john curl said:
Okay, I have a question.
Analogue material transferred to CD.
The analogue material have information way above 20 KHz, (material harmonics and may be distortion), information carried by the SHAPE of the signal.
What happens to that information (of above 20 KHz) when the CD player cuts off sharply anything above 20K?
Hi Scott,
I hide in the basement with all my test gear, wifey can't stand to be there.
Hi Joshua,
The CD players your are talking about are the first generation machines. They used 7th order, analog filters. Wicked sharp.
Most of today's machines actually oversample so the noise is up near the MW band (below). They can then use more gently filters to get rid of the nasties. They had already found out how audible a sharp filter can be. So, today's machine will have response beyond 22 KHz.
Some phono cartridges really didn't get up to 20 KHz. If they did, the phono pre may roll off, and any mis-tracking would tend to reduce HF information. Most cartridges need the proper loading to resonate with the coil to obtain the high end response, that included the tone arm wiring, leads and preamp input circuitry. Few systems have been measured and padded properly.
This is similar to a tape head. Many tape machines you did work with would easily get past 20 KHz if their heads were in good shape. Certainly, Revox or Studer products will. So will Teac / Tascam, Nagra, MCI and others.
-Chris
What's a "man cave"?
I hide in the basement with all my test gear, wifey can't stand to be there.
Hi Joshua,
The CD players your are talking about are the first generation machines. They used 7th order, analog filters. Wicked sharp.
Most of today's machines actually oversample so the noise is up near the MW band (below). They can then use more gently filters to get rid of the nasties. They had already found out how audible a sharp filter can be. So, today's machine will have response beyond 22 KHz.
Some phono cartridges really didn't get up to 20 KHz. If they did, the phono pre may roll off, and any mis-tracking would tend to reduce HF information. Most cartridges need the proper loading to resonate with the coil to obtain the high end response, that included the tone arm wiring, leads and preamp input circuitry. Few systems have been measured and padded properly.
This is similar to a tape head. Many tape machines you did work with would easily get past 20 KHz if their heads were in good shape. Certainly, Revox or Studer products will. So will Teac / Tascam, Nagra, MCI and others.
-Chris
Hi John,
I have been checking combinations over the years. I measure both the TT wiring and phono preamp input capacitance. That is really the only way to get the capacitive loading right. I know this is a rare procedure in the audio world.
From what I have seen, almost no one has their cartridge loaded properly.
Now, in the context where a good cartridge is installed and set up properly, you are quite right there John.
-Chris
That is only true of MM cartridges as long as they are loaded correctly. MC cartridges have fewer problems because they are normally only resistive loaded.
I have been checking combinations over the years. I measure both the TT wiring and phono preamp input capacitance. That is really the only way to get the capacitive loading right. I know this is a rare procedure in the audio world.
From what I have seen, almost no one has their cartridge loaded properly.
Now, in the context where a good cartridge is installed and set up properly, you are quite right there John.
-Chris
anatech said:
Chris, repeat after me:
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
The digital oversampling filter has a brickwall in it. It's just cheaper (and easier to make accurate) than an analog brickwall.
anatech said:
How do you determine what the correct loading of a cartridge is? A moving magnet cartridge has a "network" with the coil and the loading capacitance and resistance that approximately flattens the response (essentially a filter on its output) but moving coil cartridges are a different story. The inductance and resistance of the source is so low that loading doesn't really make a lot of electrical sense. Further, loading a 3 Ohm source enough to affect it would through away a lot of hard won signal to noise ratio.
I have never bought into the argument of mechanical damping from the loading- the numbers don't work. The "motor" is so loosely coupled that a short won't change its compliance.
Test records may be a way but test record collectors will vouch for finding a test record that will get you what ever response you want. There are plenty and they are all different. (I am working on a new test record for what its worth.)
Charles Hansen said:
I was saying the same thing to John on the phone today. Digital or analog the end result is identical- except that its easier and cheaper to make a complex filter in the digital domain. But you can't make the numbers do something different in the digital domain than what they would do in the analog domain.
While I really respect attention to detail, I think we are making this phono cartridge frequency response question a little too confusing. Most phono cartridges are not nearly as sensitive to loading as they once were. This is because the frequency response has been extended and the series coil has been reduced in value, significantly.
I have found test records to vary somewhat, but not like loudspeakers, or room characteristics.
I have found test records to vary somewhat, but not like loudspeakers, or room characteristics.
Charles Hansen said:
Hi Charles,
You are absolutely correct. However, some lunches are more expensive than others.
Any filter that approximates a brickwall with a very sharp cutoff will ring and smear in the time domain. This is a direct result of the inevitable link between the time domain and the frequency domain; something sharp in the frequency domain will be spread out in the time domain; something sharp in the time domain will be spread out in the frequency domain.
However, there is an important difference between sharp continuous-time analog filters and sharp finite-impulse-response digital filters that goes beyond the differences in difficulty of accurate implementation.
A sharp analog filter at 20 kHz will have an extremely non-linear phase response and variable group delay. An over-sampled digital reconstruction filter with equal sharpness of cutoff at 20 kHz will usually have a linear phase response. Some believe that this can make a very big difference.
I believe it is also the case that the sharpness of the anti-aliasing filter at the recording end is far more important than the sharpness of the recondtruction filter in the CD player.
You have a very well-reviewed CD player at Ayre which I have had the pleasure of using and listening to (John Atkinson loaned us his when we did our audio workshops at the HE2007 show). What approach have you taken with the reconstruction filters in that player?
Cheers,
Bob
Hi Bob,
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
"There's no such thing as a free lunch."
It's like Demian said, "But you can't make the numbers do something different in the digital domain than what they would do in the analog domain."
In theory, at least, you could put enough analog all-pass filters in the analog brickwall filter to also make it linear phase. I wouldn't want to actually have to do it, but if you did, it would give *exactly* the same response as the digital filter.
We used to use linear phase digital filters just like everyone else. However, we normally used a slow roll-off type as pioneered by Wadia in the late '80s. Last fall we spent four months listening to different digital filters. One thing we found is that minimum phase filters sound better -- giving our digital filters the same phase response as the comparable analog filter (without the all-pass filters added).
However, unlike everyone else who ever even played around with a minimum phase digital filter, we are *not* using IIR's. Instead we are using FIR's. You can read more at:
http://www.ayre.com/pdf/Ayre_MP_White_Paper.pdf
Bob Cordell said:
There are other, major problems with analog filters of high order as well. The Butterworth filter makes a great example. If one takes a Butterworth LPF of order N and -3 dB frequency of 1 rad/sec, it can be shown mathematically that the poles of its transfer function lie on the unit circle in the s-plane with equal angular spacing between them of pi/N radians. Let's say we wanted to make a 100th order Butterworth analog filter for a CD player to compete with the available digital filters. For this filter, the angular spacing between poles on the circle in which they lie will be 1.8 degrees (=pi/100 radians). This means that many of the poles will end up being just a few degrees from the j-axis.
One common, but very poor approach to designing such a filter is to pick out complex conjugate pairs of these poles and implement each pair as a second-order Sallen-Key filter. Using this approach on the 100th order case, you'd end up with 50 such cascaded filters. But as mentioned, many of these pole pairs will be very close to the j-axis. If you try to implement a Sallen-Key filter for such a pole pair, it's likely it would be physically unrealizable because of the capacitor ratio, or it could oscillate as well. Even if it worked, its peaking in the frequency domain would be ridiculous, and its time-domain performance unbelievably bad. The only hope for implementing the overall filter would be with a state-variable design where you're working with the entire polynomial of the transfer function denominator and not having to factor it.
Even if you could implement this 100th order filter successfully, its time-domain performance would be totally unacceptable. The time-domain response of ideal Butterworth LPFs gets worse and worse as the order increases, both in the case of percent overshoot and in the duration of the ringing. The 100th order case would be a total joke.
- Status
- Not open for further replies.