I'll agree the voltage source for the pin (VCOM) comes from the DAC IC's internal sources. AK gives the values for those resistors. You can also prove it by putting a 1k resistor to ground and seeing how much the voltage sags on the pin. I have done this on quite a few DAC IC's. Even to 1,000uF on the pin as Cirrus suggested.
If you use an op-amp to drive the pin, it will override the IC's internal resistor network and that's it really. It doesn't damage anything. If you note AK mentions the time-constant for this pin on power-up, when the capacitor is discharged (shorting the pin to ground) it takes a while to come up.
But SunRa is using an op-amp buffer to drive the REF pin?
If you use an op-amp to drive the pin, it will override the IC's internal resistor network and that's it really. It doesn't damage anything. If you note AK mentions the time-constant for this pin on power-up, when the capacitor is discharged (shorting the pin to ground) it takes a while to come up.
But SunRa is using an op-amp buffer to drive the REF pin?
My point is that the VREF has to be an input, and an op-amp is a good idea, especially if your reference voltage might droop if more current is drawn from it. I have designed with dedicated reference voltage chips, but note that they're rather weak in terms of output current, so if I need to drive more than one pin from the reference, I'll use an exceptionally accurate op-amp to increase the current supply without allowing the voltage to drop.
VCOM is not supposed to be an input, so my recommendation is not to put an op-amp on that pin just because an op-amp is a good idea on VREF.
My supposition is this: If you have to ask whether you can use an op-amp for VCOM, then you probably have no business going against the data sheet recommendations. If you knew enough to improve things by using an op-amp for VCOM, then you wouldn't have to ask. I'm not saying you shouldn't feel free to experiment, but I would not recommend designing a first AK4396 board with a circuit that goes against the data sheet. But do whatever you want and tell us how it works out.
Using an active regulator on VCOM raises the question of; what exactly does one hope to achieve by doing so? The whole point of the decoupling capacitor on VCOM is to shunt any noise or signals appearing there to A.C. ground. While an op-amp could provide a lower impedance A.C. ground at the lowest frequencies than could any reasonably sized bypass capacitor, an op-amp itself generates noise and is subject to power-supply coupling.
Again, I haven't tried this on AK parts- but put a 'scope on that pin and what would you expect? On many DAC IC's it's not a flat line... Without a cap to GND, the VCOM pin has an impedance equal to the two divider resistances (from AK datasheet 750-975 ohm ea.) or 375-488 ohms and this is drastically lowered by the bypass cap, until you hit the low frequency cutoff.
Yes the op-amp adds noise unless you filter it, but it firms up the low end and gets rid of some mush in the bass. Note the Buffalo VREF circuit has two 1uF o/p caps to filter noise there.
This is just something to try, for those who experiment outside the datasheet.
This is a very important observation. Even a so-called rail-to-rail op-amp will fall short of the full 5 V by a few mV. Most op-amps do not run 'saturated,' they simply put out 4.95 V or less - you can find the maximum output voltage on the data sheet for the op-amp.
If you need a voltage-follower to perform correctly at 5 V, then you need to supply the op-amp with much more than 5 V power.
This may not be too inconvenient, since a DAC fed with 5 V may indirectly create higher voltages than 5 V after the low-pass due to Gibbs phenomenon. Therefore, it's beneficial to have an analog supply that is higher than 5 V by about 2 dB, and once you have that then you can handle both the voltage-follower on the 5 V VREF and handle the analog output of the DAC without clipping post low-pass.
On that note, perhaps such designs as prairiemystic is examining have a hidden benefit: By using an underpowered op-amp voltage-follower, VREF becomes 4.9 V instead of 5 V. Thus, the entire conversion of the DAC runs at a smaller peak-to-peak voltage. Then, when low-pass filtering introduces overshoots in the waveform, there just happens to be some headroom up to 5 V (assuming that there are better rail-to-rail op-amps in the audio section, which I guess is not really a safe assumption in this case). Anyway, there are global considerations in the circuit as well as focused details.
P.S. Some ADC chips have VREF at 1/2 scale (which only works with certain types), so it is much easier to use an op-amp where the DC input and output are at 50% of the power.
The AK4396 datasheet (pg.34, paragraph 2) says that VCOM is a signal ground, and that no load current can be dawn from it. Which would mean that any active D.C. regulator will draw load current and cannot or, at least, should not be used. The DAC may not self-destruct, but I should think that some performance parameter would be adversely affected. Else, why would AKM explicitly state that no load current can be drawn from that pin. However, sometimes it's good not to be a slave to the datasheet, if one is prepared to take the risks which come with taking such liberties.
Thanks for all the comments regarding this vcom pin, really important stuff here, espeaceally the comments regarding the operation of the buffer opamp in the buffallo. I am not certain, but based on their schematic of the first buffalo, I believe the opamp might be supplied directly from the incoming voltage line (which can be a >5v pre-regulated source) feeding the onboard regulators. So it might be possible that the opamp has enough voltage to let the full 5V to pass through the vcom pin.
I hope in the following days I'll come with an updated schematic and with a few interesting design choises. I am very decided to go through this as the exa U2I with the breadboard AK4396 really shines and I want to get the most out of it.
I hope in the following days I'll come with an updated schematic and with a few interesting design choises. I am very decided to go through this as the exa U2I with the breadboard AK4396 really shines and I want to get the most out of it.
That would be an interesting choice. Any noise on an op-amp supply pin will usually appear on the output, so it seems like a really bad choice to buffer an important reference signal with an op-amp fed from a pre-regulated supply.
An externally hosted image should be here but it was not working when we last tested it.
the tst1 pin should be left open?
Something happened with the forum, the buttons in the reply interface are not working anymore (quote buttons, smiles, etc). At least not on my desktop and laptop...
"SunRa, I was wondering if you had looked at the AK usb receivers? A full AK usb dac sure would be fun
"
I wasn't aware of any USB interfaces from AK, thanks for pointing this up. I have so much fun with the U2I that I doubt I will go after other interfaces. I've even played 384Khz and 352.8Khz, files on the AK4396.
"the tst1 pin should be left open?"
Yes according to the datasheet. The schematic above is the one I build on a breadboard and tested the U2I interface with. It's not the final one. I will come back with details.
"SunRa, I was wondering if you had looked at the AK usb receivers? A full AK usb dac sure would be fun
"I wasn't aware of any USB interfaces from AK, thanks for pointing this up. I have so much fun with the U2I that I doubt I will go after other interfaces. I've even played 384Khz and 352.8Khz, files on the AK4396.
"the tst1 pin should be left open?"
Yes according to the datasheet. The schematic above is the one I build on a breadboard and tested the U2I interface with. It's not the final one. I will come back with details.
Why don't you provide 45.1584 MHz master clock to AK4396 for playing 2L genuine DXD sources without "hiss" noises? As long as you use 22.5792 MHz master clock, an interpolating 8 x over sampling filter in AK4396 underruns, stalls and generates unpleasant noises.
Last edited:
Thanks for your reply Bunpei. Well one of the reasons is that I am using the exa device and I don't know if that can be accommodated to support a 45MHz clock.
Also I am not sure it is a hardware problem because as I stated repeatably I've played upsampled files to 384 and 352.8Khz without any noise. The only files I have problems with are those DXD files downloaded from the 2L site.
I believe the source of the noise is the way the DXD format is specified. I learned that the antialiasing low pass filter employed in the ADC for DXD recording is much more shalower than ususal (requirement for DSD files), so I dont believe I am experiencing the same problems you guys had with the Buffalo DAC but rather the effects of this passband noise.
In support of this supposition, I've applied a digital low pass filter to the DXD files and they played just well, without noise (at 352.8Khz).
What are your opinions on this?
Dear SunRa,
First, may I ask you what frequency of MCLK are you providing to AK4396?
(My assumption is you are injecting 22.5792 MHz from your exaU2I device.)
I think you know very well that 22.5792 MHz MCLK is only effective for 128fs of 176.4 KHz according to AK4396 datasheet. If you extrapolate the relation for 352.8 kHz, 128fs of 352.8 = 45.1584 MHz MCLK is essential for a "normal" play.
If you use 22.5792 MHz for 352.8 kHz play, it is just an under-clocking abnormal play. I hope you can understand this point at first.
Bunpei
First, may I ask you what frequency of MCLK are you providing to AK4396?
(My assumption is you are injecting 22.5792 MHz from your exaU2I device.)
I think you know very well that 22.5792 MHz MCLK is only effective for 128fs of 176.4 KHz according to AK4396 datasheet. If you extrapolate the relation for 352.8 kHz, 128fs of 352.8 = 45.1584 MHz MCLK is essential for a "normal" play.
If you use 22.5792 MHz for 352.8 kHz play, it is just an under-clocking abnormal play. I hope you can understand this point at first.
Bunpei
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.