EMU0204 mods for FFT measurements

I've done some simple mods to the EMU 0204 input stage which has resulted in both improved dynamic range and lower measured noise floor.

I will post schematics and more on this later today if there is interest.

For now here are some results.


  • emuspec 10.JPG
    emuspec 10.JPG
    134.2 KB · Views: 1,417
  • emuspec 110.JPG
    emuspec 110.JPG
    132.3 KB · Views: 1,369
I'm just home for lunch and don't have a lot time right now so I'll just go over the idea of what led me to this..

The EMU 0204 is set up for musical application is is not necessarily scaled well for what we use them for. The first thing I did was to trace out a schematic for the ADC input. Compared this to what AKM suggested on their data sheet for an input circuit. AKM naturally used the best configuration to meet the data sheet specs for the ADC.

Looking at EMU's input circuit I noticed that the full range of the ADC is not being used efficiently for our purpose. What is better for us is to use the full scale of the ADC for our measurement range. For me that is 0dB (1Vrms) reference.

I set out to rescale the input so that 0dB is just slightly below clipping of the ADC.
Due to an inability to measure resistors in-circuit my gain is a bit out and I ended up with clipping at about -6dB. I'll fix this later.

The other problem is with the gain control set at minimum, 0dB on the gain knob scale, only half of the differential voltage appeares at the diff inputs of the ADC. This can't realize the FS of the ADC. I also noticed that increasing the gain control increased the noise floor of the ADC proportionally. I also sighted that for best noise performance the input source R needs to be reduced.

All of this was accomplished by the addition of two resistors and minor rearranging of the
input buffer op amps. The op amps used are the JRC2068 (NJR2068). With this performance I see no need to replace these.

After rescaling the input to the ADC ARTA must be re-calibrated to reflect the new scaling.
The calibration need not be done exactly at 0dB. -10dB can be used just as well. Simply tell ARTA what level is being input.

The result was far better than what I had expected. I think once I rescale again to 0dB the noise floor will rise a little. However one can scale to what ever range is used. A 500mV scale could be used and the entire ADC range applied to this.

We can do this because of we use fundamental suppression, THD is a relative measure and the software allows scaling down to 1mV.

Last edited:
Sunstantial development! Everyone on the planet will benefit from this spin-off project.
can such be applied to other ADC? Such as, the QA400?


Hi RIck,

The QA400 is about 3dB below FS as it is. I don't think there is much that can be done with it. The input impedance can be brought down by replacing the 100k resistors with lower value
R's on the VCOM bias circuit to the buffer amps. I did observe lower noise from doing this but not that much.


These are very nice results. Thanks for starting a separate thread on this. I look forward to seeing the details. I have a [email protected] that I really like, and will be interested to see if these mods may be applicable in some way.


Hi Bob,

I think these mods should be applicable to any sound card. I don't have a [email protected] or 0404 to play with but if you look at the chip set you can see what ADC or codec ESI used then get the data sheet for it.

The manufactures of the cards stick very closely to the IC's data sheet recommended input. If a high input Z is not needed then the noise can be brought down by lowering the input Z. But think the key here is to utilize the full input range.


Reason I asked was that there are forum posts (not in DIYAudio) indicating that E-MU driver for WIN 7 (what I'll be using) are not the best but the 0404/0204 both look ideal from a performance perspective. I haven't selected a soundcard yet.

Hi Matt,

I've heard that too. There is problem with the XP driver at 192ksps as well. I get some jumping of the noise floor every now and then. This same problem came up with beta version of SW for the QA400. It turned out to be a timing problem with one of the XP API. IIRC, QuantAsylum fixed the problem by increasing the delay time for sending the data packets.

I would think Creative would give the win7 issues more attention since the 0204 is still a current product. XP has it own issues and the reason for ASIO.

Last edited:

Looking at EMU's input circuit I noticed that the full range of the ADC is not being used efficiently for our purpose. What is better for us is to use the full scale of the ADC for our measurement range. For me that is 0dB (1Vrms) reference.



What is the original full range? The spec says 6.5 dBv maximum input. Is this the original full scale?
Last edited: