The voltage does go up when the selected channels are exclusively in "pre-out" mode (at least this is how the 3800 behaved).
-
Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
Denon avr-x3900h preamp measurements
- Thread starter alex.schroedsen
- Start date
That’s a lot of work, thank you so much.
I think a possible reason for the strange looking FFT spectra was part of the recorded test signal was muted. For example, the X3900H, for some reason, delayed starting to output the test signal when the RMAA program was expecting it. So RMAA, while expecting to record a full length test signal, instead recorded a brief silence first then followed the actual test signal. When the FFT computations did not account for the unexpected initial silence, the FFT spectra plots will have the kind of spectral spreading shown in OP's test results.
OP
alex.schroedsen
Member
If you don’t mind, I would love to see another SINAD measurement on tbe pre outs in preamp mode and analog input in direct mode, just to see if the results will get closer to your THD numbers, and/or the >95 dB range.
OP
alex.schroedsen
Member
- Thread Starter
- #46
@peng here are those measurements you were looking for (DAC source is the Matrix Mini-i Pro 2 measured earlier):
RightMark Audio Analyzer test report
| Testing device | Denon x3900h Analog CD IN Pure Direct |
| Sampling mode | 32-bit, 96 kHz |
| Interface | |
| Testing chain | External loopback (line-out - line-in) |
| RMAA Version | 6.4.5 |
| 20 Hz - 20 kHz filter | ON |
| Normalize amplitude | ON |
| Level change | -11.6 dB / -11.5 dB |
| Mono mode | OFF |
| Calibration singal, Hz | 1000 |
| Polarity | inverted/inverted |
Summary
| Frequency response (from 40 Hz to 15 kHz), dB | +0.02, -0.27 | Very good |
| Noise level, dB (A) | -116.1 | Excellent |
| Dynamic range, dB (A) | 116.1 | Excellent |
| THD, % | 0.00109 | Excellent |
| THD + Noise, dB (A) | -96.6 | Excellent |
| IMD + Noise, % | 0.00159 | Excellent |
| Stereo crosstalk, dB | -61.0 | Average |
| IMD at 10 kHz, % | 0.00195 | Excellent |
| General performance | | Very good |
Frequency response
Left | Right | |
| From 20 Hz to 20 kHz, dB | -0.91, +0.02 | -0.75, +0.17 |
| From 40 Hz to 15 kHz, dB | -0.27, +0.02 | -0.11, +0.17 |
Noise level
Left | Right | |
| RMS power, dB | -111.2 | -112.2 |
| RMS power (A-weighted), dB | -116.5 | -115.8 |
| Peak level, dB FS | -62.0 | -61.9 |
| DC offset, % | -0.0 | -0.0 |
Dynamic range
Left | Right | |
| Dynamic range, dB | +111.2 | +112.2 |
| Dynamic range (A-weighted), dB | +116.4 | +115.8 |
| DC offset, % | -0.02 | -0.02 |
THD + Noise (at -3 dB FS)
Left | Right | |
| THD, % | 0.00110 | 0.00109 |
| THD + Noise, % | 0.00138 | 0.00132 |
| THD + Noise (A-weighted), % | 0.00148 | 0.00147 |
Intermodulation distortion
Left | Right | |
| IMD + Noise, % | 0.00160 | 0.00157 |
| IMD + Noise (A-weighted), % | 0.00114 | 0.00114 |
Stereo crosstalk
Left | Right | |
| Crosstalk at 100 Hz, dB | -20 | -104 |
| Crosstalk at 1000 Hz, dB | -24 | -96 |
| Crosstalk at 10000 Hz, dB | -24 | -84 |
IMD (swept tones)
Left | Right | |
| IMD + Noise at 5000 Hz, | 0.00151 | 0.00152 |
| IMD + Noise at 10000 Hz, | 0.00253 | 0.00257 |
| IMD + Noise at 15000 Hz, | 0.00178 | 0.00180 |
OP
alex.schroedsen
Member
- Thread Starter
- #47
In order to try to attempt avoidance of the delayed starting problem @NTK mentioned, I have also recorded a test over SPDIF using a CM6631A SPDIF transmitter. Sadly the results look a lot like the HDMI capture. I made sure to specify the 24-bit width to avoid truncating/padding of bits as could be the reason for the spectral spreading in the THD graphs.
RightMark Audio Analyzer test report
| Testing device | Denon x3900h SPDIF IN Pure Direct 82.5 |
| Sampling mode | 24-bit, 48 kHz |
| Interface | |
| Testing chain | External loopback (line-out - line-in) |
| RMAA Version | 6.4.5 |
| 20 Hz - 20 kHz filter | ON |
| Normalize amplitude | ON |
| Level change | -12.0 dB / -11.9 dB |
| Mono mode | OFF |
| Calibration singal, Hz | 1000 |
| Polarity | inverted/inverted |
Summary
| Frequency response (from 40 Hz to 15 kHz), dB | +0.04, -0.23 | Very good |
| Noise level, dB (A) | -112.6 | Excellent |
| Dynamic range, dB (A) | 112.4 | Excellent |
| THD, % | 0.00087 | Excellent |
| THD + Noise, dB (A) | -84.7 | Good |
| IMD + Noise, % | 0.015 | Very good |
| Stereo crosstalk, dB | -85.4 | Excellent |
| IMD at 10 kHz, % | 0.00166 | Excellent |
| General performance | | Very good |
Frequency response
Left | Right | |
| From 20 Hz to 20 kHz, dB | -0.86, +0.04 | -0.73, +0.16 |
| From 40 Hz to 15 kHz, dB | -0.23, +0.04 | -0.10, +0.16 |
Noise level
Left | Right | |
| RMS power, dB | -108.3 | -108.4 |
| RMS power (A-weighted), dB | -112.8 | -112.4 |
| Peak level, dB FS | -61.5 | -61.4 |
| DC offset, % | +0.0 | -0.0 |
Dynamic range
Left | Right | |
| Dynamic range, dB | +108.2 | +108.3 |
| Dynamic range (A-weighted), dB | +112.6 | +112.2 |
| DC offset, % | -0.00 | -0.01 |
THD + Noise (at -3 dB FS)
Left | Right | |
| THD, % | 0.00093 | 0.00081 |
| THD + Noise, % | 0.00530 | 0.00528 |
| THD + Noise (A-weighted), % | 0.00583 | 0.00580 |
Intermodulation distortion
Left | Right | |
| IMD + Noise, % | 0.01456 | 0.01455 |
| IMD + Noise (A-weighted), % | 0.00383 | 0.00382 |
Stereo crosstalk
Left | Right | |
| Crosstalk at 100 Hz, dB | -63 | -102 |
| Crosstalk at 1000 Hz, dB | -70 | -99 |
| Crosstalk at 10000 Hz, dB | -70 | -89 |
IMD (swept tones)
Left | Right | |
| IMD + Noise at 5000 Hz, | 0.00141 | 0.00134 |
| IMD + Noise at 10000 Hz, | 0.00140 | 0.00140 |
| IMD + Noise at 15000 Hz, | 0.00218 | 0.00220 |
Unfortunately, may be it is time to try different measurement software, such as multi-tone analyzer or REW.
popej
Addicted to Fun and Learning
- Joined
- Jul 13, 2023
- Messages
- 775
- Likes
- 682
Noise is measured without signal.
THD is measured with max signal.
You can see on graphs, that background noise increases with signal. That is the reason, that measurements don't add up. Difficult to guess, if this is property of DAC or problem in measurements.
Last edited:
popej
Addicted to Fun and Learning
- Joined
- Jul 13, 2023
- Messages
- 775
- Likes
- 682
I think the OP understands my points. Without repeating things mentioned before, THD+N -54.6 dB just doesn’t seem to make sense, not for a 3000 series Denon avr, regardless.
Hopeful we will see another set of measurements soon, by Amir.
Last edited:
OP
alex.schroedsen
Member
- Thread Starter
- #52
@peng I am also looking forward to Amir's measurements. I plan to try the REW or multi-tone analyzer software @NTK suggested when I have another chance to do so. The analog pure direct throughput made it up to -96dB THD+N with Righmark. We'll see if any other software yields different results.
@popej thanks for the suggestion that it could be a problem in the signal path. The current signal path has mostly been Rightmark->FlexASIO->WDM-KS->digital audio into the Denon x3900h->rca to XLR cables->E1DA Cosmos ADCiso (1.77v input)->WDM-KS->FlexASIO->Rightmark. The only exception are the post 1 measurements that had a couple problems: (1) the ADCiso was set to mono mode (2) FlexASIO was set to WASAPI, which was resampling the signal. Those have since been corrected. The Denon x3900h has been set for crossovers to be full range and for processing mode to be pure direct. Let me know if you think of something I may have missed.
@popej thanks for the suggestion that it could be a problem in the signal path. The current signal path has mostly been Rightmark->FlexASIO->WDM-KS->digital audio into the Denon x3900h->rca to XLR cables->E1DA Cosmos ADCiso (1.77v input)->WDM-KS->FlexASIO->Rightmark. The only exception are the post 1 measurements that had a couple problems: (1) the ADCiso was set to mono mode (2) FlexASIO was set to WASAPI, which was resampling the signal. Those have since been corrected. The Denon x3900h has been set for crossovers to be full range and for processing mode to be pure direct. Let me know if you think of something I may have missed.
That’s put it in the same ball park where the X4700H was when Amir measured it.
In the past Amir would measured it with HDMI, Coax and analog input, and the results, as was the case with the X4700H, would be roughly the same. In this case, your digital input result was -54.6, that’s a lot worse than the -96.6 dB with the analog output.
So it does look like there is something in the signal path when you measured the preout and using digital input that might have causef the poor SINAD.
Last edited:
I asked chatgpt: What is the likely cause of the THD+N measured at -54,6 db while the analogue input yielded -96 THD+N ?
Pretty good answer, I think:
The analog-input measurement of roughly -96 dB THD+N is actually in line with what Denon specifies for the analog stages (around 0.008% distortion and ~100 dB SNR in Direct mode). �
Denon User Manual
A THD+N of -54.6 dB corresponds to about 0.19% THD+N. That's so poor that it would imply either:
severe clipping,
a DSP arithmetic issue,
a level mismatch,
or some processing artifact.
It is not consistent with the performance of the DAC chips or analog output stage themselves.
What stood out in those posts is that the FFT showed distortion products characteristic of signal overload rather than a noisy DAC. If the digital source was fed at or near 0 dBFS and the AVR subsequently applied any internal gain (speaker trims, Audyssey filters, bass management, Dynamic EQ, etc.), an internal DSP stage can run out of headroom. Many AVRs historically have had limited digital headroom when processing full-scale signals.
The clue is the enormous gap:
Path
THD+N
Analog input
~ -96 dB
Digital input
~ -54.6 dB
If the DAC itself were responsible, you'd expect perhaps a 5–15 dB difference, not 40+ dB.
My reading of those posts is that the digital measurement is revealing internal DSP overload/headroom limitations in the HDMI/digital path, whereas the analog measurement was taken under conditions that bypassed or avoided that overload mechanism.
Another clue is that several Denon/Marantz generations have shown behavior where reducing the digital signal level by a few dB dramatically improves measured distortion because the DSP no longer clips internally. That kind of result is exactly what you'd expect if a 0 dBFS test tone is being processed with insufficient internal headroom.
So if I had to assign probabilities:
Internal DSP clipping/headroom issue: 70%
Measurement setup or level calibration error: 20%
Actual DAC/output-stage limitation: 10%
Last edited:
popej
Addicted to Fun and Learning
- Joined
- Jul 13, 2023
- Messages
- 775
- Likes
- 682
Another possibility:
AVR can have a short delay, when playback starts. It can truncate the beginning of test signal. It could be not noticeable with music, but it can disrupt RMAA analysis.
I think you should try other software for measurement. Or you can try to add a bit of noise at the beginning of RMAA signal.
This is how RMAA signal looks like:
You can record ADC input and check if it is complete.
AVR can have a short delay, when playback starts. It can truncate the beginning of test signal. It could be not noticeable with music, but it can disrupt RMAA analysis.
I think you should try other software for measurement. Or you can try to add a bit of noise at the beginning of RMAA signal.
This is how RMAA signal looks like:
You can record ADC input and check if it is complete.
Last edited:
popej
Addicted to Fun and Learning
- Joined
- Jul 13, 2023
- Messages
- 775
- Likes
- 682
I have done RMAA measurements of my LX305 and the conclusion is, that you need recording level around -1dB. You got good results with Matrix, where level was -1.6dB (level change in report), but for Denon you got always about -12dB. Try to increase volume or ADC sensivity.
Hi, good work.
For your measurement setup I would change E1DA ADCiso input range to 4,5V. ADCiso has very small input impedance (for 1,7V only around 500 Ohm for unbalanced connection).
With 4,5V range it is around 1,1 kOhm.
Info from the E1DA Cosmos ADCiso web:
That is why you have -12 dBFS of the signal because ADCiso is too big load (500 Ohm) for your DENON and interacts with the output impedance of the pre-outs. Or use Cosmos Scaler in front of the ADC.
Than, try not to use FlexASIO - you can use Win drivers directly for your setup with Rightmark. I suppose, it can help with the "wrinkles/waves" at your FFT graphs.
There could be some problem with a "digital leak" from somewhere, also - the noise level graph should look like this (without 1 kHz tone):
Also IMD should look more like this:
Last edited:
OP
alex.schroedsen
Member
- Thread Starter
- #58
Good afternoon everyone!! After following the advice from @eeMGee and also boosting the volume just a little. I'm getting the numbers I was hoping for. Thanks everyone for the education you've given and the advice you've provided throughout the process of my learning in this.
RightMark Audio Analyzer test report (48-kHz)
| Testing device | Denon x3900h preamp full range xover 89v 1.66kohm |
| Sampling mode | 32-bit, 48 kHz |
| Interface | |
| Testing chain | External loopback (line-out - line-in) |
| RMAA Version | 6.4.5 |
| 20 Hz - 20 kHz filter | ON |
| Normalize amplitude | ON |
| Level change | -8.1 dB / -8.0 dB |
| Mono mode | OFF |
| Calibration singal, Hz | 1000 |
| Polarity | correct/correct |
Summary (48-kHz)
| Frequency response (from 40 Hz to 15 kHz), dB | +0.03, -0.20 | Very good |
| Noise level, dB (A) | -113.7 | Excellent |
| Dynamic range, dB (A) | 112.9 | Excellent |
| THD, % | 0.00030 | Excellent |
| THD + Noise, dB (A) | -102.3 | Excellent |
| IMD + Noise, % | 0.00095 | Excellent |
| Stereo crosstalk, dB | -105.3 | Excellent |
| IMD at 10 kHz, % | 0.00210 | Excellent |
| General performance | | Very good |
Frequency response (48-kHz)
Left | Right | |
| From 20 Hz to 20 kHz, dB | -0.52, +0.03 | -0.42, +0.13 |
| From 40 Hz to 15 kHz, dB | -0.20, +0.03 | -0.10, +0.13 |
Noise level (48-kHz)
Left | Right | |
| RMS power, dB | -112.8 | -112.2 |
| RMS power (A-weighted), dB | -114.0 | -113.4 |
| Peak level, dB FS | -97.9 | -97.1 |
| DC offset, % | +0.0 | +0.0 |
Dynamic range (48-kHz)
Left | Right | |
| Dynamic range, dB | +112.6 | +112.0 |
| Dynamic range (A-weighted), dB | +113.2 | +112.6 |
| DC offset, % | +0.00 | +0.00 |
THD + Noise (at -3 dB FS) (48-kHz)
Left | Right | |
| THD, % | 0.00033 | 0.00026 |
| THD + Noise, % | 0.00077 | 0.00075 |
| THD + Noise (A-weighted), % | 0.00079 | 0.00075 |
Intermodulation distortion (48-kHz)
Left | Right | |
| IMD + Noise, % | 0.00096 | 0.00094 |
| IMD + Noise (A-weighted), % | 0.00073 | 0.00075 |
Stereo crosstalk (48-kHz)
Left | Right | |
| Crosstalk at 100 Hz, dB | -107 | -110 |
| Crosstalk at 1000 Hz, dB | -105 | -103 |
| Crosstalk at 10000 Hz, dB | -85 | -85 |
IMD (swept tones) (48-kHz)
Left | Right | |
| IMD + Noise at 5000 Hz, | 0.00121 | 0.00120 |
| IMD + Noise at 10000 Hz, | 0.00168 | 0.00173 |
| IMD + Noise at 15000 Hz, | 0.00336 | 0.00340 |
Last edited:
OP
alex.schroedsen
Member
- Thread Starter
- #59
Here are the results at 44.khz. Still has a funky frequency response, but the other stats have improved almost as much.
RightMark Audio Analyzer test report (44.1kHz)
| Testing device | Denon x3900h preamp full range xover 89v 1.66kohm |
| Sampling mode | 32-bit, 44 kHz |
| Interface | |
| Testing chain | External loopback (line-out - line-in) |
| RMAA Version | 6.4.5 |
| 20 Hz - 20 kHz filter | ON |
| Normalize amplitude | ON |
| Level change | -8.1 dB / -8.0 dB |
| Mono mode | OFF |
| Calibration singal, Hz | 1000 |
| Polarity | correct/correct |
Summary (44.1kHz)
| Frequency response (from 40 Hz to 15 kHz), dB | +1.17, -3.24 | Poor |
| Noise level, dB (A) | -112.9 | Excellent |
| Dynamic range, dB (A) | 93.5 | Very good |
| THD, % | 0.00033 | Excellent |
| THD + Noise, dB (A) | -102.1 | Excellent |
| IMD + Noise, % | 0.00097 | Excellent |
| Stereo crosstalk, dB | -105.2 | Excellent |
| IMD at 10 kHz, % | 0.00212 | Excellent |
| General performance | | Very good |
Frequency response (44.1kHz)
Left | Right | |
| From 20 Hz to 20 kHz, dB | -10.40, +1.06 | -10.29, +1.17 |
| From 40 Hz to 15 kHz, dB | -3.34, +1.06 | -3.24, +1.17 |
Noise level (44.1kHz)
Left | Right | |
| RMS power, dB | -111.7 | -111.3 |
| RMS power (A-weighted), dB | -113.1 | -112.8 |
| Peak level, dB FS | -98.4 | -97.7 |
| DC offset, % | +0.0 | -0.0 |
Dynamic range (44.1kHz)
Left | Right | |
| Dynamic range, dB | +94.5 | +94.3 |
| Dynamic range (A-weighted), dB | +93.6 | +93.5 |
| DC offset, % | +0.00 | -0.00 |
THD + Noise (at -3 dB FS) (44.1kHz)
Left | Right | |
| THD, % | 0.00037 | 0.00028 |
| THD + Noise, % | 0.00080 | 0.00078 |
| THD + Noise (A-weighted), % | 0.00081 | 0.00076 |
Intermodulation distortion (44.1kHz)
Left | Right | |
| IMD + Noise, % | 0.00099 | 0.00096 |
| IMD + Noise (A-weighted), % | 0.00079 | 0.00076 |
Stereo crosstalk (44.1kHz)
Left | Right | |
| Crosstalk at 100 Hz, dB | -108 | -107 |
| Crosstalk at 1000 Hz, dB | -105 | -104 |
| Crosstalk at 10000 Hz, dB | -84 | -84 |
IMD (swept tones) (44.1kHz)
Left | Right | |
| IMD + Noise at 5000 Hz, | 0.00131 | 0.00123 |
| IMD + Noise at 10000 Hz, | 0.00170 | 0.00167 |
| IMD + Noise at 15000 Hz, | 0.00344 | 0.00339 |
Thank you! @alex.schroedsen
All looks very good.
Except for the frequency reponse at 44.1 khz.
Surely this can't be right if the frequency response at 48khz is almost a straight line?
44.1khz
48 khz
All looks very good.
Except for the frequency reponse at 44.1 khz.
Surely this can't be right if the frequency response at 48khz is almost a straight line?
44.1khz
48 khz
Similar threads
