted above.

Again on the other hand, ultra-wide component bandwidth has potential issues of its own, as you correctly alluded to, including increased susceptibility to rfi, the possibility that some components in the chain may not be able to gracefully handle ultra high frequency components that are input to them by the preceding component(s), and the possibility that transmission line effects which occur at rf frequencies could cause signal reflections to enter feedback loops.

The bottom line, as I see it, is that in itself that spec does not have a great deal of usefulness. Although FWIW I would have to say that 100kHz strikes me as at or somewhat below the lower limit of my intuitive comfort zone, and all other things being equal I would give preference to a somewhat higher number.

-- Al
-3bd @ 100khz vs 300 khz can be critical - if you are a bat.
Al nailed it IMO. I can't think of any amp I have considered that had me worry about bandwidth.
I just did a Google Search, and found this article. It's in a Positive Feedback, and it's surprising about high frequencies. Link.[]
I agree that most of us can hear only up to perhaps 15kHz but higher system bandwidth guarantees low phase shift. For instance my Rowland 102 amp has -3dB bandwidth of 65kHz but at 15kHz has phase shift of about -20 degree. Even at 10kHz it is still around -15deg. This phase shift affects sound (summing of harmonics). I would feel more comfortable with at least 100kHz bandwidth but agree with Al that bandwidth too high has problems of its own. Obvious problem is HF noise pickup but less obvious is increased chance of Transient Intermodulation (TIM) since reduction of the bandwidth at the input reduces TIM.
I have a Sony ICF-2010 receiver which has a long wave band, from 150khz. and continuous coverage to 30mhz. this covers all the traditional SW bands and than some.
What I'm getting at is that if you have an amplifier with bandwidth to 100khz, you'll also need an antenna of suitable length. IOW, well over 6000 feet full wave for 150khz. much longer at lower frequencies. Submarines utilizing ULF for communications trail an antenna of some unknown but huge length. The transmitters antenna are huge arrays, visible from space.

I would suggest, however, that the bandwidth limit of 'd' amps is one of the subtle (or NOT?) problems people complain about.

Just by the numbers, I'd suggest a minimum of 2x upper human limit + a guard band....Maybe 44khz with rolloff above that of 3 or 6 db/octave. Now, I'm slippin' here, but where have I seen that number before? hmmmmmm.

Just thinkin' out loud.
That's true, but sensitivity of the amp does not end at -3dB point. Also antenna reception still exists at 1/10 of the wavelength. If we take, in your example, level drop of 20dB/decade it will be only -43dB at 10x10x150kHz=15MHz. Your full wave antenna is now 60ft offering reception up to 1/10 of the wavelength = 6ft.

Noise might be also capacitively coupled. At high frequencies any connector might become an input. Speaker output for instance is also an input of negative feedback. It has very low output impedance but only for low frequencies. Properly designed amplifier will have filters either RC and/or common mode chokes etc., but it only reduces noise pickup and not eliminates it completely. Some amplifiers (common to most of opamps) exhibit rectification phenomena where small amounts of very high frequency signal that is modulated (radio stations) converts to even smaller amount of audible signal because of uneven rise and fall times.

First obvious remedy is to avoid long cables if possible (where cable is still some antenna but skin effect does not provide shielding). Using shielded balanced ICs helps as well as power supply filter/conditioner, but it is better to avoid 500kHz amplifier because it is just asking for trouble. 44kHz seems a little low to me - I would settle for 100kHz. On the other hand newest Rowland 625 extends to 350kHz and Jeff Rowland is definitely a guy who knows.
I personally find that amps with a usable bandwidth out to 150 KHz hit a sweet spot for speed, clarity, and transparency without opening up the can of worms that ultra-wide bandwidth introduces nor the too-sweet, closed-in sound of amps that just make it out to 20KHz.

Amps linear to 150 KHz (and I'm simply relating my personal experience from preference--I've owned around 20 amps over the last 40 years) have a combination of body and warmth, plus transparency and clarity that I like. Although few of us hear *frequencies* beyond 20Khz, I suspect all of us can hear the difference that the wider bandwidth confers on square wave *rise time* in regard to clarity and transparency that accompanies a bandwidth of 150 KHz and beyond.

Amps with ultra-wide bandwidth have an unmistakably stunning clarity, but are sometimes accompanied by over-ring and treble brightness at the expense of warmth and body, both musical values. And sometimes an ultra-wide bandwidth design can run into an oscillation problem with the speaker interface. It's my understanding that this is why MIT designed a special network for their speaker cables to prevent oscillation in their collaborative audio show system with electronics from ultra-bandwidth Spectral and speakers by Hales.
For Your Information!

High frequencies above 20Khz matter!

So what is your concern? Do you feel like you are missing something in the high end? Air, space around instruments? Not always the amplifiers fault. A super tweeter might resolve issues.

There is a good article on wide bandwidth here:
So far I've not seen this mentioned yet, IMO this is one of the most important reasons to have some bandwidth:

Phase shift accompanies bandwidth limitations. In general, you can expect phase shift components to manifest to 1/10th the upper cutoff frequency, so if the amp cuts off (begins its rolloff) at 50KHz we will hear artifacts announcing that at only 5KHz.

This is true on the bottom end as well, phase artifacts will be heard at 10X the cutoff frequency, so a 20KHz rolloff will have artifacts up to 2KHz.

A premature HF rolloff will indeed manifest to our ears as a darkness or slowness in the amp at high frequencies. Oddly, if the rolloff is severe, it can have a brightness and a darkness at the same time! Phase shift can have a pronounced effect on the amp's ability to portray an accurate soundstage as well.

An LF rolloff is heard as a lack of impact. So yes, 2Hz is the minimum cutoff frequency if you want the bass to play right.

The fact that a speaker does not have this sort of bandwidth really does not seem to be relevant- you can hear these artifacts easily despite a lack of bandwidth in the speaker.

Excessive bandwidth on the high end can lead to RF problems. If the amplifier is capable of amplifying broadcast frequencies, you can expect to fry a lot of tweeters. You can also expect the amp to run hot, in the case of transistors. It is for this latter reason that many SS amps are bandwidth limited. It is for the former reason that we limit the bandwidth in our amps (our output section goes well into the megahertz region). Another reason to limit bandwidth is that if the amp is amplifying an out-of-band signal, it can gooble up power that might otherwise be used to drive the speaker with music.

Finally, if you have too much bandwidth you do run the risk of stability. This is particularly true if the amp runs feedback (ours get the bandwidth without feedback, BTW). Due to propagation delay times in the amp, negative feedback can become positive feedback if the frequency gets too high! For this reason feedback loops have to be handled with extreme care by the designer to prevent oscillation (heck, you *always* have to be careful with that anyway), as super HF bandwidth is tricky- you don't want the amp to have a reputation for blowing up, or blowing up speakers!
Thanks Atmasphere for taming famous American "Bigger-Better"!
Down 3db at 100khz or 300khz is insignifcant. Down 3db is good in that it is designed to reduce oscilation tendencies. At what power is the amp rated at, what is the power output at max or rated power vs. Frequency response and distortion. Measure all at 1wpc and full smoke. Key is the square wave at 10hz. If it deforms, notches or rounds off investigate. I have seen highly rated 100wpc monoblocks that roll off 10db at 20khz, and roll down15db at 50hz. And the square wave was rounding severely at 60hz. They were also voted "best of show".....likely not the product consumers get. I never liked the sound myself of the consumer product. A soft bass confirmed with a bench test, revealed a compromised product. Quality transformers are found with music reference, Quicksilver, EAR, conrad johnson, ARC, and others, but there is junk out there as well....beware jallen
Many thanks for all the responses. To be specific, I am not interested in debating the human hearing bandwidth. Mine is 25hz to 17khz.

I am interested in the how the ultrasonic bandwidth impacts the reproduction of audible frequencies. Many responses have pointed out phase issues, noises (or ultrasonic noise modifying how a component behaves in the audible freqs.)These are very helpful to me. I also realize that bandwidth for amp is dependent on load and output.


Why 10 times the audbile bandwidth is neccessary. A lot of amps are -3db at 100khz. This means faithful playback up to only 10khz, if you are correct.
"Why 10 times the audbile bandwidth is neccessary" - Because of phase shift. Rule of thumb says that phase shift at 1/10 of -3dB cutoff frequency is about 5 deg and that might be significant for proper summing of harmonics.
Glai, there you have it. Our ears are not so sensitive to phase shift at a single frequency, but when a band of frequencies is involved its another matter entirely!
I have never been a big fan on Mhz frequency response especially in an amplifier. Having a bandwidth that high can cause major problems including parasitic oscillation, rf interference etc. This can cause the top end to sound hard, grainy, and strident. This is especially true if the amp uses global feedback. To answer Kijanki, ten times the bandwidth is just a rule of thumb but it is something I like to see especially in a preamp. Personally I find -3dB down @ 200Khz to be sufficient. Also, when talking about amplifiers there are differences in specs. between power bandwidth and small signal bandwidth. By the way, most transformer coupled amplifiers are not capable of ultra wide bandwidth frequency responses.
No one brought "harmonics" into this debate?
I clearly understand the statement about our ears ability, but what about speakers? Where can you place them in discussed equation? Are they identical to our ears?
Orpheus10. That's all part of this. Sine waves don't really tell you much about the bandwidth of an amp or preamp since all your getting is the fundamental frequency without the harmonics. I tend to look at square waves. That tells me a lot about the amp or preamp I'm looking at. The reason is, a 1Khz square wave shows you not only the fundamental, but the harmonics as well. A perfect 1Khz square will tell me that this piece of gear is flat from 100Hz to 10Khz or ten times the fundamental frequency in both directions. It can also shed some light on the interesting things some manufacturers do with their output coupling transformers to make it seem like the transformer has more high frequency bandwidth then it really has. So for example for perfect 20kHZ square waves, an amp or preamp needs to have a bandwidth out to 200Khz. Sad thing is, it really won't tell you how the gear under test is going to sound. For that you need the best test instrument in the world, your ears.
Is the bandwidth thing the problem with 'd' amps? they drop like a rock somewhere around or just OVER 20khz in order to get rid of the switching noise. Early 'd' amps were also noted for RF effects. I once picked one up on a filling.
Going along with the steep frequency dropoff, which is also linked to load impedance with low impedance having a quicker dropoff, but you also apparently get some phase problems.

I'm saving my Pesos for a Pass integrated. (look for a website soon since i'll be accepting donations....'Pesos for Pass. org'. I'll buy it without even an audition. Just let me plug it in and give 'er a rip. Reasonable bandwidth, low feedback and near-impeccible square wave response. This won't be buying specs, either since virtually nobody who hooks 'em up to an appropriate speaker from SDFR to Panels has anything BAD to say about 'em.
Magfan, my class D Rowland 102 has 5Hz-65kHz -3dB bandwidth but I'm saving my Pesos for Rowland 625 with small possible temporary setbacks in the future like divorce, bankruptcy and foreclosure.
Right you are...
The open loop for the ASP500 module is 80khz@8 ohms and 60khz24ohms.
All at some specified tolerance.
Phase displacement is also noted......

While you don't have to be a "rocket scientist" to know what you hear, you do have to be a scientist in order to explain what you hear.

"Harmonics" are what separates the "high end" from "midfi". For example, if the fundamental frequency is 25 HZ, the frequencies of the harmonics are 50 HZ, 75 HZ and 100 HZ. Although the amplitude of the harmonics decreases in order, they determine the quality of the sound.

When you go to a live event, these harmonics are quite audible; they give the violin that rich tone and timbre, they resonate with each guitar note. These harmonics are present as low level detail at home.

If you have been to a live event and had that sensation of a sound that was ringing right in your ear, that was due to harmonics.

We're all talking harmonics here. I mentioned summing of harmonics twice while Al mentioned it also few times. Highest fundamental is around 8kHz (church organs) but most often it is only up to about 2kHz as it is shown in this interactive chart:

I noticed that Harp has absolutely incredible range - never expected that.
Marakanetz, as I mentioned earlier, it does not seem to matter that the speakers lack the bandwidth of the amps. The phase shift that will occur in the amplifier due to bandwidth limitations is something that you hear well within the range of human hearing and nearly any speaker can reproduce it.
Don c55, nice link. I can't find the articles anymore, but it has been proven in the last twenty years or so that the human brain does definitely detect frequencies above what the ear can hear, scientists just don't understand exactly how yet.
Here is a YouTube video with a guy testing a couple of old tube amps. They're still going FWIW.[] []
I watched the video and here is my interpretation of the performances. What he calls slant is actually called tilt or phase shift and it's measured in degrees. The Mac and the Dynaco look similar at maybe 20-25 degrees of tilt. This indicates that the low frequency response is rolled off. Possibly -3dB at 10-15 Hz. The Dynaco does a better job as the tilt is linear at low frequency. Therefore it's rolling off smoothly. The Mac on the other hand has tilt that is not linear at low frequency. At very low frequency the Mac frequency response might actually rise indicating some instability possibly caused by a coupling cap or power supply issue. High frequency response is similar on both amps with a bit of overshoot and only one or 2 cycles of well damped ringing. Some manufacturers do this so the amp sounds like it has greater high frequency response then it actually has. The Fisher looks like it has some major issues as it's high frequency response is not linear. It probably dips then rises and dips again. Either the amp has coupling cap issues, feedback cap issues, and/or the output coupling transformer is not properly compensated for. When the frequency response or any amp becomes non linear, i.e. rises and falls dramatically the test is pretty much over since that non linear response in not useable. I normally use sine waves to determine -3dB down points of an amp or preamp under test and confirm it with square waves.