I took a look. I still choose not to follow those recommendations in this particular instance.
For example on 1. I choose not to use an RC filter for supply purity, rather I use an LC filter (a ferrite bead provides the L). I consider LC to give superior performance here - it negates the need for so many supply decouplers. Of course if a series R is used for filtering the need for decoupling becomes much more critical.
On 2 I am not using a ground plane nor am I splitting return current paths.
On 3 I use two decoupling caps for 6 chips. This is based on experience with my multiple paralleled DAC (GrossDAC) where I used a total of 12 ceramics for 36 chips. That board though did have a groundfill. I choose not to shield the digital input lines with ground traces.
You have lost me - I am not seeing any voltage compliance spec apart from at DC for the TDA1387. The TDA1543 has an AC voltage compliance spec and I'll just mention in passing every one of the passive I/V implementations I've seen violate it.
Why is this relevant? I'm 'respecting' the DC voltage compliance spec in this design. If I violate that the DAC won't put out the stated current so it makes a large difference to the operation of the DAC. Unlike the case of changing the supply decoupling details.
The application information isn't the same as the limits in the DS table so I'm not sure of your point.
You mentioned 'a bit disappointed' that's what I was referring to. Why mention that if you didn't want a comment on it?
Yes, its the way used on the Audiophonics board.
I agree that's the textbook way to resync an async signal. But it does require knowledge of the metastability resolution time of the flip flop, not a parameter that is given in the DS.
I made an omission when talking about asynchronous reclocking - the metastability isn't the most serious issue. I forgot to point out that the peak to peak jitter when an audio clock is resynchronized with an asynchronous 50MHz clock becomes 20nS. Not at all an amount to be proud of when typical S/PDIF receivers achieve of the order of 0.1nS.
Maybe because you are avoiding something critical in digital circuit design without some sensible reason? Isn't this a valid reason for disappointment? Sorry but I can't follow you...
Last edited:
You didn't follow my explanation then?
Disappointment is the result of having unreasonable expectation. See : Dealing With Disappointment | Psychology Today
Prototype 9th order filter
My wife managed to find a way to squeeze 8 inductors onto a board outline designed for 6
The stop band rejection for this filter is in theory better than 100dB but in reality this number is going to depend on circuit parasitics which I'm not able to quantify accurately. Listening will reveal whether there's any benefit of this compared to the 7th order filter. If anyone's interested in the FR plot (simulated) just ask and I'll post it up.
My wife managed to find a way to squeeze 8 inductors onto a board outline designed for 6
The stop band rejection for this filter is in theory better than 100dB but in reality this number is going to depend on circuit parasitics which I'm not able to quantify accurately. Listening will reveal whether there's any benefit of this compared to the 7th order filter. If anyone's interested in the FR plot (simulated) just ask and I'll post it up.Attachments
You mean this?
abraxalito said:
I don't see anything that can be qualified as an explanation in technical aspect here. If you can, I think that this is the main problem with the whole discussion on this subject. You criticize me with being "subjective" but you come up with an entirely subjective explanation like "I did this because I did same on this..". This is funny.
Last edited:
No, that's not the main part of it, this is :
For example on 1. I choose not to use an RC filter for supply purity, rather I use an LC filter (a ferrite bead provides the L). I consider LC to give superior performance here - it negates the need for so many supply decouplers. Of course if a series R is used for filtering the need for decoupling becomes much more critical.
I see I missed out an explicit mention that the configuration worked fine on the multiple parallel DACs board so I transferred it to the PhiDAC hex. Any clearer now?
You reasoning is entirely subjective here. Your choice of using LC filter on PSU part doesn't mean that you can avoid local decoupling.
Local decoupling is mandatory for each chip in order to ensuring low-inductance path to ground. This is nothing to do with PSU schema (LC or RC) of choice. That's why this is mentioned on different section (3. Topology).
This doesn't mean that following DS directions will not work fine or even better. Also, this doesn't mean that you should unconditionally follow DS directions either. This is your design and you are free to whatever you like.
Unfortunately not. After seeing your efforts on defending your decision of avoiding local decoupling, I guess your belief on your scheme is stronger than I thought. Otherwise you could easily spend your effort on optimizing decoupling scheme instead of this discussion.
Let's swap roles here for a moment. You give us your interpretation of the application's paragraph #1. In particular the values of R and C in the RC filter its suggesting to be used to purify the supply. Then we can move forward from there.
I don't know what you are up to. It seems you are loosing your calm with temper. Paragraph #1 is about purifying supply voltage. If you did it, then put a check mark on it then go next.
Btw, who is "us" ? Is there anyone else cooperating with you in order to excel the design?
Correct, and you don't need to know.
Then it looks as though you're projecting your own anger on to me.
I'm asking for you to demonstrate how to implement the application's paragraph #1. Is that really so difficult to understand?
Here 'us' means me and any readers of the thread.
Attachment 1, 2: This is what came out of this test at 1kHz test frequency.
I was thinking that the extra gap due to mylar foil might change from it’s initial value over time but this does not reflect on L and Q values which remained quite stable over a week.
The sharp drop on Q on the 1st coil is due to breaking of two of the 8 0.007mm enamel wires The coil ends were on the air during measuring period. At the last day after the 6th measurement, I properly soldered the leads to the bobbin tags and L, Q values changed slightly (last measuring point on each diagram).
Attachments 3, 4, 5 are impedance sweep results (0-46kHz) on the coils with my E-MU 0404 sound card. Signal level:1Vrms
The faulty coil is to be replaced by another one (the 7th coil at the sweeps).
George
Attachments
Decoupling experiments
As a result of the decoupling discussion my wife got curious and decided to modify her PhiDAC hex by soldering up 5 extra 1uF 0805 X7R caps across the DAC chips, between pins 4 and 5.
The result was quite a surprise to me as I'd expected no difference would be easily heard. Instead we heard a substantially less musical result - we were playing a Beethoven string quartet at the time which I wasn't familiar with. I asked my wife 'Who is this composer?' and was shocked to hear 'Beethoven' as a reply as he's one of my all-time favourites. But the presentation was really rather a turn-off to listen to - strings had lost their usual delicacy.
So then I thought - maybe this isn't about caps per se, its about X7R caps. So next up we changed the X7Rs to Wima 330nF film caps (the kind with 5mm pitch). The jury is still out on how this sounds but one thing we're both clear on - its not an improvement.
As a result of the decoupling discussion my wife got curious and decided to modify her PhiDAC hex by soldering up 5 extra 1uF 0805 X7R caps across the DAC chips, between pins 4 and 5.
The result was quite a surprise to me as I'd expected no difference would be easily heard. Instead we heard a substantially less musical result - we were playing a Beethoven string quartet at the time which I wasn't familiar with. I asked my wife 'Who is this composer?' and was shocked to hear 'Beethoven' as a reply as he's one of my all-time favourites. But the presentation was really rather a turn-off to listen to - strings had lost their usual delicacy.
So then I thought - maybe this isn't about caps per se, its about X7R caps. So next up we changed the X7Rs to Wima 330nF film caps (the kind with 5mm pitch). The jury is still out on how this sounds but one thing we're both clear on - its not an improvement.
I wouldn't have a clue what to measure. I've already measured the DAC's distortion (about what the DAC chip's DS leads me to expect) but so far not the amp's. But I doubt the amp has anything to do with it as my wife got similar results when listening with headphones (different amp involved).
Looks like a textbook case of confirmation bias
I still have a class II ceramic in circuit.
ABX with what as A and what as B?
Incidentally hadn't you said many months ago that you were staying away from this thread in future? What changed your mind?