Hi Mark, Thanks for the detailed explanations. Still will need some time to wrap my head around them. When I tried connecting with the Sparkfun board, I forgot to remove the 1k pullup resistors. Maybe that's why it does not connect. Anyway I will modify the i2c program for the pullup. I will also order the diodes and implement them. Your comments are always very clear, thanks for taking the time. In the meantime, I will do it carefully. Connecting i2c only when both the devices are powered and disconnect them before shutting down. Regarding the programs, your notes are very easy to follow. Though the language will take sometime for a beginner like me. And I didnt even know that Windows Calculator had the programmer function! Will have more questions going forward as always.
Kay
Kay
Hi Kay,
The I2C program should not be modified for the pullup. The program already sets the internal arduino pullup resistors to disabled. My point was a warning not to change that.
Also, the 1k resistors probably don't account for the translator not working. They are simply in parallel the existing dac board pullup resistors, so that is 4.7k in parallel with 1k. That may be verging on a heavy load to be switched by the MCU if it was hooked up, and maybe a heavy load for the translator (not sure off the top of my head, I would need to check the data sheet). In any case, 4.7k should be a reasonable value for I2C pullups.
The important thing with the translator is that it is connected to Vcc on both the dac side and the arduino side, and that the higher voltage Vcc (should be the arduino) is connected to port 2 on the translator. If Vcc is present on both sides (and ground is connected to the translator on both sides), and you short one of the inputs to ground on one side, say you short SDA to ground on one side. Then the SDA pin on the other side of the translator should also go low near ground. You could see if it works with a DC voltmeter and test clip to short one side to ground.
EDIT: To be clear, the translator should have one common ground that is shared by both sides somewhere. It is not a galvanic isolator device.
-Mark
Last edited:
Hello,
I recently read a paper from TI where the THS4551 is used to optimize the THD+N in a Data Aquisition System incooperating a low noise ADC and I wonder if the THS4551 will be a good choice for audio applications.
The only down side with the precision fully differential Opamps from TI is that they only accept 1Vpp max on the inputs.
BTW, for galvanic isolation the DAC and the Arduino you could use a chip out of the ADUM series from Analog Devices.
I recently read a paper from TI where the THS4551 is used to optimize the THD+N in a Data Aquisition System incooperating a low noise ADC and I wonder if the THS4551 will be a good choice for audio applications.
The only down side with the precision fully differential Opamps from TI is that they only accept 1Vpp max on the inputs.
BTW, for galvanic isolation the DAC and the Arduino you could use a chip out of the ADUM series from Analog Devices.
There are other differential opamps. TP uses one their Mercury output stage. Some of the differential opamps tend measure as having very low distortion. However, a successful audio designer I know says they don't sound as good as some of the numbers might suggest. Could be, IMHO. Haven't tried any of them myself.
Si860x can be used. The same guys make galvanic isolators for I2S. They work well at DC and low frequencies, but there is often some capacitive coupling between ground planes that are physically close to each other that may limit isolation effectiveness at HF/RF.
Hi Mark/members,
Yay, I successfully set DPLL of PCM & DSD to best i.e. no. 1. So far I've listened to about 6 tracks. Initially I had a couple of pops. This I believe is because of the connection of the lifted pins which I reconnected simply with pieces of pins trapped under. Apart from that, its playing flawlessly. Tried direct flac and also upsampled to dsd256 & 512. Regarding the SQ, my impression is that bit less background noise and more open. Don't know, could simply be placebo effect. Anyway am really satisfied.
Kay
Yay, I successfully set DPLL of PCM & DSD to best i.e. no. 1. So far I've listened to about 6 tracks. Initially I had a couple of pops. This I believe is because of the connection of the lifted pins which I reconnected simply with pieces of pins trapped under. Apart from that, its playing flawlessly. Tried direct flac and also upsampled to dsd256 & 512. Regarding the SQ, my impression is that bit less background noise and more open. Don't know, could simply be placebo effect. Anyway am really satisfied.
Kay
Yes, you definitely need to appropriately layout your PCB.
But the isolated sides of suchs ICs have a distance that is greater than 3mm and therefore have enough isolation and the chance that there is a coupling between the isolated sides is very low.
In one of our designs we used a multiplexed 8MHz SPI signal to sport four different DACs each of which were isolated through an ADUM250N.
We used an isolation distance of 3.5mm and there is absolutely no problem.
Hi Kay,
If you reconnect the MCU, it may set the DPLL back to the default values. Did you listen to the DPLL settings with the MCU disconnected, or after you reconnected it?
-Mark
Last edited:
Of course. The implementations I usually see use about a 1mm gap between ground planes. The also use the narrow profile isolator chips. Probably, they do it that way to save board space. Of course, that's not the only type of commonly seen problem either.
if you hear any background noise, then something is wrong with whole setup of yours.
unless, you listen vinyl rips of course..
Looks like DigiKey has created an order page for AK4499, although they show it as back ordered. Maybe they are tired of people asking about it?
https://www.digikey.com/products/en?keywords= 974-1192-ND
https://www.digikey.com/products/en?keywords= 974-1192-ND
There are registers to control the volume. There is also a bit you can set somewhere to control the volume of both channels using only one register. The registers actually turn down the volume, so I guess one might call them attenuators.
Hi Ivo,
ESS has some info on that type of AVCC circuit on the downloads page of their website: http://www.esstech.com/files/4514/4095/4306/Application_Note_Component_Selection_and_PCB_Layout.pdf
They just used 10uf or 47uf elctrolytics. That's what I used for an earlier dac too. Those should probably be fine, IMHO.
Also perhaps worth mentioning about some good opamps to use, IME. I have been using OPA1612 for everything as of recent (AVCC, I/V, differential summing). They and many other opamps tend to sound better with some several uf of film caps from each of the +-15v rails to ground. Some people have reported better sound quality with as little as 15uf of film caps per rail. Similarly good results may obtainable from using high gain audio-friendly voltage regulators such as Jung-Didden, or especially if cost and or size is an issue, perhaps Nazar regulators. We can probably provide some more info about the latter if you are interested.
Another opamp that is very good for AVCC is AD797. However, those particular opamps usually work better with no film caps on the +-15v rails. Too much low ESR capacitance on the power rails can cause stability issues in that case, IME.
However, film caps are probably worth a try if using other good audio opamps such as OPA1612 or LME49720/LME49860. The caps tend to help get rid of the last bit of graininess in the sound that may perhaps be most easily noticeable for things such as cymbals.
Also if interested, the AVCC schematic we use can be found attached to post #3003 in this thread.
-Mark
ESS has some info on that type of AVCC circuit on the downloads page of their website: http://www.esstech.com/files/4514/4095/4306/Application_Note_Component_Selection_and_PCB_Layout.pdf
They just used 10uf or 47uf elctrolytics. That's what I used for an earlier dac too. Those should probably be fine, IMHO.
Also perhaps worth mentioning about some good opamps to use, IME. I have been using OPA1612 for everything as of recent (AVCC, I/V, differential summing). They and many other opamps tend to sound better with some several uf of film caps from each of the +-15v rails to ground. Some people have reported better sound quality with as little as 15uf of film caps per rail. Similarly good results may obtainable from using high gain audio-friendly voltage regulators such as Jung-Didden, or especially if cost and or size is an issue, perhaps Nazar regulators. We can probably provide some more info about the latter if you are interested.
Another opamp that is very good for AVCC is AD797. However, those particular opamps usually work better with no film caps on the +-15v rails. Too much low ESR capacitance on the power rails can cause stability issues in that case, IME.
However, film caps are probably worth a try if using other good audio opamps such as OPA1612 or LME49720/LME49860. The caps tend to help get rid of the last bit of graininess in the sound that may perhaps be most easily noticeable for things such as cymbals.
Also if interested, the AVCC schematic we use can be found attached to post #3003 in this thread.
-Mark
Last edited:
1.65v = AVCC/2 and is the quiescent and average output voltage of the dac. Using it for Vref makes the dac output current = 0ma on average. Whether or not that is optimal in all ways is not something that ESS has said.
Some people may like to adjust the Vref voltage divider so as to make the average I/V output voltage = 0v. In that case, low level audio signals will be operating right in the area where opamp output crossover distortion sits. Maybe not so good for small signals coming out of the I/V stage, but maybe better for the differential summing stage. Distortion in the differential summing stage may be lower if the quiescent or average input common mode offset is = 0v. So, we could adjust Vref for lowest distortion coming out of the differential summing stage.
Another possible source of distortion with bipolar opamps we use for audio, such as OPA1612, is related to input bias currents. The bias currents are large enough that dc input resistance looking out through the opamp inputs combined with the bias current will create a voltage drop, which is an error source. If we can't keep dc resistance really low as seen by both opamp inputs, we should at least make the dc resistance of both of them look the same. That way the bias error is mostly due to the input bias offset current. However, none of the commonly seen Vref voltage dividers makes the noninverting input see the same dc resistance as the inverting input. The voltage divider might have to have a source resistance that looks more like 400-ohms for that. Then again, maybe the error is only a dc offset error, not a dynamic one affecting sound quality. Although there is a graph in the OPA1516 data sheet showing distortion as a function of input resistance, the test circuit they use is not the exact same circuit we are using. If unsure, probably best to run a suite of measurements with an AP to see what changes and to what makes distortion better or worse. Doing a measurement only at 1kHz and one input level is probably not adequate, IMHO.
What it means, I think, is that there are many possible sources of distortion in modern, ultra-low distortion SD dacs. There are no Self or Cordell books on dacs with lists of all sources of known distortion to look out for. People are still finding more everyday, it seems like.
Also, despite the fact that we know about many sources of distortion, such as clock jitter, inappropriate power supplies, improper use of ground planes, etc., lots of people are still going ahead doing wrong things like there won't be any bad effect from it.
What can you do? People behave like people do.
Last edited: