Connect with us

USB Sound Cards? Good to adapt for stuff?

Discussion in 'Electronic Components' started by Lostgallifreyan, Jun 13, 2006.

Scroll to continue with content
  1. Hi.
    I just bought this thinger:
    http://tinyurl.com/zloek
    (USB full duplex audio 'card' probably based on Micronas IC)

    Has anyone here played around with one? I was thinking that with a copy of
    Sound Forge or similar app, a very very cheap realtime datalogger can be
    made. The main sticking point would be DC blocking, which would have to be
    removed.

    Any thoughts?
     
  2. Pooh Bear

    Pooh Bear Guest

    I think most A/Ds have high pass filtering in DSP. It may simply not work.

    Graham
     
  3. Which would be annoying. I guess someone might have made a DirectShow
    plugin that can full-wave-rectify a signal, that way I could get a unipolar
    trace by filtering the rectified input from a 20 KHz oscillator whose
    output amplitude is controlled by the signal I want to log. That defeats
    the point though, I'm hoping that someone made a sound card that leaves
    nothing more than removal of DC blocking caps as the only thing to do to
    get a data logging input.

    Incidentally, I know that a bit of code in a hardware DSP might come cheap,
    but cheaper than providing and mounting two 10 µF caps? Surely not? >:)

    I can see why it might be done to save space in a widget like the thing I
    linked to, but as a standard practise, I'd have thought it was not done. If
    is, I guess it might explain why it's impossible to modify an Echo Layla 20
    bit interface to take DC inputs. In that, there certainly is something
    blocking DC on the ADC itself.
     
  4. Dave Platt

    Dave Platt Guest

    Another issue is power. If I recall properly, power-over-USB provides
    only a single supply. If you want to measure DC voltages which swing
    below ground, you're going to either need some way of developing a
    secondary, negative-rail voltage supply, or need an external bias
    network (possibly with a buffer) to shift the voltage up into the
    common-mode range of the single-power-supply USB chip.

    Standard USB-for-audio-purposes adapters probably do the latter...
    capacitor-couple the signal and then use a resistive ladder to shift
    the centerpoint up to .5 Vcc.

    That's probably cheaper than having to provide a second power supply
    of some sort to the USB chip, so that its common-mode input range
    extends well below ground.
     
  5. Stef Mientki

    Stef Mientki Guest

    isn't this a audio output only device ;-)

    Stef Mientki
     
  6. Joerg

    Joerg Guest

    The DC block for sound cards might be on-chip and then you are pretty
    much stuck. A better option could be one of those TI EX430 kits. They
    cost $20 and contain an MSP430F2013 which has a 16-bit ADC on board. No
    DC blocking plus it's freely programmable.
     
  7. Thankyou. That also answers my unanswered question in another thread, for
    suggestions for a specific device with ADC and some space for code. You
    wouldn't know one that also has 16 bit DAC in addition to that lot, would
    you? I'd want one for making a log convertor for analog voltages, with a
    gadget that makes the simplest and cheapest board layout possible.
     
  8. Nope :) Definitely input, one of those sockets is a stereo mic input. This
    thing might not be suitable, as people have said, but that would be because
    it's stuffed full of DSP I don't want and maybe can't bypass (although if
    it IS the Micronas chip, it is apparently user controllable by some means I
    don't know yet). I just want a cheap way to get 16 bit voltage conversion
    in and out, and the USB idea is appealing. If I can get this in something
    cheap and mass-produced, I will, there's no point in re-inventing the wheel
    at inordinate expense..
     
  9. Joerg

    Joerg Guest

    Lostgallifreyan wrote:

    Not from the MSP430 series. Atmel or Microchip might have something but
    I am not familiar with their devices, except for some 8051 family parts.
     
  10. Stef Mientki

    Stef Mientki Guest

    Texas has some of these chips,
    if I remember well 24 bit ADC + ucontroler (8051?) + 20 bit dac in 1
    device (maybe even a few ADC/DACs),
    I just found out after making a log converter with separate components ;-)

    Stef
     
  11. Joerg

    Joerg Guest

    There may be a quick and dirty option: Find out the highest frequency it
    can process reliably and with enough dynamic range. Then chop the slow
    input signal at that rate. A simple CMOS oscillator and a gate or FET
    can do that. You may lose some bits but hey, if it's good enough it
    might work.
     
  12. Nice. :) I think that might just do it.

    One other thing I'm thinking is that the HPF on an IC might be disabled, if
    the IC designer was generous about this.

    The Echo Layla 20 bit audio rack unit is a VERY tempting device for laser
    scanner control and other things, monitoring laser power, general lab
    monitor and control. Echo Audio kindly sent me the I/O datasheet in PDF.
    I've found that the ADC is a Cirrus Logic CS5335-KSEP and with a bit of
    fine surgery pin 1 can be made to control DC offset, and a DC blocking cap
    can be bypassed. Currently, if I try to bypass the DC block cap, the result
    is white noise in that channel, I have NO idea why, it really is odd, that
    one..

    If I can solve this, the CS5335 pin 1 will be a nice control, when low, it
    tracks an onboard op-amp's DC output and cancels it, when high, it makes
    the offset compensation freeze, so I can short an input, hold low for a
    second, then raise high to eliminate any offset between the short and the
    DAC input inside the CS5335, regardless of source, an extremely useful
    feature. I hope the Micronas or similar IC in those cheap USB thingers can
    be modified for this, but I suspect they might not provide a means to
    bypass the HPF, even though they do apparently offer some control.

    All this sounds tedious, but it beats having to build from scratch every
    time, especially when 8 ins, 10 outs, at 20 bits or better (plus 2-channel
    S/PDIF I/O at 24 bit), can frequently be found on eBay for £70 or so.
    Beside the prospect of adapting that, all ideas of self-build look as
    appetising as a pair of used boots on a plate.

    Your fast-chopping idea is nice though, if it works, it won't matter what
    the input processes are, at 20 bits I might still get to keep an accurate
    16 bit log. These digital multi-channel things put out sync signals too, so
    I guess if I can derive something from that I can reduce bit-loss by making
    the chopper sample-accurate.
     
  13. That one sounds cool. Anything that takes the pain away... :) As I
    mentioned in the big post a few minutes ago, it's usually far more
    attractive to adapt existing commercial devices, given that there are so
    many cheap ones, and surely some might be spectacularly useful if this
    works.

    I've looked at log convertors with analog parts, and decided not to try it
    that way, though one analog log amp still IC interests me, the AD8307. An
    8-pin DIL. The datasheet talks high, wide and plentiful about RF though, I
    don't know enough to know if it could do DC for 3 octaves or so of accurate
    freq-pitch scale conversion, and no-one's been able to tell me yet so this
    one is still a mystery to me.
     
  14. Joerg

    Joerg Guest

    Lostgallifreyan wrote:


    My ISP's news server is dropping posts, some of which later pop up (like
    yours right now). Would have answered sooner but couldn't see it.
    Basically this is modulating your input signal onto a carrier, the
    chopping frequency being the carrier.

    They usually aren't. They are after that one big market of audio capture
    and anything else is peanuts to them from a business point of view.

    Look at the DC level on the other side of the cap with a scope. If it
    isn't zero at all times (doubt that it is) then a bypass would force it
    down and one or more bias levels could go out of whack. Once they are it
    can take a longer time than expected to return to normal.

    True. If you want to cheat and can spring about $400 then this one might
    be a nice option:

    http://www.labjack.com/labjack_ue9.html

    You can also synchronize with a nifty software PLL that "learns" when to
    expect the sample. But that can run astray if the signal is too low for
    a long time. As with building your own hardware that'll be a hassle you
    just might not need right now.
     
  15. Sorry about delay. :) That was a good post, so I wanted to return some effort.
    http://tinyurl.com/mt97e is a direct link to a PNG file where I made a
    diagram of the part of the circuit I need to change, including the simplest
    modification I could make. It removes just one DC blocking capacitor in each
    channel and uses three wires to connect 4 resistors and one op-amp stage to
    mix the 2.2V reference to the incoming DC-coupled signal, feeding the mix
    directly to the 2.2VDC side of the now-absent capacitor.

    I might also try the chopper idea, as the 'word clock' output is regular
    and always present. I wouldn't try to decode it in software if it's regular
    enough, I'd just filter it to get a simple regular pattern and lock an analog
    PLL to it if possible.

    Please let me know if the proposed modification in the PNG file looks ok, or
    might be bettered (bearing in mind the need to minimise to extreme any mods
    made on the original board).

    As mentioned in the PNG file's text, the ADC does allow direct coupling by
    bypassing the DSP HPF it uses to block DC. Actually it either monitors it
    and perpetually readjusts the output to compensate, or it freezes the offset
    at the last known value. I mentioned that in some more detail in my previous
    post in this subthread:
     
  16. Joerg

    Joerg Guest

    Lostgallifreyan wrote:


    Well, it's my news server that does that sometimes.

    It was a bit hard to see, maybe next time you could post a zoomed out
    schematic (browsers are often bad at zooming). And add designators (R1,
    R2...). :)

    I'd drop the 10M input dividers to 1M, still shouldn't mess up your
    bias. A concern: This opamp may not like it if you feed it just 5V
    single supply. Also, IIRC it's input common mode range must stay about
    1V away from its negative rail and if you grounded the neg supply this
    is cutting it really close. You might want to make that new input
    divider asymmetrical to push this a little higher.
     

  17. I should have mentioned, the main circuit's op-amps are on a dual rail 15V
    supply.
    Sorry, that was definitely a gross oversight of mine, not showing that. About
    the
    resistors, I agree, I'd considered that too. I often reduce to 1M in situations
    where
    it first occurs to try the highest I have, things are usually more resilient to
    induced
    noise that way, and any error is more predictable. I did try shrinking the
    image but
    it made a pig's ear of the text, probably another dodgy decision to put it
    there, but nm..
    My main concern is whether there might be some horrible gotcha to look out for
    when
    using the non-inverting summing amp in this way. Also, the possiblity that some
    vastly
    superior answer might be staring me in the face while I can't see it. :)
     
  18. Joerg

    Joerg Guest

    Lostgallifreyan wrote:


    Watch out for power sequencing. If your 15V supplies come on before that
    of the CS5335 or if your opamp bias points aren't immediately stable
    when those 15V rails come up the top right 150ohm resistor might source
    up to 100mA into the AINL+ input. This could cause it to latch up or
    fry. I don't know how much it could stomach, might be worth to think
    about raising that resistors value.
     
  19. I'll trust that one :) That's a network the unit's maker applied right from the
    IC maker's datasheet, two 150R resistors and a 2n2 cap. I'll tap the same
    supply the main op-amp's getting, for my addition (though I'll maybe modify the
    PSU slightly to carry the extra power, as some bits already get hot). I think
    it will be ok, as the 2.2 V won't be doing anything it didn't already do
    before, as the CS5335 sees it. I'm more cautious about that reference itself, I
    think the CSS5335 sources it at a very low current, hence my initial choice of
    10M for resistances on the adder input. I'm thinking of using the LF412, one
    for each pair of channels, four IC's in all.
     
Ask a Question
Want to reply to this thread or ask your own question?
You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.
Electronics Point Logo
Continue to site
Quote of the day

-