Audio Science Review of Expert 200.

[markush]

Yes that’s actually what they claim - perfectly measured amp. Don‘t know if we are investigating a flaw in that design or being on the wrong track. But I think it’s worth getting into more detail. And maybe Someone can have a feedback from Matheau.

[thumb5]

...And maybe Someone can have a feedback from Matheau.

Watch this space...

[Jean-Marie]

With reference to a number of posts on page 2 of this thread, and latterly @Bwaze 's post above, I was interested in making a few "real world" measurements myself, just because I wanted improve my own understanding of these measurements, and what might going on in the tests.

I do not own an Audio Precision analyser, or anything remotely similar, so I am not able to repeat or challenge the measurements made by ASR, Hifi News or anyone else.  But I do have a multi meter.

So, I connected some speaker plugs the cables on my multi meter, plugged it into the spare terminals on one speaker, ran a continuous 10kHz test tone, and checked the voltages at different gains on the Devialet.  Initially I checked the voltage at very low gains, and then calculated the wattage this equated to.  Reasonably confident that power levels were low enough, I then increased the gain. The first thing I discovered is my multi-meter is not that accurate with low voltage.  You can set it for different voltage ranges, I used the 2V range and 20V range.  As an example, the reading with 10Khz at -30dB was 0.1V using the 20V range, or 0.38V using the 2 V range.  So one thing I have learnt is that I probably need to buy a better multi-meter, although my old one has served me well for what I need.  I then repeated the above using a 15kHz tone.  (which I was quite please to note that I could hear very clearly)

What I was particularly interested in understanding was exactly how much power a 10kHz or 15kHz test tone would require at anything up to 0dB gain.

To calculate the power, I needed to know the resistance that the Blades are presenting at the particular frequency.  I do not have an impedance curve for the Blade, but Stereophile have published one for the Blade 2, which I think has the same HF driver, so this should be near enough.  Here I  conservatively took the impedance at 10kHz as 5 ohms, and at 15kHz 4.5 ohms.

These are the results:

10kHz tone:

-30dB gain, 0.08V, approximately 0.001W
-20dB gain, 0.38V, approximately 0.03W
-10dB gain, 1.44V, approximately  0.41W
0dB gain, 5.8V, approximately 5.8W

15kHz test tone:

-30dB gain, 0.02V, approximately 0.0001W
-20dB gain, 0.21V, approximately 0.01W
-10dB gain, 0.88V, approximately 0.17W
0dB gain, 4.0V, approximately 3.6W

In an earlier post @ogs made the point "Your Blades are quite sensitive at 91dB (2.83V/1m). 90dB SPL at your listening position would be around 2 watts..."    In terms of normal listening, I agree with ogs here.  Normally I use HQPlayer with upsampling, and run HQPlayer at -4 dB to prevent issues with inter-sample clipping.  Even with this, I am usually at -20dB or lower, and rarely go above -10dB even with "quiet" recordings played loudly.  I mention this because with the above figures, HQPlayer was set to 0.0dB, so even with a level of gain way above what I would ever normally use, the highest wattage I have measured is 5.8W for the HF tone.  In normal use, running my system with 0dB gain would lead to very high volumes, certainly enough to cause hearing damage, possibly enough to cause speaker damage.

This makes me think that to get higher wattages during the kind of tests ASR and Stereophile are doing, you would need to run a very high voltage, equivalent to huge positive gain, into the Devialet's analogue input.  This is the only way I can see how you could get to 20 or even 10 watts.  Have I got this right?  If so, I assume the Devialet sees the high voltage on the analogue input, it thinks this is wrong, and shuts down.  I am speculating a little here, so would be interested in the views of others here.

In practical terms, say if you wanted to use your Devialet for something like listening to music at a reasonably high volumes, I would say the above numbers suggest this testing anomaly is not of too much concern.  Thankfully, trying this test did not cause my Devialet to shut down, and it has not fried my speakers.  For this I am very grateful (I was a little worried about trying this) and I shall now proceed to using my Devialet for the above mentioned purpose.

Confused,

Is your multi-meter analog or digital?
If it is digital, and a “normal” general public one, I highly doubt that it can be used to measure a signal beyond a few 100s Hz. 
Analog ones, usually are able to go higher, and digital ones able to go up to 20kHz would certainly cost several 100s of € at least. 

your measurements at 0dB show 3dB difference between 10kHz and 15kHz, way worse  any linearity that have ever measured on any amplifier I know of, including the Devialets. 

Therefore I highly suspect that the multi-meter is to be investigated whether or not it is able to handle properly a 15kHz signal. 

Jean-Marie

[Confused]

@Jean-Marie - To clarify, yes it is a digital multi-meter.  As for its accuracy, as I mentioned in my earlier post the readings are not even consistent between the 2V and 20V settings, and as I also said earlier I am not trying to challenge ASR or Stereophile, I am just trying to improve understanding, and not claiming to be making definitive measurements here.

As for the difference between the 10kHz and 15kHz, the test tones were taken from the internet and happened to be from different sources, so it is highly likely that most of the 3dB difference is accounted for by differences in the test tone.  I agree that the multi-meter may lack accuracy, but I doubt its behaviour differs much between 10kHz and 15kHz, when it is probably optimised for nearer 50Hz.  That said, this was not intended as test for linearity, just a simple test to try and improve understanding.

Anyway, I am very pleased to see you are taking an interest in this thread.  One thing we have all failed to understand is exactly why a Devialet Expert amp will shut down at very low watts during certain high frequency test scenarios.  I believe you might have a contact who could perhaps provide some technical insight here?  If you could find some relevant information, that would be greatly appreciated.

@thumb5 - you make a very good point regarding the actual input level resulting from the test tone.  I had a quick look at the input level meters on the amp itself.  I have no idea how accurate this meter is but with the test tone it was just running into the "orange zone", it is a tiny scale, but it looks to be maybe -18dB or so.  I certainly have actual real recorded music that will push these meters higher into the red zone.  This is very interesting, at it shows that overall my test was running maybe -18dB gain with respect to a "full voltage" test tone, not the full 0dB that I mentioned.  In a way I am quite pleased about this, I am slightly uncomfortable about possibility of causing damage running pure tone tests, so it looks like the test tone itself provided a little headroom here.  I'm thinking that it would be easy enough to analyse the test tones to get a very accurate idea of the absolute level of the signal, but as the multi-meter is likely inaccurate, this is a fairly moot point. 

@"alaw" - enthusiasm can get you into trouble sometimes, I am more than happy that I have tried this with the amp and speakers unscathed.  An exploded Blade tweeter or something during the current UK "lockdown" would not be good.  I am happy to give testing a rest now and return to the listening of music. 

As an aside, the multi-meter issues raised here reminds me of an old article by Archimago

http://archimago.blogspot.com/2017/09/mu...o-you.html

The multi-meter shown in the article is not that different to mine, so as an open question for anyone wanting to take such measurements with a multi-meter, what type is best for high frequency audio?  Archimago seamed happy enough with his, but maybe he was wrong?  Is there something more suitable?  Maybe analogue is better?  Are there issues with multi-meters presenting a resistance of their own?  There is a whole topic here I think.

[disarmamant]

Is there something more suitable?  Maybe analogue is better?  Are there issues with multi-meters presenting a resistance of their own?  There is a whole topic here I think.

Multimeters are unsuitable for making AC measurements at high frequencies, they are mainly designed for making AC measurements at 50 or 60 Hz. You are quite lucky you did no damage to your speakers by attempting to make high frequency measurements at potentially such high power levels.

If I understand the level control on the Devialet correctly, if for example you were using a 2 volt input signal into one of the analogue line inputs, and the maximum level for the line input was set to 2 volts on the Devialet, with the volume set at 0dB then the amplifier should be outputting a signal at full rated power.

If that was the case you would surely not be able to be in the same room with the power at that level!

The only way to achieve a meaningful measurement would be to replace the speaker with an 8 ohm resistive dummy load rated at at least one kilowatt. You could then measure the peak to peak sine wave AC waveform with an oscilloscope across the terminals of the dummy load.


Sent from my iPad using Tapatalk

[Confused]

Is there something more suitable?  Maybe analogue is better?  Are there issues with multi-meters presenting a resistance of their own?  There is a whole topic here I think.

Multimeters are unsuitable for making AC measurements at high frequencies, they are mainly designed for making AC measurements at 50 or 60 Hz. You are quite lucky you did no damage to your speakers by attempting to make high frequency measurements at potentially such high power levels.

If I understand the level control on the Devialet correctly, if for example you were using a 2 volt input signal into one of the analogue line inputs, and the maximum level for the line input was set to 2 volts on the Devialet, with the volume set at 0dB then the amplifier should be outputting a signal at full rated power.

If that was the case you would surely not be able to be in the same room with the power at that level!

The only way to achieve a meaningful measurement would be to replace the speaker with an 8 ohm resistive dummy load rated at at least one kilowatt. You could then measure the peak to peak sine wave AC waveform with an oscilloscope across the terminals of the dummy load.


Sent from my iPad using Tapatalk

I tend to agree with your points here.  However, I think I may not have been clear enough myself regarding what I was actually trying to achieve with my earlier measurements.  I mentioned in my earlier post that I wanted to make some "real world" measurements to improve my own understanding of this HF power issue.  To be clearer, what I did NOT want to do is to try to attempt a "full power" test.  You mention I was lucky not to do any damage and that I would not be able to be in the room during a full power test.  In fact, the approach was to start at a very low gain, I was not only using a volt meter but I also had a SPL meter.  I was creeping the gain up slowly to see the influence on both volume and SPL.  I would have stopped the test before I got to very high SPL's.  There might be a degree of bravery in doing tests like this, but fear of damaging the kit kept me very much on the cautious side of things.  For the record, the 10kHz tone ran to about 95dB @ 1m, no more, which was subjectively a little uncomfortable, but I kept it at this level for only a little over a second or so.  Plus, in terms of my own bravery in testing, this was far enough for me to quit, and had the test tone been at a higher level, I would certainly have quit at about -20dB on the Devialet's gain, as I said, I was just trying to get a feel for real world power levels, I was not attempting a full power test.

Or to put this another way, if someone were to try a test as you describe, with a one kilowatt resistor and a 2V analogue input, we know what the end result would be, the amp would shut down.  Which I guess takes us full circle to one of the key points of this thread, why do Devialets shut down during HF testing at relatively low power levels?

Thinking about my own results a little more, I got to just under 6 watts with SPL's in the mid 90's, at 0dB gain.  Not that loud or that much power.  Considering the 10kHz tone was at about -18dB, these numbers actually look about right.  Certainly not super accurate, but in the right ball park.  So in theory, if the gain is then increased to +18, or a full 0dB test tone is used, then mathematically this would need about 380 watts to produce the tone at 113dB.  So maybe digital mutli-meters are not suitable for high frequency measurements, but in this case I suspect the results are not actually that far out, they look reasonable at least.

A few comments in this thread mentioning that digital multi-meters are no good for HF measurements, but does anyone know any details?  Do they overread, under read, or are they just "inaccurate".  Plus, what types of multi-meter are good for the job?  I noted earlier that Archimago seemed happy with a similar use of a digital multi-meter, but I also not he is suggesting using a 220Hz tone, somewhat lower than 10kHz.  So, are digital multi-meters maybe OK up to 300Hz, then inaccurate beyond?  Or is Archimago wrong to suggest a digital multi-meter for even 220Hz?  Does anyone have any decent information with regard to this?  I did a bit of searching on line, the information I found is not great, but it does look like some digital multi-meters will struggle at about 10kHz, but should be reasonably good below that.  From what little I have found, it looks like Archimago's 220 Hz test should be OK, but over 1000Hz it depends on the specific meter design.  As I said, the information I can find is not that good, if anyone has some decent information, I would be interested.

Anyway, no tests are planned today - Time for some music ….. 

[thumb5]

Thinking about my own results a little more, I got to just under 6 watts with SPL's in the mid 90's, at 0dB gain.  Not that loud or that much power.  Considering the 10kHz tone was at about -18dB, these numbers actually look about right.  Certainly not super accurate, but in the right ball park.  So in theory, if the gain is then increased to +18, or a full 0dB test tone is used, then mathematically this would need about 380 watts to produce the tone at 113dB. 

It would make sense to me if the meter on the front panel displayed peak level rather than RMS, in which case a reading of -18 dB would mean the test tone was at -21 dB (RMS).  If that assumption is right, scaling up from the ~6 W you measured would mean about 750 W output into 4 ohms for 0 dB (RMS) input.  In the right ball park, as you pointed out.

Anyway, no tests are planned today - Time for some music ….. 

Good plan

[Jean-Marie]

For digital multi-meters, there is an ADC involved. Now to read the RMS value of an AC signal, you have the choice between two options: either you sample with a high enough frequency to be able to do the calculation in the digital domain, or you use a electrical circuit that does the integration for you and read the outcome of that circuit.

Such a circuit it typically act as low pass filter with -20dB per decade and therefore is only accurate around the frequency it is designed for.

If I were to design a multi-meter I would center my design around 55Hz to cope both the 50Hz and 60Hz AC you can encounter.

10kHz would be more that two decades past 55Hz, therefore I doubt the readings would be really meaningful.

On the other hand if I was doing it in the digital domain, typical integration in the digital domain need at minimum 4 times the sampling frequency as the higher signal they integrate.
That would mean 240Hz for an 60Hz AC signal and 60kHz sampling rate for a 15kHz signal.
Unless this is a very expensive meter, I doubt that anyone would have a design at 60kHz then a design at 240Hz would do the job.

My two cents.... for the moment.

[thumb5]

Just as a data point, I have a Fluke 79 Series II DMM which has AC voltage and current ranges. According to the user manual, these are guaranteed accurate to +/- 1% from 45 Hz to 1 kHz and are typically within +/- 1.5 dB up to 20 kHz.

[Confused]

Amir has reached out to Devialet for a response.
There are questions that could easily be cleared up but I wonder if they will be as woefully uncommunicative on this as they are with pretty much everything else….

Since the device is not under warranty, does Devialet really need to respond? How many other manufacturers would respond to similar requests?

Someone on ASC posted the following links for 170 measurement. It's the same hardware. Registration is required to read these report.
This seems to be pretty comprehensive tests, but done only at 1khz.
http://www.milleraudioresearch.com/downl...t_170.html
http://www.milleraudioresearch.com/downl..._lpcm.html
http://www.milleraudioresearch.com/downl...0_usb.html

On the other hand, its about time for Devialet to explain if their units are tested for full power using multiple sign waves. Many people do refer to Stereophile's measurement of original D-premier. I thought that would have been taken care of in last 8 years.

These are good links from @sam1000

For anyone interested, the registration that is required to view the measurements is free, click the "registration" button per the link below:

http://www.milleraudioresearch.com/avtech/index.html

(note the comment: Javascript must be enabled in your Browser to activate the Registration button.

Measurements for the D-Premier can also be found there:  (Nearly 10 years old now.....)

http://www.milleraudioresearch.com/downl...nput).html

http://www.milleraudioresearch.com/downl...nput).html

I know Paul also tested the O'dA, I would have been interested to see these but unfortunately the website only seams to show measurements up to 2014, the Od'A review was March 2016:

http://designwsound.com/dwsblog/2016/03/...hare=email

Mindful of the recent posts here, I am sure we can all agree that Paul Miller has a touch more expertise with audio measurements than I do.  When hunting for other things I found this, it looks like Paul uses Devialet amplification in his own listening room.  (Devialet for 2 channel, Krell for 7 channel AV)  So whatever he has found in the measurements, it seems Devialet is good enough for him.

https://www.hifinews.com/content/paul-miller