I don't know your design or the AK4497SVQ well enough to give you much of an answer. Of course the LT3042 has a pretty good ripple suppression at 1.2 MHz and LC filters can also be effective at such frequencies, but the reference voltage has to be very clean indeed.
Regarding SMD ferrite beads, the Murata BLM18RK series (such as the BLM18RK102SN1D) may be of interest, as they have more impedance at 1.2 MHz than most SMD ferrite beads.
On the PCB, you will have to keep the loop areas of the loop where the big 1.2 MHz currents flow and of the loop from the reference decoupling to the AK4497SVQ and back small and make sure they are at some distance from each other to minimize magnetic coupling.
Me too.
Because XMOS (and CM66xx) — only with them I can write my own firmware as I need. Number of channels, control of DAC, ADC and other devices, and implement various functions.
Xing, Amanero, etc. do not provide such an opportunity.
This is what I was talking about - to do better.
Alex.
Not me. With STM32 I can do all above and more and in addition publish my source code in entirety.
It's almost the same thing, because a device that suits you is better than one that doesn't suit you, isn't it? 🙂
I know.
Forget to mention also STM32 and PIC32.
Alex.
Last edited:
The lesson of this story would seem to be that unless you are willing to give up control of the direction of the discussion, don't.
I forgot the elephant in the room: and use magnetically shielded inductors.
talk about VREF.
IMHO, I think AKM's DAC is similar to SAR ADC's VREF.
The current is also not large, so the key is not 2200uF C, but
keeping the output impedance low outside MCLK.
Anyway, I simulated three ways.
1. Reference voltage IC and ADR4805
(Add RC filter to the output of reference voltage to reduce noise,
and use opamp as voltage buffer.
ADA4805 can drive capacitive load,
and is recommended opamp for similar applications.)
2. LT3042.
3. LT3040 (Reference voltage IC Buffer based on LT304x series regulator topology.)
Conclusion:
1. Too many disadvantages. Since it is an opamp, it has low impedance, but it is not the best,
and it gradually increases as you go to high frequency, and the impedance peaks at 6 ohms around 100 kHz.
And the PSRR is low around 1.2 MHz, so if you have to double regulate it, it's kind of funny.
It also takes up quite a lot of space.
2. Very good. Very low across the entire band.
3. Very good. Slightly better than the LT3042.
Also, it should be noted that the LT3042 and LT3040 have in common
that they require some output current. VDDL/R satisfies this condition,
but VREFHL/R may require additional load.
In conclusion, to control the impedance of the LT304x and the high frequency band,
it seems that paralleling a Tantalum 10uF with low ESR and ESL and a MLCC with lower values will probably give good results.
The ceramic dielectric may be important, but many builders have reported good measurements with the LT3042,
so there's probably no need to sweat it too much.
IMHO, I think AKM's DAC is similar to SAR ADC's VREF.
The current is also not large, so the key is not 2200uF C, but
keeping the output impedance low outside MCLK.
Anyway, I simulated three ways.
1. Reference voltage IC and ADR4805
(Add RC filter to the output of reference voltage to reduce noise,
and use opamp as voltage buffer.
ADA4805 can drive capacitive load,
and is recommended opamp for similar applications.)
2. LT3042.
3. LT3040 (Reference voltage IC Buffer based on LT304x series regulator topology.)
Conclusion:
1. Too many disadvantages. Since it is an opamp, it has low impedance, but it is not the best,
and it gradually increases as you go to high frequency, and the impedance peaks at 6 ohms around 100 kHz.
And the PSRR is low around 1.2 MHz, so if you have to double regulate it, it's kind of funny.
It also takes up quite a lot of space.
2. Very good. Very low across the entire band.
3. Very good. Slightly better than the LT3042.
Also, it should be noted that the LT3042 and LT3040 have in common
that they require some output current. VDDL/R satisfies this condition,
but VREFHL/R may require additional load.
In conclusion, to control the impedance of the LT304x and the high frequency band,
it seems that paralleling a Tantalum 10uF with low ESR and ESL and a MLCC with lower values will probably give good results.
The ceramic dielectric may be important, but many builders have reported good measurements with the LT3042,
so there's probably no need to sweat it too much.
Are you familiar with this thread?
In it, @bohrok2610 found measurably worsened amplitude noise sidebands (the thread title says phase noise, but it turned out to be amplitude noise) due to the microphony of the class 2 ceramic reference filtering capacitor of an LT3042 used to supply the reference voltage for a DAC. He did not find a measurable effect of the microphony of the output decoupling of the LT3042.
There is some other thread where @Markw4 dug up articles about 1/f noise in electrolytic capacitors, both aluminium and tantalum types. There was some 1/f noise associated with their leakage, so you can best use low-leakage types for reference filtering (and low ESR, of course).
It's all a matter of perfectionism, of course, as these sidebands should be very well masked anyway.
In it, @bohrok2610 found measurably worsened amplitude noise sidebands (the thread title says phase noise, but it turned out to be amplitude noise) due to the microphony of the class 2 ceramic reference filtering capacitor of an LT3042 used to supply the reference voltage for a DAC. He did not find a measurable effect of the microphony of the output decoupling of the LT3042.
There is some other thread where @Markw4 dug up articles about 1/f noise in electrolytic capacitors, both aluminium and tantalum types. There was some 1/f noise associated with their leakage, so you can best use low-leakage types for reference filtering (and low ESR, of course).
It's all a matter of perfectionism, of course, as these sidebands should be very well masked anyway.
And here's some interesting stuff I found on the AKM site today.
It turns out that AKM also uses all three methods to supply VREF to their 32bit DACs.
It turns out that AKM also uses all three methods to supply VREF to their 32bit DACs.
@MarcelvdG Thank you. Very realistic and practical information. Also, can easily find related papers.
Not according to datasheet. Most performance characteristics are better at 1mA load.
@bohrok2610 Yes. I was just talking about output impedance.
As you said, it starts working well from very little, around 500uA.
I haven't built AKM, so I don't know the minimum IREF.
Anyway, as you say, don't need the extra load.
As you said, it starts working well from very little, around 500uA.
I haven't built AKM, so I don't know the minimum IREF.
Anyway, as you say, don't need the extra load.
@bohrok2610
Just checked the related post. It's awesome.
I took the STM32 cube ide class a year ago and it was just implementing simple functions.
People like you keep my DIY passion alive.
It's time to unpack the reflow machine I bought for my DAC build.
As Altor said, this is a perfect example of "to do better".
STM32CumeMX does not provide USB UAC2 functionality at all.
Even UAC1 is very, very limited there.
🙂
This is AK4497 DAC schematic (partially), which I used in the past.
The measurements were very good and the SQ was so high that I didn't even turn towards the ESS (despite several attempts) until the Big Fire happened.
Alex.