feedback on proposed board layout

Designing hardware that uses VLSI Solution's devices as slave codecs such as an external MP3 decoder chip for a host microcontroller.
Post Reply
danmcb
User
Posts: 7
Joined: Fri 2015-09-04 3:32

feedback on proposed board layout

Post by danmcb »

Hello,

I am in the design process for a Eurocard which will have 4 x VS1053b, each running as slave to its own PIC32MX micro. All the digital section runs from a single 1.288MHz clock. and not much else there apart from regulators and an SD Card. There is also a bit of analogue processing before the VS1053 outputs get to the edge connector.

I have been reading your very helpful notes on board layout, and thinking how to apply them. I made a rough pencil sketch (scan attached) of the layout. I split analogue and digital grounds apart, joining them under each of the four VS's. All power is derived from +24V, so I would also put a bit of filtering on that as it goes into the digital zone.

If you have any feedback on this, or can suggest any improvements, I would be very grateful.

thanks

Daniel
Attachments
layout-sketch.jpg
layout-sketch.jpg (297.48 KiB) Viewed 15204 times
User avatar
Panu
VSDSP Expert
Posts: 2829
Joined: Tue 2010-06-22 13:43

Re: feedback on proposed board layout

Post by Panu »

Hi!

Basically looks ok to me. Be sure to make your powers strong enough so that they are stable. Use lots and lots of capacitors, e.g. large one (electrolytic?) and small one (plastic / ceramic) together. Make also the ground as strong as possible. Place the RCAP capacitors first, very near the VS1053, almost touching the pin. That's the analog reference pin, any noise there will be amplified everywhere in the system. Our RCAP pin is what other manufacturers might call VREF. We have an internal band-gap and with good capacitor and clean ground it's good, but you can also drive that pin from your own reference if you like.

If you have analog inputs, use good quality (film? low ESR?) capacitors as the coupling capacitors and instead of a decoupling capacitor, use a 10nF sampling capacitor in analog inputs.

If you are going for high high quality analog inputs, you can use some extra analog filtering for the ADCs. RC filters are often good enough, but might require an extra digital filter inside the VS1053 to get the frequency response correct. For the absolute best, use active filters. There's a lot of online applets that help you design different kinds of analog filters, I like to use both units of a dual op-amp for filtering one channel. Personally I'm very fond of elliptic filters, because they can remove so much noise because they have the steepest transition band of all filters. But some people feel that their phase response is too unlinear. It's up to you.

For the gap between ground planes, the theory says that if no signals at all in any layers are routed over the gap (there shouldn't be), then you don't need the gap. But you do know what is the difference between theory and practice, don't you? - Theoretically there's no difference, but in practice there's a lot. :)

Finally, if you intend to drive external amplifiers or have a line out, you need an analog filter in the output. In my signature below, there's a link for a "Line in & Line out" schematic. That one has an RC filter (marked 3rd stage sigma-delta reconstruction) which is good enough. But again, you can also use active filtering with op-amps. If you do, place the middle of the transition region somewhere below 50kHz.

Our signals are centered around the RCAP voltage - 1.23 or 1.65 volts. And when the analog drivers are switched on, there's an unavoidable "pop" sound when that happens. You might want to take measures to eliminate that pop from the output - if you use on-board amplifiers, keep them switched off until the VS1053 output is active and stabilized. And use differential amplifiers; connect the first input signal ("IN-") of the differential input pair to LEFT or RIGHT and the second pin ("IN+") to GBUF pin to minimize the pop - GBUF rises up together with the LEFT and RIGHT signals. (See http://www.vlsi.fi/fileadmin/software/V ... der_03.pdf for an example.) If you have a line out, consider to use analog switches to cut the signal and charge the coupling caps slowly instead of the sudden rise from zero to rcap of the VS1053 output. There should be a schematic for that lying around here somewhere.


Well, here were some initial thoughts.. not sure if I got the correct point of view or not for you.. :) Just get your ground right and your powers strong. Separate LDO's (low drop-out regulators) for each voltage. And for the analog signals, basically the "Line In & Line Out" schematic with passive RCs is good enough for Hi-Fi. After all, our crystal locked DACs are high quality and we rely on the purity and mathematical exactness in our signal handling instead of trying to mask out the errors by using, for example, ridiculously high sample rates. So you have a pretty good signal to begin with. But for those who feel that normal Hi-Fi is not good enough, there's a world of adventure and discovery in analog filters.
danmcb
User
Posts: 7
Joined: Fri 2015-09-04 3:32

Re: feedback on proposed board layout

Post by danmcb »

Thanks Panu, that's great info! I will take heed. I already have separate 3V3 supplies. The one on the digital side is a small switcher, the analogue one is linear. They are not LDO though, don't need to be as they run from +24V.

There are no analogue inputs (this application reads from SDCard and plays back the small mp3 files it finds, in a loop). Better to leave analogue ins unconnected or maybe ground them? perhaps via a small resistor?

Yes, the output goes to amplifiers, via some buffering. Because this is speech, I do have high and low pass filters, 1st order, at 100Hz and 10kHz, which is fine for this app. I do have the differential amps in my design, using GBUF.

At the moment my code does a software reset of the VS between each playback, which I found was needed to get rid of a nasty click. Now there is a very slight click, but not serious. There is also analogue muting, so I could also mute the analogue circuit from the PIC between each loop, which should make the transition completely silent. I will look into that.

thanks again!
User avatar
pasi
VLSI Staff
Posts: 2102
Joined: Thu 2010-07-15 16:04

Re: feedback on proposed board layout

Post by pasi »

I hope you meant 12.288MHz :D
Visit https://www.facebook.com/VLSISolution VLSI Solution on Facebook
zzz123
User
Posts: 1
Joined: Mon 2015-11-02 9:50

How could I make the VS1053 separate the stereo to

Post by zzz123 »

How could I make the VS1053 separate the stereo to left and right output,I have found it just mix left and right channel as I play a stereo music which the left and right are different.
User avatar
Henrik
VLSI Staff
Posts: 1294
Joined: Tue 2010-06-22 14:10

Re: feedback on proposed board layout

Post by Henrik »

Hello zzz123,

could you put your question in other words? I don't understand it.

Kind regards,
- Henrk
Good signatures never die. They just fade away.
User avatar
pasi
VLSI Staff
Posts: 2102
Joined: Thu 2010-07-15 16:04

Re: feedback on proposed board layout

Post by pasi »

Check that your board does not have any short-circuit between the channels.
Visit https://www.facebook.com/VLSISolution VLSI Solution on Facebook
danmcb
User
Posts: 7
Joined: Fri 2015-09-04 3:32

Re: feedback on proposed board layout

Post by danmcb »

Thanks Panu for the advice. It was very useful. The design came out very well. This was a small run custom design, only two boards have been made, so it was important to get good boards first pass, if possible. In fact that is how it worked out. Here is a pic of the finished PCB.

I took a lot of trouble over power layout - it's a 4 layer board with separate analogue and digital 3V3, and a lot of planes for ground, also one for 3V3 digital. Plenty of decoupling caps as well.

The only thing I wish is that this IC was available in a larger pitch format - when you are doing hand assembly it is a bit nerve racking putting four of these on a board, as any damage to even one pad and there is probably way to fix it. So I really took my time soldering the VS1053 IC's. The PIC microcontrollers were much easier because of the larger format. But anyhow, it all works perfectly.

thanks again for the help.
Attachments
IdentCard1.jpg
IdentCard1.jpg (613.58 KiB) Viewed 14854 times
User avatar
Panu
VSDSP Expert
Posts: 2829
Joined: Tue 2010-06-22 13:43

Re: feedback on proposed board layout

Post by Panu »

Congratulations, looks great!

-Panu
AsaWyon
User
Posts: 1
Joined: Wed 2016-04-13 15:56

Re: feedback on proposed board layout

Post by AsaWyon »

Hi...as per my experience,if you are going for high quality analog inputs, you can use some extra analog filtering for the ADCs. RC filters are often good enough, but might require an extra digital filter inside the VS1053 to get the frequency response correct.
For the absolute best, use active filters. There's a lot of online applets that help you design different kinds of analog filters.
Post Reply