Hi Mark,
Good decision to go back to Marcel’s clock doubler and RTZ.
I suppose you still have the Rubicon caps.
The best of both worlds.
Are you still going to try the 2x8 instead if 4x4 shiftregisters ?
Hans
Good decision to go back to Marcel’s clock doubler and RTZ.
I suppose you still have the Rubicon caps.
The best of both worlds.
Are you still going to try the 2x8 instead if 4x4 shiftregisters ?
Hans
Are you saying that the 9039q2m does not support mono output to both L/R outputs? AFAICT this can still be acconmplished by setting registers 64/65 appropriately.
Like to try that at some point. Would like to figure about the resistors first.
True. There is some other stuff I also need to do that might take a few days. Don't want to put off the other project too long. It will involve taking down the dacs.
That is not the problem. Summing currents by shorting outputs together results in high THD regardless of input level. My guess is that it is the same issue as IVX pointed out with ES9039PRO.
It sounds to me like you have this discussion in the wrong thread.
Talked to Acko about metal foil resistors. He may be able to have some sent directly here to try out in the dac. Have to wait and see how that works out.
Thanks Mark, I am fine with this mod/upgrade using foil resistors which I intended to do anyway so don’t worry about reverting later on🙂
Thanks bohrok2610, I am happy with what you have presented and your plans going forward. All good references for future evaluations…
@Endo2112 you are continually pushing your own agenda here As such you will find you are unable to post further in this thread, a thread which has nothing to do with your beloved ESS ES9039Q2M
Concerns have been voiced over your identity and motives from the outset. If this continues your membership will be terminated.
Jinkies! I just looked up the prices of foil 0805 resistors at Mouser. 😱
It's like soldering 20 dollar bills to the board! 😉 Serious ouch if one of those pings off and away during soldering!
I am wondering though if swapping out the Susumo 0.5% thin films for something with a better tolerance would represent a worthwhile improvement? Say Vishay/Dale 0.1%? They can be had for around £20 for the lot and maybe that is a cost effective way to gain some higher accuracy? Or will that extra accuracy not be noticed due to the nature of the circuit?
I bought 100x 1% MMA Vishay/Beyschlag MELF0204 resistors for 5 EUR and hand matched them with a quick-and-dirty Wheatstone bridge to 2x 32 resistors with 0.1% tolerance (or better).
A single-bit FIRDAC is not nearly as sensitive to resistance tolerances as a multibit DAC, but tighter tolerances won't do any harm.
The main effect of a tighter resistance tolerance is that the notch in the response of the FIRDAC gets deeper, giving a better suppression of the idle tones around fs/2 (which could mix into the audio band when there is a spur on the reference or the clock around the same frequency) - if there are any such idle tones, which depends on the modulator algorithm. A secondary effect is less sensitivity to bit clock duty cycle fluctuations.
Obviously I don't expect any major improvements. If I did, I would have used 0.1 % tolerance resistors myself.
Some info on metal foil resistors attached.
An Excerpt from one document:
Also some info on current noise in metal film resistors.
An Excerpt from one document:
Also some info on current noise in metal film resistors.
Last edited:
Mark,
I know you're not the op-amps best friend, but when it comes to noise reduction with metal foils resistors, there is another way to achieve that, right in front of you.
When steering the Dac with a 0dB signal, you will get 66mVrms on the 8.2nF cap.
The 8 resistors, in parallel are producing ca. 339nV/rtHz over a 20Khz BW, which is slightly less as with a straight FR.
This calculates in a S/N of 20*log(339e-9/66e-3) = 105.8dB.
Now instead of tapping the signal at the 8.2nF cap but on the output of Marcel's first op-amp, output voltage will be 910mVrms at 0dB digital in.
With the OPA2210 noise at the output will be 1.37uV over a 20Khz BW.
S/N now becomes 20*log(1.37e-6/910e-3) = 116.4dB.
So, you would have gained 10.6 dB in S/N, which is much more than could be achieved with metal foils.
How's that possible almost like magic, well that's because the junction of 255R and 845R is a virtual input, so this low input impedance is in parallel to the eight Firdac resistors effectively lowering their voltage noise.
Hans
I know you're not the op-amps best friend, but when it comes to noise reduction with metal foils resistors, there is another way to achieve that, right in front of you.
When steering the Dac with a 0dB signal, you will get 66mVrms on the 8.2nF cap.
The 8 resistors, in parallel are producing ca. 339nV/rtHz over a 20Khz BW, which is slightly less as with a straight FR.
This calculates in a S/N of 20*log(339e-9/66e-3) = 105.8dB.
Now instead of tapping the signal at the 8.2nF cap but on the output of Marcel's first op-amp, output voltage will be 910mVrms at 0dB digital in.
With the OPA2210 noise at the output will be 1.37uV over a 20Khz BW.
S/N now becomes 20*log(1.37e-6/910e-3) = 116.4dB.
So, you would have gained 10.6 dB in S/N, which is much more than could be achieved with metal foils.
How's that possible almost like magic, well that's because the junction of 255R and 845R is a virtual input, so this low input impedance is in parallel to the eight Firdac resistors effectively lowering their voltage noise.
Hans
FYI, Texas Components has a better price for Vishay metal film resistors.
A few hundred dollars in resistors per board if you get to the 15% discount level.
A few hundred dollars in resistors per board if you get to the 15% discount level.
Update: Pro hi end audio designer guy dropped by and listened to Marcel's dac with the RTZ circuit restored. He said, the space is back but the male voice doesn't sound right, its too weak in the midrange. He said, put the X5R caps back in. So I did. He then said it sounds way more open this way, much better. He said the voice is better but still isn't quite right so you have to work on that. Then he left.