What is the value of choosing an output transformer with a very high frequency response? The importance of wide bandwidth is easier to understand on the low end because most output transformers have a low frequency response that is still reasonably within the audible range. But what is the impact of going high into the ultrasonic range?
Suppose a manufacturer offers a range of similar transformers that differ only in bandwidth: (10Hz - 35kHz) or (10Hz - 60kHz) or (10Hz - 100kHz). What are the compelling reasons for choosing the 100kHz OT?
Suppose a manufacturer offers a range of similar transformers that differ only in bandwidth: (10Hz - 35kHz) or (10Hz - 60kHz) or (10Hz - 100kHz). What are the compelling reasons for choosing the 100kHz OT?
The output transformer is usually the limiting factor for stability in a tube amplifier.
If the plan is to use lots of negative feedback to get low THD (like the Williamson amps and relatives), then you need high bandwidth in the OT to maintain stability with all that feedback.
But, these days, if you want low THD, you are far better off with a solid-state amp. Nowadays valve amps seem to be prized for their high distortion and lots of nonlinearity. Under these circumstances, I can't think of any reason for OT bandwidth above 20 kHz.
Keep in mind the median age of an active diyAudio member is somewhere in their mid-fifties. At that age, a majority of people can't hear much above 12 kHz at practical SPL levels.
20kHz is only audible by young ears at intolerable SPL (the threshold of permanent hearing damage), and so is a very over-generous spec for audio bandwidth at the best of times.
-Gnobuddy
Also the value of H. It easy to find a good declaration on response but not on inductance; generally a trafo with a great BW hasn't a high value of H. There are a little number of trafo with both value are scarce. One of the main test on trafo is the THD versus frequencies, normaly you can see like a U where the distortion increase at low and high frequency due the low self inductance at low freq. and parasitic at high. This means that the quality of the iron and the architecture of the coils are fundamental for BW.
Walter
Walter
Last edited:
Bandwidth is typically specified at the -3dB corners.... Also called POLES in a transfer function... Keep in mind that there is 45 degrees of Phase shift at these Poles.....
So Phase shift is really bad for music and is very noticeable in the high frequencies more so..... This is why you need to stretch out the Corner Frequencies so far, So that your usable audio range has good Phase Linearity.... Now when you close the Loop for the Negative Feedback..you have better Phase Margin, thus better stability..
So Phase shift is really bad for music and is very noticeable in the high frequencies more so..... This is why you need to stretch out the Corner Frequencies so far, So that your usable audio range has good Phase Linearity.... Now when you close the Loop for the Negative Feedback..you have better Phase Margin, thus better stability..
Wrt feedback requiring a wider bandwidth, for both low and high end, it typically gets a bit more complicated than knowing the OPT -3dB frequencies.
The transformer response is worth measuring to beyond the required amp's operating bandwidth (not just the transformer bandwidth), due to bandwidth extension from feedback, and as margin is required at and beyond the operating bandwidth.
Also, feedback is sometimes not applied uniformly across the operating bandwidth of the amp, but rather is lowered at the shoulders using (typically) shelf networks.
The transformer response is worth measuring to beyond the required amp's operating bandwidth (not just the transformer bandwidth), due to bandwidth extension from feedback, and as margin is required at and beyond the operating bandwidth.
Also, feedback is sometimes not applied uniformly across the operating bandwidth of the amp, but rather is lowered at the shoulders using (typically) shelf networks.
Terman in the text Electronic & Radio Engineering, 4th Edition claims the effect of phase shift not detectable by ear. But obviously required for stereo reproduction. Most music power is in the mid-range, not at the band edges. Unless we are listening to a large church organ. In a large space where the sound can be reproduced.
Attachments
The O/P transformers primary resonance frequency and its impedance has to be such as to not effect the stability of an amplifier using negative feedback over a range that far exceeds the audio frequency bandwidth. However when running valves in Class AB, the interaction between the primary halves has to be considered when a valve approaches or actually is cut off by a signal. A condition once known as notch distortion occurs when there is a significant flux leakage between the transformer primaries.
Norman Crowhurst discussed this problem and J. Somerset Murray described something similar in a series of articles in 'Hi Fi News' many years ago. Those articles , by the latter author, revealed the cause, the effects and designed a transformer to overcome the problem of what he labelled as Leakage Reactance Distortion. A way of measuring a transformer for this defect was included. I do not have these articles available as I read them via the company library that I worked for.
I can only recall the basics and a manufacturer, Transformer Equipment that supplied transformers made to Somerset Murray's design. Partridge Transformers supplied ultra linear transformers, types P4200, P4160 that specified the primary resonance frequency as well as the primary to secondary resonance freq., so it would appear that this is something to be considered in a transformer's specification.
Norman Crowhurst discussed this problem and J. Somerset Murray described something similar in a series of articles in 'Hi Fi News' many years ago. Those articles , by the latter author, revealed the cause, the effects and designed a transformer to overcome the problem of what he labelled as Leakage Reactance Distortion. A way of measuring a transformer for this defect was included. I do not have these articles available as I read them via the company library that I worked for.
I can only recall the basics and a manufacturer, Transformer Equipment that supplied transformers made to Somerset Murray's design. Partridge Transformers supplied ultra linear transformers, types P4200, P4160 that specified the primary resonance frequency as well as the primary to secondary resonance freq., so it would appear that this is something to be considered in a transformer's specification.
Last edited:
Let's see what the data shows:
That excerpt was taken from here: Audibility of Phase Distortion
The facts seem to be that:
(a) Using a computer, it is possible to artificially butcher the phase of a signal so much that you can easily hear a change in sound. The 3 piano notes in the link illustrate this.
(b) Without such extremely artificial manipulation, phase shifts might perhaps just barely be detectable under artificial conditions such as listening in an anechoic chamber, but there is no definitive evidence of this.
(c)Much more importantly, under normal listening conditions, normal phase shifts created by audio electronics cannot be heard - at all.
Incidentally, if you had a perfect zero-phase loudspeaker located two metres from your ears in an anechoic chamber (no reflected sound), signals at 15 kHz would still experience roughly thirty-one thousand, one hundred seventy two degrees of phase relative to signals at 150 Hz. Simply travelling through the air caused the treble to phase-shift by over thirty one thousand degrees - and we can't even hear that enormous amount of phase shift. No wonder we can't hear the far, far smaller phase shifts in our amplifiers.
In a normal room, sound would reflect multiple times from objects at multiple distances, and there would be additional phase shifts that might be ten times bigger than 31000 degrees (reflections from a wall 10 metres away would do that, for example, or multiple reflections from two walls 5 metres away.)
For this calculation, I've assumed the speed of sound in air in the room is 343 metres/second. It will vary fractionally with temperature and barometric pressure, but not enough to make any significant change to the facts - you get tens of thousands of degrees of additional phase shift between bass and treble signals, just because of sound travelling from the speakers to your ears.
Keep in mind, any amplifier with negative feedback has less than 180 degrees of phase shift between bass and treble frequencies. That's a fraction of a percent of the phase shift caused by two metres of sound propagation through the air. And we can't even hear the much bigger phase shift.
Anyone who thinks they can hear that infinitesimally tiny extra bit of phase clearly has superhuman senses, and should also be able to detect the additional weight of the flea on their cat when they pick up the cat.
Worrying about phase shift in amplifiers is a case of "He can't see the forest for the trees." In other words, it's a case of losing sight of what actually matters, and focusing instead on something that's utterly irrelevant.
-Gnobuddy
Since the issue of unequal phase shift of different frequencies at a fixed distance was mentioned, I raise another thing to consider:
Phase Shift is one thing. Group Delay is another . . .
Use left and right amps, and left and right two way speaker systems. Put one speaker 10 feet in front of the other speaker. Connect only the tweeter of the front speaker. Connect only the woofer of the back speaker.
If you have an anechoic chamber, this will be the definitive test. But you might be able to hear the same effect in an acoustically correct auditorium. If you do not have either, an outside open field will do very well.
Use a test tone that is at the speaker crossover frequency, and apply it one at a time to the one speaker, and then to the other speaker. adjust the sound amplitude separately of the front and rear speakers so that they appear to be the same amplitude at the listener.
There will be a 10 msec delay of sound from the woofer versus the tweeter at the listener.
Now, play only the left channel music through both left and right amps and left and right speakers (only one tweeter and only one woofer active). Use transient music (especially percussive) drums, symbols, bells, vibraphone, clapping hands, shouts, etc. See if the sound is fractured, versus when the speakers are at the same distance from the listener (and with the apparent amplitudes re-adjusted for equal apparent amplitudes of crossover frequency test tones at the listener).
And since testing this way is so difficult, get that computer out, create some filters, and delay all the frequencies below 1 kHz by 10 msec, versus all the frequencies above 1kHz. Then play music through that, first with the delay filters, and then without the delay filters.
Now I know most speakers do not have a 10 msec group delay at mid frequencies, and even less amplifiers have that much group delay, but now at least you can see and hear the results of group delay.
Group delay is always about relative arrival time. Phase is not always about arrival time.
Phase Shift is one thing. Group Delay is another . . .
Use left and right amps, and left and right two way speaker systems. Put one speaker 10 feet in front of the other speaker. Connect only the tweeter of the front speaker. Connect only the woofer of the back speaker.
If you have an anechoic chamber, this will be the definitive test. But you might be able to hear the same effect in an acoustically correct auditorium. If you do not have either, an outside open field will do very well.
Use a test tone that is at the speaker crossover frequency, and apply it one at a time to the one speaker, and then to the other speaker. adjust the sound amplitude separately of the front and rear speakers so that they appear to be the same amplitude at the listener.
There will be a 10 msec delay of sound from the woofer versus the tweeter at the listener.
Now, play only the left channel music through both left and right amps and left and right speakers (only one tweeter and only one woofer active). Use transient music (especially percussive) drums, symbols, bells, vibraphone, clapping hands, shouts, etc. See if the sound is fractured, versus when the speakers are at the same distance from the listener (and with the apparent amplitudes re-adjusted for equal apparent amplitudes of crossover frequency test tones at the listener).
And since testing this way is so difficult, get that computer out, create some filters, and delay all the frequencies below 1 kHz by 10 msec, versus all the frequencies above 1kHz. Then play music through that, first with the delay filters, and then without the delay filters.
Now I know most speakers do not have a 10 msec group delay at mid frequencies, and even less amplifiers have that much group delay, but now at least you can see and hear the results of group delay.
Group delay is always about relative arrival time. Phase is not always about arrival time.
Last edited:
Most ignorant bunch of BS I have heard in a while...So according this ....no need to spend money on wide band audio OT's ...just by the cheap stuff with minimal bandwidth..because nobody will notice the difference...
Look at the Fourier Series with phase shifted highs and look at the resultant waveform... When you close the loop some amps with OT's of limited bandwidth... the transient response is total crap.. and any step response will have slight ringing that will inter-modulate the high frequency content...
These blind test don't have direct application to the issue at hand...
The LF behaviour depends on inductance, so for good LF bandwidth you need high H.waltube said:
You may be confusing two different things. Phase shift is almost not noticed on music (or anything else). At higher frequencies we cannot even distinguish pitch, let alone phase. However, well-behaved phase shift allows for lots of feedback to be applied while maintaining stability.cerrem said:
No. A pure time delay preserves relative phase.Gnobuddy said:
That isn't what he said. There are good reasons to buy good OPTs, but the phase of what you hear is not one of them.cerrem said:
It is very common for audiophiles to frighten themselves quite unnecessarily by looking at time domain waveforms.
If you mean SE amplifier without global feedback, it mostly doesn't matter.
With push pull and global feedback situation is completely different. 20 dB global negative feedback and output transformer with 50 - 60KHz resonant frequency are very likely to result in around 5% THD and inter-modulation distortions at max output power @ 20KHz (caused by phase shift). These are not empirical values, but rather result of experiments with improved Williamson and Mullard 520 variant (both 6550 / KT88 60W).
Please note I meant resonant frequency, which is a product of leakage inductance Ls and stray capacitance Cs, F=1/(2*PI()*SQRT(Ls*Cs)).
Higher top -3dB frequency usually means higher resonant point (unless design is awkward or broken).
Additionally, Ls > 12mH and global feedback > 20 dB often results in instability and spurious oscillation, which may be not visible at first sight and is not easy to fix.
Attachments
1) In travelling a distance "x", a wave experiences a phase shift of 2*pi*x/lambda. That's 2pi radians, or 360 degrees, for every wavelength travelled. This follows directly from the equation for a propagating wave, i.e. exp(wt - kx), where w=2 pi f and k = 2 pi / lambda.
2) Ergo, If a 150 Hz sound wave travels one wavelength, it experiences 360 degrees of phase shift.
3) One wavelength of a 150 Hz sound wave equals one hundred wavelengths of a 15 kHz sound wave (assuming the speed of sound changes negligibly over that frequency range.) This follows directly from lambda = v/f, where v is the speed of sound in air, and f the frequency.
4) Ergo, if a 15 kHz wave travels the same distance through air as a 150 Hz wave, it accumulates one hundred times as much phase shift as the 150 Hz wave.
If that distance was one wavelength at 150 Hz (the same as 100 wavelengths at 15 kHz), the 15 kHz wave accumulated 36000 degrees of phase, while the 150 Hz wave accumulated 360 degrees of phase.
5) Ergo, if a 150 Hz wave and a 15 kHz wave start out together in phase, they will experience increasing amounts of relative phase shift as they travel through air. After travelling a distance equal to one wavelength of the 150 Hz wave, the phase difference between the 15 kHz wave and the 150 Hz wave will be (36000 - 360) degrees, or 35640 degrees of phase.
Which step do we disagree on?
-Gnobuddy
So: carefully collected scientific data from carefully conducted experiments by one of the most respected individuals in the field is "ignorant BS", according to you.
What do you suggest we replace it with? Unsubstantiated opinion and emotion and belief? Because that route leads back to the Dark Ages of superstition and ignorance. We've been there, it was horrible, we don't want to go back.
A lifetime of brain-washing by advertisers telling you that "You get what you pay for" can make this hard to believe. It's important to remember that this claim is just brain-washing, not reality. (If you found a big chunk of gold by the river, and paid nothing for it, is it "bad gold", quite worthless? If you pay $50,000 for a rusted-out 1980 Chevy pickup, does that make it a better vehicle? Of course not. The price you pay for it doesn't change the item itself.)
Reality isn't advertising, and it isn't opinion, and it usually isn't common-sense, either. Reality is discovered, one bit at a time, by smart people using the scientific method, generating hypothesis, and conducting experiments to test if the hypothesis was true or not. You don't always find what you expect to find - reality is more complex than the limited human imagination, and our limited human prejudices.
In this case, reality is beautifully described in the Floyd Toole quotation I linked earlier: "Within very generous tolerances, humans are insensitive to phase shifts."
Yes, the waveform changes. Yes, we cannot hear that. This has been known for decades - we do not hear waveforms, we hear spectral content.
Agreed, and several people (including myself) pointed this out early in this thread. If the intention is to apply negative feedback, then a wider transformer bandwidth will allow more negative feedback, and/or better stability margins.
This is the only real reason one might need a transformer with bandwidth wider than 15 - 20 kHz.
And why is that? Science is not applicable, but unsupported opinion / emotion / belief is?
-Gnobuddy
The OP did not indicate whether the application was SE or PP, it is certainly possible even in SE applications to achieve an HF bandwidth of 100kHz or greater, but at considerable expense.
Lundahl makes affordable PP transformers with bandwidths of 100kHz or greater.
I'm of the opinion that unless the intention is to run a lot of feedback that very wide bandwidth does not confer audible benefits commensurate with the increased cost in many (most?) cases. I've found output transformers with 30kHz -1dB to be more than satisfactory, but I'm a deaf old coot. LOL
Lundahl makes affordable PP transformers with bandwidths of 100kHz or greater.
I'm of the opinion that unless the intention is to run a lot of feedback that very wide bandwidth does not confer audible benefits commensurate with the increased cost in many (most?) cases. I've found output transformers with 30kHz -1dB to be more than satisfactory, but I'm a deaf old coot. LOL
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.