phwatts
New Head-Fier
- Joined
- Jan 2, 2010
- Posts
- 2
- Likes
- 0
Dear all,
Since the Essence ST's performance is finally good enough to compete with serious component hi-fi, I'm taking the plunge to replace my Meridian setup.
However, I think there are still a few areas where the ST can be improved, but despite hundreds of postings & threads it seems most are only going in circles with trying opamp permutations. Without treading on toes, simply spending lots of money on exotic opamps while there is so much room for improvement elsewhere is IMO similar to using sandpaper on a rough tree log. At least there aren't long discussions on how one brand of boutique solder made a quantum leap difference Don't get me wrong, I also do it, but I would prefer to leave component tweaking by ear for last.
I was wondering who might be interested in being part of it, either just by means of comments & input or actively participating.
I'm an EE specializing in digital design and multilayer PCB layout, mostly digital media interfaces for audio&video so I'm familiar with the environment - I do not consider the proposed project as over-ambitious, but it will definitely take considerable time.
As my intention is both stereo & HT, it will require the HDAV module too. However, simply swapping the PCM1796 with PCM1792 DAC's is not good enough.
I have identified four problem areas:
1) Despite the on-board regulation and "HyperGrounding", the power & grounding environment in a PC is hardly clean and a far step from component hi-fi.
2) The quality and complexity of the I-V and summing stages are low-quality and similar to sub-$500 CD players. Although the DAC's and opamps are rather good, filter & decoupling capacitors are ceramic and MKT instead of MKP.
3) Unless the expansion board connects with TDM, the digital audio appears to be daisy-chained I2S with single-ended lines; hardly a good idea given the high clock frequencies. Given the high clock accuracy, going for ASRC will (probably) not improve things though.
4) It appears that volume is adjusted either in the processor or in the DAC's by means of the SPI interface. Either way, digital attenuation is an undesirable method since it reduces dynamic range.
I therefore intend on doing the following. The exact details and extent will depend on the actual implementation as I haven't bought my ST yet and would first like to get some answers from current owners:
1) Build a proper external supply for the ST card with a 50VA toroidal transformer and regulation. This will reside within the PC chassis.
2) Disable/remove all unnecessary circuitry to reduce power consumption and ground pollution. This includes the DAC, should the headphone out not be used.
3) Add a small daughterboard that plugs into the expansion header, with iCouplers and differential line drivers (LVDS) for the digital audio signals. This board has its own high-speed breakout connector. The SDATA signal from the front channels will be tapped from the board and added here, so that all 8 channels are presented on the header. This method assumes that I2S is used for the audio, but it could be TDM. This connector now provides properly tailored audio, completely isolated, for all eight channels. The LVDS electrical format also allows a fairly long cable without loss, compared to the limitation of 3.3V single-ended TTL of the stock connection. There are neat 110ohm differential cables to choose from; CAT6 bundles could be used if really necessary.
4) Build a high-end 8-channel external DAC; residing outside the PC chassis. This will be similar to the HDAV board but with a few notable differences:
- Dedicated power supplies with transformers. This will include a properly regulated 3.3VDD, 5VCC and +-15V.
- Much better quality filter components. Not going overboard with $1300 Audio Note capacitors though! Wima MKP/FKP has worked very well in past projects.
- Volume control in the analog domain by means of TI's PGA2320. I've used it in numerous projects and it works excellent. Admittedly it's an extra opamp in the signal path, but it's a much better trade-off compared to digital attenuation.
- Proper line drivers for long cables and improved bass. Choices I've tried in the past are BUF634 or THS6012; the latter is very similar to TPA6120A2 but intended for line drivers.
- PCM1794A DAC's. The reason to use it above the PCM1792 is simple, all the settings that is currently software controlled is unnecessary and can be hard-configured, and the circuit path is much cleaner without having more noisy digital lines. The only caveat is volume control, which depends on how the ST implements it. Hopefully it's done in the DAC's via the SPI bus. In that case, a small MCU can be used to hijack the SPI bus, identify the volume control signalling and re-format it for the PGA chips. However, if attenuation is performed in the processor, a different strategy will be required. By far the easiest would be to leave all the software at max volume and build a custom volume control on the DAC itself. This is easy by means of a cheap pot acting as voltage reference with an MCU with internal ADC, which translates it into SPI for the PGA chips. I have done all of this before.
There's a choice to opt for a single large board, or rather 4x identical stacked stereo boards, each with its pros&cons. The implementation of the volume and TDM vs. I2S, and generally how much duplication will be possible will play a role in this. The increased real estate and lower production cost will allow adding more separation such as individual regulation, and dual-mono for increased dynamic range by summing the currents and using one DAC per channel. This I have also done before.
Anyways, I'm looking forward to comments. Of particular help would be current ST owners (preferably those with HDAV boards) to do some reverse-engineering in places. Simply checking connectivity with a continuity meter will already help a stack in order to determine some of the unknowns. I hope that this will allow me to avoid buying an HDAV board. The exact clocking will also determine whether the excessive signal tailoring is necessary - even though benign, buffers & line drivers add jitter so it might be that cons outweigh the pros. All the same can be done with straight TTL though.
Naturally this will not be cheap, but rather benign compared to the price of the component hi-fi it will contend with.
Regards,
Pierre
Since the Essence ST's performance is finally good enough to compete with serious component hi-fi, I'm taking the plunge to replace my Meridian setup.
However, I think there are still a few areas where the ST can be improved, but despite hundreds of postings & threads it seems most are only going in circles with trying opamp permutations. Without treading on toes, simply spending lots of money on exotic opamps while there is so much room for improvement elsewhere is IMO similar to using sandpaper on a rough tree log. At least there aren't long discussions on how one brand of boutique solder made a quantum leap difference Don't get me wrong, I also do it, but I would prefer to leave component tweaking by ear for last.
I was wondering who might be interested in being part of it, either just by means of comments & input or actively participating.
I'm an EE specializing in digital design and multilayer PCB layout, mostly digital media interfaces for audio&video so I'm familiar with the environment - I do not consider the proposed project as over-ambitious, but it will definitely take considerable time.
As my intention is both stereo & HT, it will require the HDAV module too. However, simply swapping the PCM1796 with PCM1792 DAC's is not good enough.
I have identified four problem areas:
1) Despite the on-board regulation and "HyperGrounding", the power & grounding environment in a PC is hardly clean and a far step from component hi-fi.
2) The quality and complexity of the I-V and summing stages are low-quality and similar to sub-$500 CD players. Although the DAC's and opamps are rather good, filter & decoupling capacitors are ceramic and MKT instead of MKP.
3) Unless the expansion board connects with TDM, the digital audio appears to be daisy-chained I2S with single-ended lines; hardly a good idea given the high clock frequencies. Given the high clock accuracy, going for ASRC will (probably) not improve things though.
4) It appears that volume is adjusted either in the processor or in the DAC's by means of the SPI interface. Either way, digital attenuation is an undesirable method since it reduces dynamic range.
I therefore intend on doing the following. The exact details and extent will depend on the actual implementation as I haven't bought my ST yet and would first like to get some answers from current owners:
1) Build a proper external supply for the ST card with a 50VA toroidal transformer and regulation. This will reside within the PC chassis.
2) Disable/remove all unnecessary circuitry to reduce power consumption and ground pollution. This includes the DAC, should the headphone out not be used.
3) Add a small daughterboard that plugs into the expansion header, with iCouplers and differential line drivers (LVDS) for the digital audio signals. This board has its own high-speed breakout connector. The SDATA signal from the front channels will be tapped from the board and added here, so that all 8 channels are presented on the header. This method assumes that I2S is used for the audio, but it could be TDM. This connector now provides properly tailored audio, completely isolated, for all eight channels. The LVDS electrical format also allows a fairly long cable without loss, compared to the limitation of 3.3V single-ended TTL of the stock connection. There are neat 110ohm differential cables to choose from; CAT6 bundles could be used if really necessary.
4) Build a high-end 8-channel external DAC; residing outside the PC chassis. This will be similar to the HDAV board but with a few notable differences:
- Dedicated power supplies with transformers. This will include a properly regulated 3.3VDD, 5VCC and +-15V.
- Much better quality filter components. Not going overboard with $1300 Audio Note capacitors though! Wima MKP/FKP has worked very well in past projects.
- Volume control in the analog domain by means of TI's PGA2320. I've used it in numerous projects and it works excellent. Admittedly it's an extra opamp in the signal path, but it's a much better trade-off compared to digital attenuation.
- Proper line drivers for long cables and improved bass. Choices I've tried in the past are BUF634 or THS6012; the latter is very similar to TPA6120A2 but intended for line drivers.
- PCM1794A DAC's. The reason to use it above the PCM1792 is simple, all the settings that is currently software controlled is unnecessary and can be hard-configured, and the circuit path is much cleaner without having more noisy digital lines. The only caveat is volume control, which depends on how the ST implements it. Hopefully it's done in the DAC's via the SPI bus. In that case, a small MCU can be used to hijack the SPI bus, identify the volume control signalling and re-format it for the PGA chips. However, if attenuation is performed in the processor, a different strategy will be required. By far the easiest would be to leave all the software at max volume and build a custom volume control on the DAC itself. This is easy by means of a cheap pot acting as voltage reference with an MCU with internal ADC, which translates it into SPI for the PGA chips. I have done all of this before.
There's a choice to opt for a single large board, or rather 4x identical stacked stereo boards, each with its pros&cons. The implementation of the volume and TDM vs. I2S, and generally how much duplication will be possible will play a role in this. The increased real estate and lower production cost will allow adding more separation such as individual regulation, and dual-mono for increased dynamic range by summing the currents and using one DAC per channel. This I have also done before.
Anyways, I'm looking forward to comments. Of particular help would be current ST owners (preferably those with HDAV boards) to do some reverse-engineering in places. Simply checking connectivity with a continuity meter will already help a stack in order to determine some of the unknowns. I hope that this will allow me to avoid buying an HDAV board. The exact clocking will also determine whether the excessive signal tailoring is necessary - even though benign, buffers & line drivers add jitter so it might be that cons outweigh the pros. All the same can be done with straight TTL though.
Naturally this will not be cheap, but rather benign compared to the price of the component hi-fi it will contend with.
Regards,
Pierre