Let me preface this post say a couple of things. One, The Denon AVR-1909 design is fairly good one. My system really is a rather modest one, using Polk R30's and R15's with a CS101 center and Definitive Technologies Pro-Sub-1000. I started by breaking the elements into subsystems and what I could do with little money a project to see just what could be done. One of the things, is the analog chain, and no matter the system, it always has to come back analog to be played over the speaker system. This was one of the crucial starting blocks, and having done reworks since the 70's, I know that signal chain caps aree the first place to start. People who think various brands and deisgns of caps make no matter simply know no better. Long gone are the days or Rel's or WonderCaps, and now we have lines like MUSE and my personal choice for analog, and that the ELNA SILMIC II, purposely built for this application, and hands down the sweetest and finest signal chain electrolytic on the market today.
There is a certain sweetness and tonal quality only to be found from an electrolytic, and is my favorite. After examining the schematic for the AVR-1909, Denon made a good choice in the value to use, but SUCKED in their usage of SAMYOUNG caps. You couldn't have chosen a worse cap to pass signal through. So I purchased a hundred 10uF/35v from DigiKey for $12 and reworked from the schematic any cap that signal passed through on the mainboard to the finals and the input board for the SR and SL channel to compare. I reassembled the unit to get a feel of what affect that had and was pleasantly surprised at the tonal change that was emerging in the EXT. IN channels which carried my SACD info, however this lead to some interesting finds.
I had used a Lissajous pattern to look at what was going on in comparing the input signal against the analog chain, and I found that 100Hz was often as much 60 degrees off from the source, with high overshoot, ring and poor recovery, and high end almost strained to well over -120 degrees from the source, which lead me to the op amps. Sure enough, JRC2068's are stuck in this thing which has a slow slew rate of 5~6V/μs, and not even specs in the schematic, but the schematic wasn't much better with the AZ4580 which is only 7V/μs, both having really crummy noise levels and THD for this usage...maybe a car amp, but not a home entertainment system. Slew rates needs to be 3 times that. In the pinnacle of analog we came to find that 20V/μs is a good target for signal chain and can keep up with complex passages, which the JRC was doing a really horrid job with. The JRC has an advertised overshoot of guaranteed 15% and could even be more if at rail to rail voltages, which was clearly appearent on the scope. They simply cannot keep up faithfully to the source. And this is where the goofy engineer or 'bean counter' who chose this got you and me.
So back to engineering 101 we go, and selecting a quality op amp, and Burr-Brown is known as an industry standard and the SoundPlus line is a good solid base to choose from. There are others, but the TI/Burr-Brown's are known for their warmth and good performance specs hard to beat. So selecting an op amps started with a total look at what I had to work with inside the AVR-1909, and examining the ADC as well as the DAC would yeild a good choice to use in the analog chain as well. One of the criteria I had was to find the 'right' op amp so that I wouldn't have to change loading, decoupling or resistors (basically without having to redesign Denon's original one) and keeping the current draw simular to what was already there.
Several key elements must coalesce for effective analog front-end design in the signal path. The typical signal path’s analog front-end includes an op amp that drives the ADC, an RC filter, and the microcontroller or digital signal processor (DSP). When interfacing an ADC with an op amp, it’s imperative to understand the specifications that are important to get the expected performance results. Modern ADC ac specifications such as THD, SNR, settling time, and SFDR are critical for filtering, test and measurement, video, and reconstruction applications. The high-performance op amp’s settling time, THD, and noise performance must be better than that of the ADC it’s driving to maintain the proper system accuracy with minimal or no error.
Noise is a very important specification for both the op amp and the ADC. Three main sources of noise contribute to the overall performance of the ADC - quantization noise, noise generated by the ADC itself (particularly at higher frequencies), and the noise generated by the application circuit. For example, the 24-bit ADC of the ADAU1328 used in the AVR has an SNR of 108dB. To achieve better SNR, the ADC driver noise should be as small as possible. The Rohm BA4510 used in the AVR has an abismal 80dB and slew rate of only 5V/μs. Now in designing the proper op amp for ADC side, when driven single ended to differential, you must double the noise and THD of the op amp used, since you are using both sides to drive a single channel of the CODEC. This means to starting out the gate, the op amp you have in your AVR created twice the noise in one single channel then both channels combined in the CODEC (BAD BAD) and THD of the CODEC is 0.0019% and just one side of the BA4510 is 0.01% again BAD BAD! So we flood out analog to digital converter with huge noise and harmonic distortion, not counting the fact that the op amps slew rate is so slow, it cannot even keep up with our content. Don't even get me started on overshoot and unble to stop ring. The combined settling time of the op amp and the ADC must be within 1 LSB. The 0.01% settling time of the original is 6us. That's microseconds....the ADAU1328 is 300ns...that's nano seconds.... Everything that has to go to DSP or Audssey passes through the ADC chain.
Signal-to-noise and distortion (SINAD) is a parameter that combines the SNR and THD specifications. SINAD is defined as the RMS value of the output signal to the RMS value of all the other spectral components below half the clock frequency, including harmonics but excluding dc. Because SINAD compares all undesired frequency components with the input frequency, it’s an overall measure of an ADC’s dynamic performance. Also, the op amp’s critical parameters—THD, settling time, and noise—must be taken in to account to interface a high-performance op amp with an ADC. And remember, that using this choice in Denon's design of single ended to differential input requires us to double the noise and THD figures. The choice we have made is the OPA2365 which has in comparison to the BA4510, a noise of 4.5nV/Hz at 100kHz, slew rate of 25V/μs and THD+N @ 0.0004%. This is a much better choice and meets every aspect of our need to properly drive the ADAU1328 clean, without additional noise and vanishingly low THD. Every bit is driven to 1LSB and settling time equal to the ADC at 0.3μs to 0.01%. The hard part to select this op amp is that it has to be a 3.3v device, and a good choice for audio, which it certainly is. And to hear it is to believe it!
Next we need to address DAC (digital playback and analog side as well. The criteria really falls within the same specs as we use for ADC, we simply go out the other side. This why we use an Burr-Brown OPA1602 which has even lower noise at 2.5nV/√Hz at 1kHz, lower distortion (THD+noise) @ 0.00003% at 1kHz and the slew rate we need... 20V/μs and works within Denon's original op amp power specs with actually a slightly lower current draw (400mV compaired to 500mV) With this choice we go from an measly 85dB of dynamic range with the originals to well over 130dB...stop and think about that for a second... Channel separation went from 90dB a channel to well over 130dB as well.
The real problem with many folks who choose to change op amps don't properly select one and ends up in oscillation or use a jfet where a bipolar should be used, and these two selections are both bipolar and are drop-in replacements without mods of any kind.
I contacted Mouser online and obtained 10 OPA1602's and 2 OPA2365's for around $50.00 shipped. And this is where things start getting interesting...
I take and in the SR and SL channels in the analog side place in the EXT. IN path an OPA1602 and reassemble to see what happened. I took another Lissajous view of the orginal LR channel and the modded SR channel with front right signal only from the source and find that the modded channel probed at the finals right before the darlingtons and made a quick recording of the original (A1) and the modded channel (B1) which I will attach, and the orginal is still the same but the modded channel is polar responsive nearly exactly to the source, and I mean EXACTLY...I'll include the A1 and B1 file and you can see if you examine on a spectrograph just what you are dealing with from our non-friend, the 'bean counter'... Overshoot, loss of control, loss of high frequencies, to out and out loss of program matierial. Polar opposites on high frequencies when confronted in complex passeges...what does this mean to my ear? Smeared or loss of content, unable to control bass when you know it shoudln't sound muddy, but does... Quality sound starts with reproducing the source as faithfully as possible, and it is if you use quality parts. It wouldn't have costed Denon, or me and you another 50 bucks per unit to have had sonically faithful reproduction.
The design of this amp is good, however, their using cheap parts in these two areas make it just so-so, but the building blocks are already there, just this simple signal path upgrade makes a world of difference, and that ain't no BS. Come listen for yourself.
I cannot post pics yet...pooooo
