|'); } // End -->|
In Reply to: It was probably the AV8 I tried a couple of years ago... posted by oscar on August 8, 2005 at 16:43:19:
yeah you're right - i auditioned the lexicon mc-8 and wasn't impressed - it sounded no better as a preamp than a harman kardon receiver (but the harman kardon was like 1/5 of the price)
you should try the cary cinema 6 out - don't be fooled by the relatively low price, like you i've been hunting around for a long time and this is the first unit i liked that i could actually afford. unfortunately, no balanced ins or out but then my power amp doesn't do balanced either.
I'd much prefer to take advantage of balanced outputs, though this is gonna cost. Not too many 5/6 channel balanced pre/pros out there... Sunfire ? Halcro, Theta (??), Meridian ?$$$),
out of those lot, meridian or theta is likely your best bet. not so sure about halcro or sunfire. there's also integra research rdc7.1 and halo c1.
i've decided that the prepro market is too thin to really support up to date models. you may not have noticed this, but even the cheapest receivers tend to have better DSPs and support for more formats than very expensive pre pros. because these pre pros don't sell in large volumes, they tend to be several generations behind in technology, and that's inexcusable.
The best sound for the money is to be found in pre-pros from Sunfire, Anthem and Halo, not in multichannel preamps and certainly not in receivers. The installed market in North America is very heavy to pre-pros and they not only have current features, but frequent firmware updates.
... and i meant it. i haven't heard them, and unlike some people, i don't judge an equipment until i've heard it.
and my comment on pre-pros relate to the DSPs, which are several generations behind what's available in even the cheapest receivers. that's a statement of fact, not opinion. there are NO pre pros out there based on the ti da610 chipset or the cs49500, but there are several AV receivers.
i have heard the lexicon mc-8, and compared it with a harman kardon dpr2005, and i don't think the difference in quality is significant enough given the price difference. this is an opinion though, not necessarily shared by everyone.
... just in case anyone is wondering why i care so much about the DSP, they do make a difference in the quality of the digital processing.
as a rule of thumb, the more horsepower is available for DSP, the better the overall quality of the digital section. more horsepower translates to more precision used for DD/DTS decoding, which leads to smaller quantization errors caused by truncation/roundoff (this is not just a theoretical benefit, i can hear differences between DD done in 24 bit fixed point vs 32-bit floating point). 64-bit floating point is the ideal and will result in nearly "transparent" digital processing with no degradation of sound quality.
more horsepower also leads to more taps used for digital filtering, and the ability to run more digital filters, which improve the quality of bass management and room eq.
Basically, at this point in time, given that the latest gen DSPs are starting to offer 64-bit floating point, it would be silly to spend a lot of money on pre-pros based on out of date 24-bit DSPs.
Halo, Sunfire, Halcro, Meridian are using the Freescale platform, which is fundamentally 24-bit. Meridian, of course, has done their own proprietary code development and run their DSPs in 48-bit mode, but i'm not so sure about the others.
Lexicon is still using a 24-bit processor for DD/DTS decoding, but at least is using 32-bit floating point for bass management and Logic 7.
Compare that to the Denon AVR5805 and Harman Kardon AVR635, both of which are using the latest generation TI DA610.
What really lets the receivers down is the generally poorer quality of the analog preamp section, plus higher noise levels. But if you look at the specs of supposedly high end pre pros, they don't measure any better in terms of noise levels, and that concerns me.
What would be a better approach is to buy a relatively cheap receiver with good noise level specs, rip out the power amps, upgrade the power supply, and mod the analog pre amp section to improve the audio quality. I suspect that would give a better result than a typical "high end" pre pro.
which will yield an audible improvement (potentially huge), and stop worrying about the number of bits in the DSP used to convert low resolution sources.
you will understand why the number of bits used to decode Dolby Digital (which is what the majority of DVDs use) will yield a significant improvement in the decoding quality. Also, why floating point processing will be better than fixed point.
and can you explain exactly why buying a wall shelf for the turntable will improve the playback of DVDs?
In general, I really don't think the quality of DVD sources requires more than 24 bit floating point architecture. By the way, I listened to the AV8 when I bought my "new" system, and it was by far the weakest of the pre-pros I listened to: the 5.1 analog sounded particularly bad. The Lexicon MC-8 didn't impress me much either. There are (or were) some much more impressive pre-pros you haven't mentioned: Bel Canto, EAD and Sim Audio (modified EAD) in particular.
*** I really don't think the quality of DVD sources requires more than 24 bit floating point architecture ***
There is no such thing as "24 bit floating point architecture" - that would imply a 15 bit mantissa and 8 bit exponent plus a sign bit, which is clearly not enough to process 24-bit words.
The minimum required to process 24-bits resolution in floating point is a 32-bit representation.
24-bit fixed point is susceptible to round-off/truncation errors - and requires very careful scaling of intermediate results and a 48-bit accumulator. The Meridian approach of doing it entirely in 48-bits is a smart one, but requires twice the processing power.
But if you examine the link that i provided, even a casual numerical analysis would show that the algorithm is very difficult to implement in 24-bit fixed point. A bad implementation could easily reduce the accuracy of the final result to around 8-12 bits. Even a 32-bit floating point design is susceptible to quantization if care is not taken. A 64-bit implementation is ideal because it allows full 24-bit resolution to be carried throughout the entire processing chain with absolutely no danger of truncation even in pathological situations.
*** There are (or were) some much more impressive pre-pros you haven't mentioned: Bel Canto, EAD and Sim Audio (modified EAD) in particular. ***
I haven't heard them, which is why i haven't mentioned them. EAD and Sim Audio are still based on the 24-bit Freescale DSP though, don't know about Bel Canto. In general, I would stay away from any Freescale design apart from the Meridian. In addition to the 24-bit issue, Freescale doesn't provide certified DSP decoding libraries, unlike the other manufacturers, so you either have to rely on an in-house development team (very expensive R&D - hence the high cost of Meridian), or outsource to an outfit like Vinci Labs. I have nothing against Vinci Labs - they do good work, but it reduces the differentiation across models, and I would question their ability to sustain innovation given their limited R&D budget compared to someone like TI. Look at their ability to offer room eq, for example, it's still "coming" but TI and Cirrus Logic have working implementations. And someone like Audyssey isn't likely to write code for a 24-bit platform - it will be nearly impossible to optimize something as complex as that in fixed point.
I haven't opened up my Cary yet, but I suspect (based purely on the specs) that it is using the CS49400 DSP - not a bad chipset, and comes with a good DSP library, but not the latest and greatest. And there are some annoying quirks in the Cary implementation - no DPL2x layering on top of DTS for example.
PS - You may want to do a search on the Vinyl Asylum. A wall shelf doesn't necessarily improve vinyl performance, it depends on your house and your turntable. The general idea is that in a house with a wooden floor (or if you system is on an upper floor), the walls offer better rigidity and avoid floor vibrations. I nearly bought a wall shelf a few years ago, but based on advice from knowledgeable vinyl inmates, I'm glad I didn't since it's debatable whether it would have yielded an improvement in my situation (my house is built on a concrete slab).
it took a lot of effort for Meridian to program them in 48-bits, since it involves maintaining intermediate results in two parallel sets of registers.
I doubt that any other manufacturer have done similar - maybe in pockets of the code, but not all the way through.
Which really means it's highly unlikely you are hearing anywhere close to 24-bit resolution on these devices. Those are "marketing" bits, but processing quantization errors reduce the accuracy to not much more than 16 bits at best.
What are pre pros? During this part of the thread I have been trying to figure it out, but without success.
This post is made possible by the generous support of people like you and our sponsors: