I think this may be what I was suggesting before. The distortion changes with software set output level from the D10s.
The 1K resistor may also be making the load impedance more friendly for the D10s output amplifiers.
Certainly your test shows that there's different results from the DAC when there's no changes to the ADC side.
You are wrong. Capacitor size matters. 220u and 320R (1.7Vrms range) is not enough for LF, that is all.
I changed 220uF/10V to 470uF/10V (I used damn jamicon, because I don't have anything else at home)
Now I have much better numbers at LF. 1000uF will be the final.
stock caps and bypassed
I changed 220uF/10V to 470uF/10V (I used damn jamicon, because I don't have anything else at home)
Now I have much better numbers at LF. 1000uF will be the final.
stock caps and bypassed
Last edited:
I measured with ADC's sensitivity 1.7V, and my AP SYS2522 was the sine source. I guess your ADC has a defective one of caps. Also, if 1000ohm in serial helps with H3, you have a problem with your sine source.
Did you measure in unbalanced mode ?
The stock caps I have are 10V not 16V
There is nothing wrong with my DAC Topping D10s. I found an older measurement of D10s -> Asus STX II PCIe. THD 20Hz is -120dB
The stock caps I have are 10V not 16V
There is nothing wrong with my DAC Topping D10s. I found an older measurement of D10s -> Asus STX II PCIe. THD 20Hz is -120dB
Yep, balanced or unbalanced the H3 of 20Hz is the same level -112db. Did you measure your DAC at the same input impedance yet? I guess not 😉
New measurements. D10s -> STXII
DAC Topping D10s, output impedance = 20R, 0dBFS = 2Vrms
ADC STXII, input impedance = 4K3, 0dBFS = 2Vrms
Dashed traces = additional 500R load on the DAC output.
All 20-50Hz THDs are better than Cosmos with 220uF input caps
DAC Topping D10s, output impedance = 20R, 0dBFS = 2Vrms
ADC STXII, input impedance = 4K3, 0dBFS = 2Vrms
Dashed traces = additional 500R load on the DAC output.
All 20-50Hz THDs are better than Cosmos with 220uF input caps
Cosmos ADCiso grade A 1.7V sensitivity, the input is 600mV=-10dbfs, with an original alcap 220uF.
20Hz H3 is about two times less than yours, the design flaw right? 😉
20Hz H3 is about two times less than yours, the design flaw right? 😉
biggest design flaw = small dip switches 😀
As for the input capacitors, you could have left more room for larger ones. Otherwise, a great box. 👍
I think the D10s is perfectly fine. (for the price)
Last comparison. Cosmos Rch with 220u/10v and Lch with 470u/10v super coupling capacitor Jamicon. (the price is also amazing, just 0.11euro)
As for the input capacitors, you could have left more room for larger ones. Otherwise, a great box. 👍
I think the D10s is perfectly fine. (for the price)
Last comparison. Cosmos Rch with 220u/10v and Lch with 470u/10v super coupling capacitor Jamicon. (the price is also amazing, just 0.11euro)
I told you, you have defective 220uF caps, try to replace for the same value brand new to see the same THD@20Hz. BTW, I have a few, perhaps 2-3 cases when users burned these caps by overvoltage during a poweramp testing. After caps were replaced ADC works properly.
Thanks very much. I knew I was just being unobservant.
Now, about your issue...
I see that you're using the "classic" Cosmos ADC. No USB isolation.
Here's what I think.
The ADC has limited CMRR - I think Ivan may have measured around 60+ dB. That's about what I measured, too. This degrades as you go lower in frequency. That's because the caps are in series with the inputs. Large electrolytic capacitors don't have close tolerances. Hold that thought.
CMRR in most circuits relies on balance of the input circuit and their relative impedances. Rather than me trying to explain, have a look here:
Bill Whitlock on Balanced Interfaces
So, you have a couple challenges with an unbalanced source like the D10s run into a Cosmos ADC.
The first is that those caps at the input are part of the virtual bridge circuit that ultimately determines common mode rejection. Since the caps aren't perfectly matched, they add to the imbalance in the bridge. At high frequencies the cap's impedance goes toward zero. That means that they are far less of a cause for imbalance in the bridge because the bridge resistors are the dominant factor. As you go lower in frequency, the coupling cap impedance rises and their contribution to the bridge becomes more significant. That degrades the CMRR at low frequencies. If you use larger value caps, that problem becomes less since their effective impedance is less. Try using 10,000 uF and see what happens.
The second is that your source isn't balanced with equal impedance between the "hot" connection and the load and between the "cold/ground" connection and the load. From what I just read, the D10s output impedance is 10 Ohms. That's for the normal signal connection and is what the bridge is fed with. But, on the ground side, it's close to zero Ohms. That's a significant mismatch for the bridge, especially with the low impedance of the internal attenuator of the ADC. (You might try adding a 10 Ohm resistor in the ground connection as an experiment.)
So, why do you care? Well, USB interfaces have significant noise, especially at low frequencies, that are caused by common mode current loops. That's why there's a new Cosmos ADCiso now. Previously I posted the effect of using a USB isolator with an original classic Cosmos ADC. Without isolation of some kind, there's a lot of junk, especially at lower frequencies. The Cosmos ADC's very good but limited CMRR can't ignore this junk. Since the Cosmos ADC's CMRR degrades as you go lower in frequency - see above - these common mode disturbances are detected as distortion by REW.
There was a discussion of this elsewhere: PMA on using a D10s and Cosmos ADC as a test system
So, to maximize performance, my recipe is to use a Cosmos ADCiso with the built in USB isolator. But, a proper external USB isolator does as well and is what I used to use. I also always use a Cosmos Scaler with the ADC, even if the gain is set to zero. That way the ADCiso is always fed a proper balanced signal. I also changed from a Topping D10s as a source to a Topping E50. That gives both unbalanced and balanced outputs, is powered by an external source (I use a battery), and inherently has lower distortion. (If a Cosmos DAC becomes available, I'll try that). All of that gives really great results.
BTW, this isn't just a measurement system issue. What's the CMRR of your preamp and power amp? Do you use a USB sourced DAC with or without a USB isolator? How much isolation does your preamp, power amp, and whatever else is connected to the system have with regard to common mode currents passing through the AC mains connections?
BTW2, think about how audio products are usually tested. Somebody connects the device under test to the test signal output of an Audio Precision system and examines the resulting signal through the test input of the AP. If nothing else, Audio Precsion has really gone to heroic lengths to maximize the CMRR of the generator output and the detector input. That means that the test system has minimal effect on the DUT performance. Just as advertised. But, that's not the real world. Almost all hifi components do affect the DUT performance and are susceptible to common mode and other disturbances that the AP is pretty much immune to. So, even if a DUT tests perfectly with an AP system, that doesn't mean that it performs well in a real system, or - horrors - that it sounds good. Show me a test of a complete system...
I apologize if you already know all this information. I'm not trying to school you or show what I might know. But, perhaps other people have similar questions and would like to get a better understanding.
Now, about your issue...
I see that you're using the "classic" Cosmos ADC. No USB isolation.
Here's what I think.
The ADC has limited CMRR - I think Ivan may have measured around 60+ dB. That's about what I measured, too. This degrades as you go lower in frequency. That's because the caps are in series with the inputs. Large electrolytic capacitors don't have close tolerances. Hold that thought.
CMRR in most circuits relies on balance of the input circuit and their relative impedances. Rather than me trying to explain, have a look here:
Bill Whitlock on Balanced Interfaces
So, you have a couple challenges with an unbalanced source like the D10s run into a Cosmos ADC.
The first is that those caps at the input are part of the virtual bridge circuit that ultimately determines common mode rejection. Since the caps aren't perfectly matched, they add to the imbalance in the bridge. At high frequencies the cap's impedance goes toward zero. That means that they are far less of a cause for imbalance in the bridge because the bridge resistors are the dominant factor. As you go lower in frequency, the coupling cap impedance rises and their contribution to the bridge becomes more significant. That degrades the CMRR at low frequencies. If you use larger value caps, that problem becomes less since their effective impedance is less. Try using 10,000 uF and see what happens.
The second is that your source isn't balanced with equal impedance between the "hot" connection and the load and between the "cold/ground" connection and the load. From what I just read, the D10s output impedance is 10 Ohms. That's for the normal signal connection and is what the bridge is fed with. But, on the ground side, it's close to zero Ohms. That's a significant mismatch for the bridge, especially with the low impedance of the internal attenuator of the ADC. (You might try adding a 10 Ohm resistor in the ground connection as an experiment.)
So, why do you care? Well, USB interfaces have significant noise, especially at low frequencies, that are caused by common mode current loops. That's why there's a new Cosmos ADCiso now. Previously I posted the effect of using a USB isolator with an original classic Cosmos ADC. Without isolation of some kind, there's a lot of junk, especially at lower frequencies. The Cosmos ADC's very good but limited CMRR can't ignore this junk. Since the Cosmos ADC's CMRR degrades as you go lower in frequency - see above - these common mode disturbances are detected as distortion by REW.
There was a discussion of this elsewhere: PMA on using a D10s and Cosmos ADC as a test system
So, to maximize performance, my recipe is to use a Cosmos ADCiso with the built in USB isolator. But, a proper external USB isolator does as well and is what I used to use. I also always use a Cosmos Scaler with the ADC, even if the gain is set to zero. That way the ADCiso is always fed a proper balanced signal. I also changed from a Topping D10s as a source to a Topping E50. That gives both unbalanced and balanced outputs, is powered by an external source (I use a battery), and inherently has lower distortion. (If a Cosmos DAC becomes available, I'll try that). All of that gives really great results.
BTW, this isn't just a measurement system issue. What's the CMRR of your preamp and power amp? Do you use a USB sourced DAC with or without a USB isolator? How much isolation does your preamp, power amp, and whatever else is connected to the system have with regard to common mode currents passing through the AC mains connections?
BTW2, think about how audio products are usually tested. Somebody connects the device under test to the test signal output of an Audio Precision system and examines the resulting signal through the test input of the AP. If nothing else, Audio Precsion has really gone to heroic lengths to maximize the CMRR of the generator output and the detector input. That means that the test system has minimal effect on the DUT performance. Just as advertised. But, that's not the real world. Almost all hifi components do affect the DUT performance and are susceptible to common mode and other disturbances that the AP is pretty much immune to. So, even if a DUT tests perfectly with an AP system, that doesn't mean that it performs well in a real system, or - horrors - that it sounds good. Show me a test of a complete system...
I apologize if you already know all this information. I'm not trying to school you or show what I might know. But, perhaps other people have similar questions and would like to get a better understanding.
Last edited:
I'm using Topping HS02 usb isolator between the Cosmos and PC
BTW, I'm preparing active notch filter for my cosmos (unbalanced, for 1.7Vrms range )
If better connectors were used in the Cosmos APU, I would buy it
What I definitely want to buy is the Cosmos DAC
IVX, sorry for offtopic
BTW, I'm preparing active notch filter for my cosmos (unbalanced, for 1.7Vrms range )
If better connectors were used in the Cosmos APU, I would buy it
What I definitely want to buy is the Cosmos DAC
IVX, sorry for offtopic
Last edited:
A few ASR forum members or my discord members perhaps, they asked me to add 48V power, kind of it is a good idea to let people use Neumann 103 with the APU. I said ok and took off the switchable passive LPF from the LNA to place there phantom power. Finally, no one reported if use 48V power, and people adding an external passive LPF..