There is a slight difference : Mooly said that because you said there were testimonies of breaked Arcam with old caps and because some are of bad quality with an accelerated ageing of their spec : esr... . It is a technical answer to a technical asking. If you re read, his answers are full of logic : measurement, read the datasheet, etc ! Nothing wrong and answers are didactic as well.
I'm not a technician. Mooly is a very good one with a true knowledge behind that. And the apparent simplicity of some answers hide a huge understanding and similar experience, always said and explained with humility with Mooly. You can trust him.
My answers are just based on the perspective to answer you about the difference between the simple CD9 and some waitings about its improvement on a subjective way with a modest tweaking method : swaping some caps.
I believe a good cooking on caps on a well born device adds a lot and sometimes does miracle !
Note the difference between cooking and technic !Technic is more scientific (ok cooking could be as well) and predictable.
Here it is just to allow you to understand the relativity of caps : with close specs they can give different results.... which can help you to find a better setup in relation to the whole system...
You can go further but it asks you a deeper understanding to diy power supply, involves active parts, redesign, etc, etc...
I'm not a technician. Mooly is a very good one with a true knowledge behind that. And the apparent simplicity of some answers hide a huge understanding and similar experience, always said and explained with humility with Mooly. You can trust him.
My answers are just based on the perspective to answer you about the difference between the simple CD9 and some waitings about its improvement on a subjective way with a modest tweaking method : swaping some caps.
I believe a good cooking on caps on a well born device adds a lot and sometimes does miracle !
Note the difference between cooking and technic !Technic is more scientific (ok cooking could be as well) and predictable.
Here it is just to allow you to understand the relativity of caps : with close specs they can give different results.... which can help you to find a better setup in relation to the whole system...
You can go further but it asks you a deeper understanding to diy power supply, involves active parts, redesign, etc, etc...
c36 is in 5V digital dac supply, c44/45 are in audioanalog +/-12V supply for 2114 opamps in dac output
Well I didn’t say I don’t trust Mooly of course. It was just an implicit question. Thanks for the clarification. Sorry if I sounded rude..
Last edited:
I'm still here, and thanks for the kind words folks 🙂
For myself I tend to go for known and good quality 'commercial grade' parts rather than seeking out exotic offerings. If the specs are right, the brand good and the price is right 😉 then I'm happy with my choice.
The real test of any component upgrade, and particularly where there appears to be a big difference in performance, is to do a double blind test... which is not easy on your own equipment of course.
For myself I tend to go for known and good quality 'commercial grade' parts rather than seeking out exotic offerings. If the specs are right, the brand good and the price is right 😉 then I'm happy with my choice.
The real test of any component upgrade, and particularly where there appears to be a big difference in performance, is to do a double blind test... which is not easy on your own equipment of course.
Hi Mooly. 🙂 Do you still think most or all elcaps can be replaced by their Panasonic FC equivalent? Apparently the original Rubycon smoothing caps (C36/44/45) were not all that bad, but the Panasonics (FC) are a small improvement over the old caps. And I suppose it’s possible that bypassing the 2 (reservoir) caps C47/48 may simply have changed the operation/effectiveness of the total so called 'pi-filter' of the power supply of the analog amplification stages, hence the sonic change. So, next I’m planning to replace C47/48 by Panasonic FC. If that gives a further improvement then I may try other cap replacements with Panasonic FC..
Actually I’m still quite skeptical regarding those stories that nearly all electrolytic caps and some other parts of these players need to be replaced/upgraded. I think they may be primarily an advertisement for 'Condor Audio Repair'. 🙂 Nevertheless, something must be wrong with the Alpha 9. This weekend I’ll compare it to the Carver SD/A360..
I know it can be exceedingly difficult to assess (by ear) whether some tweak of your audio system is an improvement or not, especially if there seems to be only a small performance difference. But the difference between the Alpha 9 and CD23T is clearly audible for anyone. The Alpha 9 sounds quite nice overall but clearly lacks in high and low detail/definition.
Yes they probably can all be replaced by that type but as with anything, to make a really informed decision you need to know the specs of the original and base any replacements around that.
The supplier I use doesn't even list the FC type but they do stock the FR amongst many others. These are very slightly larger (same lead pitch though) looking at random examples but they are far better specced than the FC.
Eg, 1000uf/25 volt. Pt no EEU-FC1E102 vs EUFR1E102L
2.4amp ripple against 1.6 for the FC. E.S.R. is 18milliohm vs 38milliohm for the FC. Load life at max ratings. 10,000 hrs vs 5000 hrs for the FC.
So you need to compare what you are fitting with what was originally fitted or if that is not possible, look for the 'best' or most suitable parts that will physically fit. You might the FC is still a 'better' part than the original, if not it may still be 'better' as a newly fitted part compared to the aged part it replaces, and if its temperature ratings are better it may actually last longer than the original.
The supplier I use doesn't even list the FC type but they do stock the FR amongst many others. These are very slightly larger (same lead pitch though) looking at random examples but they are far better specced than the FC.
Eg, 1000uf/25 volt. Pt no EEU-FC1E102 vs EUFR1E102L
2.4amp ripple against 1.6 for the FC. E.S.R. is 18milliohm vs 38milliohm for the FC. Load life at max ratings. 10,000 hrs vs 5000 hrs for the FC.
So you need to compare what you are fitting with what was originally fitted or if that is not possible, look for the 'best' or most suitable parts that will physically fit. You might the FC is still a 'better' part than the original, if not it may still be 'better' as a newly fitted part compared to the aged part it replaces, and if its temperature ratings are better it may actually last longer than the original.
FC and FR both come in many can sizes, FR is the smaller one 99.9% and even with the two random selected above FR has 10mm dia and FC 12.5mm dia.
Well, I swapped (reservoir) caps C47 and C48 of the Alpha 9 with Panasonic FC and it was immediately evident that the strange 'non-linearity' in the (mid) bass had returned. This doesn’t make sense at all. Logic dictates that replacing any 17 year old PS caps with brand new good quality equivalents should improve overall sound.
Now sonically the problem reminded me of a slightly misaligned sub-woofer cut-off. So I turned back the output level of my (active) sub by a tiny amount (less than 1 dBspl) and the problem disappeared almost completely! It also slightly improved clarity of the CD23T. The Alpha 9 is now a step closer in performance to the CD23T. The (mid) bass problem seems to have been (largely) a flaw in the setup of the sub, which was tuned to the Alpha 9 with the old PS caps.
Next I did some frequency response measurements. I have software to measure speaker output and room acoustics using a calibrated microphone and a CD with test tracks to be played by the CDP, i.e., not by the sound card of the computer, which is often of low quality. So this is perfect to compare the frequency response of two CD players.
First I used a pure sine sweep to measure the FR from both players. The result is shown in the first graph below (the blue curve is the Alpha 9, the red curve is the CD23T). Remarkably, the output level of both players is exactly the same and so is the FR as measured via the speakers. Then I used a pseudo-random noise track to measure FR. The result is shown in the second graph. It can be seen that when noise is played, the output level of the Alpha 9 starts to drop (relative to the CD23T) at 5kHz. The drop reaches a maximum of 10dBspl at 15kHz!
I’m not sure what this means, but apparently with the noise signal some cancellation takes place at high frequencies in the Alpha 9 which seems to confirm the lack of detail relative to the CD23T. Hmm.. 😕
Now sonically the problem reminded me of a slightly misaligned sub-woofer cut-off. So I turned back the output level of my (active) sub by a tiny amount (less than 1 dBspl) and the problem disappeared almost completely! It also slightly improved clarity of the CD23T. The Alpha 9 is now a step closer in performance to the CD23T. The (mid) bass problem seems to have been (largely) a flaw in the setup of the sub, which was tuned to the Alpha 9 with the old PS caps.
Next I did some frequency response measurements. I have software to measure speaker output and room acoustics using a calibrated microphone and a CD with test tracks to be played by the CDP, i.e., not by the sound card of the computer, which is often of low quality. So this is perfect to compare the frequency response of two CD players.
First I used a pure sine sweep to measure the FR from both players. The result is shown in the first graph below (the blue curve is the Alpha 9, the red curve is the CD23T). Remarkably, the output level of both players is exactly the same and so is the FR as measured via the speakers. Then I used a pseudo-random noise track to measure FR. The result is shown in the second graph. It can be seen that when noise is played, the output level of the Alpha 9 starts to drop (relative to the CD23T) at 5kHz. The drop reaches a maximum of 10dBspl at 15kHz!
I’m not sure what this means, but apparently with the noise signal some cancellation takes place at high frequencies in the Alpha 9 which seems to confirm the lack of detail relative to the CD23T. Hmm.. 😕
Attachments
What level (on the disc) is the noise track. Some DAC's handle 0db levels differently (post #162).
http://www.diyaudio.com/forums/mult...-power-do-your-speakers-need.html#post2883059
http://www.diyaudio.com/forums/mult...-power-do-your-speakers-need.html#post2883059
I suppose the noise track could be at 0db level (as could the sine sweep). If we take the FR graphs of the sine sweep to represent the correct output level for both CD23T and Alpha 9, then the Alpha 9 selectively depresses high frequencies when presented in a (0dB) noise signal. In music this may manifest itself as less detail/definition.
As far as I know, the DACs of both players are identical. And I don’t think any passive or active filter could do this. I replicated the result with single channel measurements. It looks like some noise induced frequency dependent phase shift between added signals occurs in both channels, somewhere in the CDP.. 😱
You need to look at the noise on a scope as I did in the other thread. That shows instantly if there is any obvious limiting event though you are only averaging by eye.
Music CD's wont come close to having HF energy at the true 0db level.
Music CD's wont come close to having HF energy at the true 0db level.
I might borrow a scope. However, given that different DACs may handle 0dB signals differently, a sound spectrum analyzer should not use 0dB test signals. Moreover, at a given volume setting I measured through the speakers that the noise signal peaks at 84.9 dBSPL, whereas at the same volume setting a normal music track (Pixies: Indy Cindy) peaks higher at 89.8 dBSPL.
I checked the DAC chips of both players and they have different serial numbers on them and so do the DSP chips, but I’ve never heard that the DAC (and/or DSP) of both players is actually different.
What if this HF roll-off with the noise signal in the alpha 9 isn’t caused by some 0dB limiting? Could it be a clock related (timing) issue? After all, relative to the Alpha 9, the CD23T has "improved buffering of clocks and audio data and improved master clock and clock output distribution circuitry"..
I checked the DAC chips of both players and they have different serial numbers on them and so do the DSP chips, but I’ve never heard that the DAC (and/or DSP) of both players is actually different.
What if this HF roll-off with the noise signal in the alpha 9 isn’t caused by some 0dB limiting? Could it be a clock related (timing) issue? After all, relative to the Alpha 9, the CD23T has "improved buffering of clocks and audio data and improved master clock and clock output distribution circuitry"..
I'm no expert at all on acoustic measurements with microphones other than very simple observations I have made with a basic SPL meter. I do know that miniscule differences in position can influence the result dramatically due to reflections/absorption and so on. To my way of thinking the only definitive test has to be on what comes out of the line out sockets on the players. Clock and timing variations will not have any influence on what you are looking at.
Well, you don’t need to be an expert to make some valid sonic FR and SPL measurements. Just get yourself a good (calibrated) microphone and software that can work with it. If you make sure that in subsequent measurements you and the microphone are in the exact same position/location, then any measured difference in FR/SPL represents an equivalent difference at the source. If you don’t want to include room acoustics in your measurements, then you simply place the microphone within about 10cm distance of the speaker to be measured. I did that to replicate the result that the reproduction of HF noise is at a lower level in the Alpha 9 than in CD23T (see the graph below; red is the CD23T and blue is the Alpha 9). And a close range measurement (with the sub off) of a sine sweep (again) corresponded exactly with the red curve for both players. So, this difference is real and as such it will be present at the line output of the players. As for the peak SPL measurements, the noise signal cannot be a 0dB signal because the measured peak SPL is (much) lower than the peak level in a normal music track. This is scientific proof of that fact.
Hm.. Don’t you think that bad timing/synchronization of additive processes could lead to cancellation of especially phase shifted high frequencies in noise? Anyway, there must be an explanation for this phenomenon other than 0dB limiting effects..
Attachments
I know when listening to a test tone via speakers, that head movement of just a few millimetres can almost null the sound completely. So I assume microphone placement will be just as critical. As I say though, acoustics isn't my field I'm afraid and so I'm not certain of the facts and how reliable your measurement technique would be.
I can't see any way timing errors would affect the audio output in this way.
I can't see any way timing errors would affect the audio output in this way.
if the microphone isn’t touched or moved in any way in between measurements, then don’t you agree that any measured FR/SPL difference must be a real difference, present at the source?
Well, the basic principle of the dCS RINGDAC is that subsequent 16bit values are randomly allocated for processing to one of several (five in fact) DACs (hence the name Ringdac). If the recombination of the output of these different DACs isn’t timed correctly, then I think it could effect (phase shift) primarily high frequencies..
I still can't see any realistic way that timing errors would produce a difference that was measurable via a mic set up.
Call me old fashioned 😉 but I think you need to dangle a scope on the end of the players and have a look at what the noise test track looks like. You could also try and record the output of the two players with Audacity, then working with just one channel from each, equalise the levels of the two, invert one and see what the difference looked/sounded like.
Call me old fashioned 😉 but I think you need to dangle a scope on the end of the players and have a look at what the noise test track looks like. You could also try and record the output of the two players with Audacity, then working with just one channel from each, equalise the levels of the two, invert one and see what the difference looked/sounded like.
I explained how timing errors could affect high frequencies in a Ringdac. If you don’t accept that line output can be validly measured via the speakers through a microphone, then that’s your problem.
Dangling a scope on the end of the players will not change the presented conclusions.
I don’t see what that could possibly add to what we already know by now..