• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. 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!

inconsistent IMD measurements in Hypex data sheets and a look at Amir's measurements

capslock

Senior Member
Joined
Jun 19, 2020
Messages
316
Likes
145
I was trying to establish the pecking order of nc400, nc500oem and ncxxxmp from the data sheets. THD+N are difficult because you get a mixed bag of HD, noise and EMI.

I thought IMD might be a nice way to establish whether the amp has sufficient loop gain at near 20 kHz but it turns out the plots are a bit of a mystery and highly inconsistent.

Let's take the nc250mp for starters. IMD at 18.5 and 19.5 kHz at P_R/2 at 4R load.
- the two signal peaks sit at about -3dB (whereas other hypex data sheets have them at -6 dB, and Amir has his at 0 dB - this is all convention, but you have to pay attention when comparing)
- the noise level is at about -134 dB or -131 dB relative to those signal peaks
- both 2nd order products (17.5 and 20.5 kHz) sit at -105 dB (relative to peaks from now on) -- usually the 20.5 kHz would be higher because of degrading loop gain
- 3rd order (16.5 kHz) at -120 dB
- 4th order (15.5 kHz) at -128 dB, higher orders not discernable
- there is a -115 dB peak at 8.25 kHz - what is this?
- there is a DC peak at -91 dB - does this have any meaning, such as in CMRR?
- the first order differenc tone at 1 kHz sits at -104 dB
- 2nd at -119
- 3rd at -126

Now something is very strange in the nc252mp data sheet, same PR/2 at 4R:
- the two peaks sit at - 6dB, so this is a different level that I will have to use as a reference
- noise is at -132 dB
- instead of the expected peak at 17.5 peak there is a doublet of two peaks at 17.75 and 18.25 kHz at about -106 dB
- instead of the expectd 20.5 kHz peak there is a doublet at 19.75 and 20.25 kHz at -104 dB
- there are three more peaks at 17.25, 16.75 and 16.25 kHz in decending magnitude
- the spectrum between 4 and 16 kHz looks very clean, just like the nc250mp except that it had a peak at 8.25 kHz
- the DC peak and some stuff that might be the difference tone and its 2nd and 3rd oder are there but at the wrong frequencies

--> This is the only Hypex data sheet that has the products at the wrong frequencies. Did they misuse their analyser or is this a spurious mode that the module can get into?

Taking Amir's review of the IOP with an nc252mp module, this looks more like the nc250mp data sheet.

His IMD is at 5 W (into 4 R?) and with 19 and 20 kHz tones.
- peaks sit at 0 dB
- noise at roughly -138 dB, i.e. 3-4 dB lower than hypex at 200 W (why?) but with a jitter-like "podestal" around the test tones
- 2nd order at 18 and 21 kHz at -98 dB (strangely this is 7 dB higher than the nc250mp data sheet says - how can the amp be more nonlinear at 5 W than it is at 200 W?)
- 3rd order at 17 kHz at -109 dB (11 dB higher than nc250mp)
- 4th order at 16 kHz at -116 dB (12 dB higher than nc250mp)
- no higher orders, same as nc250mp data sheet
- 2nd to 7th order from 21 kHz up pretty uniformly at -106 dB (this is the declining loop gain - I like Amirs wideband measurement better than hypex's)
- no 8.25 kHz peak!
- no DC!
- 1 kHz at -91 (13 dB higher than nc250mp data sheet)
- 2nd at -108 (11 db higher)
- 3rd at -117 (9 dB more)

IMD according to data sheet for nc100hf for PR/2 at 8R is very similar to nc250mp in that it has mostly 2nd and 3rd order distortion and a tiny bit of 4th. However, the 8.25 kHz peak is missing, and 2nd order seems to be higher.

IMD according to data sheet for nc122mp at PR/2 at 4R is very similar to nc250mp. The 8.25 kHz peak is present. Third and fourth order products seem a lot lower than nc250mp.

In summary:
- The nc252mp data sheet is the only outlier with strange frequencies of the distortion products.
- This behavior is not replicated in Amir's test, but Amir measures lower noise and higher low order products at much lower power.
- The 8.25 kHz is present in most hypex data sheets.
- The DC peak is present in all hypex data sheets but not in Amirs.
- All IMD measurements so far are consistent in that these modules have mainly 2nd and 3rd order distortion and a tiny bit of 4th.

(more to follow)
 
OP
C

capslock

Senior Member
Joined
Jun 19, 2020
Messages
316
Likes
145
From the nc500mp data sheet for PR/2 into 4R:
- peaks at -3dB
- noise at -136 dB relative to peaks (for once, we get lower noise at higher power, as it should be)
- 2nd order at 17.5 kHz -107 dB (2 dB better than nc250mp)
- 2nd order at 20.5 kHz at -97 dB (biggest difference so far, loop gain must be drooping!)
- 3rd order at 16.5 kHz at -109 dB (11 dB worse than nc250mp)
- 4th order at 15.5 kHz at - 116 dB (8 dB worse than nc250mp)
- 5th order at 14.5 kHz at -118 dB (this wasn't even there for nc250mp)
- all harmonics up to 9th order are there, very busy midrange!
- 8.25 kHz peak missing!
- DC contribution is there at -89 dB
- difference tone at 1 kHz at -95 dB (9 dB worse than nc250mp)
- all the harmonics of the difference tone are there up to the 7th

The IMD plot in the nc502mp data sheets seems to be the very same image, down to the last pixel as far as I can tell. One has to wonder how the tabulated THD specs and graphs can be different in that case.

I don't think we have an independent IMD measurement of either module yet?

In summary, the nc500mp/nc502 mp have significantly higher and more harmonics than the smaller modules. The bigger difference between the 20.5 and 17.5 kHz peaks in the large modules is a clear hint that the corner frequency of the loop gain is lower. In addition, overall loop gain may be lower, or the analog stage may be struggling to drive the higher capacitance of the larger MOSFETs.

I have heard it stated that some people don't like the sound of nc400 or 1ET400 and think it is sterile. One of these people just told me he likes nc500mp/nc502mp very much. I suspect that this sound is familiar from older amplifiers and enhances the listening experience.
 
OP
C

capslock

Senior Member
Joined
Jun 19, 2020
Messages
316
Likes
145
So how does the nc500OEM stack up? It is considerably cheaper than the nc400, which is hypex's top performance module, but more expensive than the MP when you count in the power supply and buffer that you have to buy separately.

One should remember that this module has no buffer and hence has 13 dB lower gain. SNR is given at 135 dB and hence a 11 - 17 dB better than for the ncxxxmp modules. Noise voltage is 9 µV typ, 20 µV max compared to 30 - 40 µV for the MP modules. I suspect the nc500oem was tested with some buffer or a higher input voltage but the data sheet does not really say.

IMD at 200 W into 4R:
- peaks at -3dB
- noise at -124 dB, ie. much worse than any of the MP modules (different analyzer setting or different analyzer?)
- 2nd order at 17.5 kHz at -95 dB (this is easily 10 dB worse than any of the MP modules)
- the other 2nd order peak is unfortunately not depicted
- 3rd order at 16.5 kHz at -106 dB (a lot worse than the small MP modules and even slightly worse than nc500mp)
- 4th order at 15.5 kHz at -113 dB (12 dB worse than nc250mp but and 3 dB worse than nc500mp)
- no higher orders (due to noise floor, we cannot say if it is better than the small MPs, but this chain of higher harmonics seen in the nc500mp should be visible)
- no 8.25 kHz peak (should easily come out of the noise floor)
- DC at - 88 dB (similar to all MPs)
- 1 kHz at -96 dB (worse than small MPs but similar to nc500MP)
- no 2nd order at 2 kHz (i.e. better even than nc250mp!)

What a strange beast! Let's look at Amir's review of the Nord One with nc500 and hypex buffer. Again, it is 19+20 kHz at 5 W (into 4R?).

- peaks at 0 dB
- noise at -130 dB and without jitter podestal unlike the review of the nc252mp (but about 8 dB higher than nc252mp - can this really be true? the hypex buffer has much better opamps than the NE5532 on the MP? maybe there is no adequate low pass filter?)
- 2nd order at 18 kHz at -106 dB (i.e. 8 dB better than his review of nc252mp and much better than hypex IMD, but that was 200 W)
- 2nd order at 21 kHz at 101 dB (this is loop gain running out, but not as severely as in the nc500mp)
- 3rd order at 18 kHz at -124 dB (15 dB better than his measurement of nc252mp and 18 dB better than hypex at 200 W)
- all other harmonics are really in the noise
- no 8.25 kHz peak
- no DC peak
- difference tone at -92 dB (1 dB better than his measurment of nc252mp, 3 dB worse than hypex at 200 W)
- 2nd and 3rd order seem strangely broadened and are amid a higher noise floor

In summary, this old hypex IMD measurement is not really comparable to newer hypex measurments (which are also not always consistent). Amir's data at 5 W seem to support the notion that this module is really a lot better than even the small MP series, even if 2nd order is not really stellar.
 
OP
C

capslock

Senior Member
Joined
Jun 19, 2020
Messages
316
Likes
145
Last part of this series, nc400, this is the high end module, and it has a buffer. Gain is the same as the MP modules, noise at 23 µV is higher than nc500 but lower than MP. S/N is just a tad higher than MP, THD+N is better than nc500OEM.

The data sheet has a properly labeled plot for PR/2 into 2 R. The plot to the right of that is not properly labelled but is in all likelyhood PR/2 into 4R.
- peaks at -3 dB
- noise floor at -138 dB, much better than nc500OEM (which was probably a bad setting), better than small MPs and even 2 dB better than nc500MP, the lowest noise part of the MP series
- 2nd order at 17.5 kHz at -100 dB (5 dB worse than nc250mp and 7 dB worse than nc500 mp!)
- 3rd order at 16.5 kHz at -124 dB (4 dB better than nc250mp, 15 dB better than nc500mp and 18 dB better than nc500OEM)
- no (!) higher orders, i.e. at least 10 dB better than 4th order in nc250mp
- no 8.25 kHz
- DC at -108 dB (strangely -100 dB into 2R, so maybe this does have some meaning?)
- 1 kHz at -124 dB (20 dB better than nc250mp)
- 2 kHz at -125 dB (6 dB better than nc250mp)
- 3 kHz at -129 dB ( 3 dB better than nc250mp)
- no higher orders

In summary, according to hypex IMD plots, this amp is quieter and cleaner than any of the other hypex amps, maybe with the exception of the 2nd order at 17.5 kHz (wonder if this could be power supply?).

Amir didn't measure IMD. His 1 kHz HD testing at 5 W was -125 dB 2nd, -112 dB 3rd, -128 dB 5th, and with one channel only, it was -132 dB 2nd and -134 dB 3rd with nothing else. We don't reall know how to compare these results to the IMD plots but they do not necessarily support the view that 2nd order is high in this amp.

Lastly, the noise floor in Amir's measurements was about -144 dB at 5 W, compared to -138 dB for his IMD of the nc252mp and -130 dB for the nc500OEM. If these measurements were done with the same settings, the nc400 is the quietest amp by far.

OK, lots of details with some questions hidden. Enjoy and comment away!
 
OP
C

capslock

Senior Member
Joined
Jun 19, 2020
Messages
316
Likes
145
As a bonus, let's compare to the Purifi 1ET400A. It'd datasheet is a little different from hypex in that it gives us 1 kHz sine and 18/19 kHz IMD into 4 R at 1, 10, and 100 W. Interestingly, the noise level is always lower for the sine measurement that it is for the IMD, even if both are FFTs that show the 0 to 20 kHz range. This becomes as much as 12 dB for the 100 W measurments.

Ok, let's look at the 100 W IMD, which is closest to what is in the hypex nc400 data sheet:
- peaks more like -2 dB (strange)
- noise at -138 db relative to those -2 dB, rises to about -133 dB from 14 kHz on (modulator noise?)
- 2nd order at 17 kHz at -120 dB (i.e. 20 dB better than nc400 at presumably 200 W)
- 3rd at 16 kHz at -137 dB (13 dB better than nc400)
- 4th at a5 kHZ at -138 db (would be in the noise for nc400)
- no 8,25 kHz
- DC at -132 dB
- 1 kHz difference tone at -128 dB (4 dB better than nc400)

We can also compare the 1 kHz 100 W into 4R vs nc400 at 200 W into 4R

- 2nd -131 / -115
- 3rd - 125 / -109
- 4th -137 / -134
- 5th -125 / -122
- 6th -142 dB / -132
- 7th -132 / -122
- 8th - 152 (noise floor) / -135
- 9th -130 / -134

It would appear the the Purifi at 100 W is a lot better yet than the nc400 at 200 W. On the other hand, Amir measured even lower values for some of these harmonics in the nc400 at 5 W with one amp operating only.

Amir's review of the Purifi evaluation amp has no IMD plot.

We can compare the 5 W results (pregain / no pregain / Amir's results on nc400 one amp only)
- 2nd -133 / -135 / - 132
- 3rd - 128 / -129 / -134
- 4th -n.f. / -138 / n.f. (< 140)

In this comparison, the nc400 seems to hold water. I don't really know what to make of Amir's bar graph of harmonics in the 1ET400 test. The bars are a lot higher a than in the FFT, e.g. -120 dB for 3rd harmonic.

Further measurements of Eigentakt modules in NAD products were not half as good.
 

boXem

Major Contributor
Audio Company
Joined
Jun 19, 2019
Messages
2,014
Likes
4,852
Location
Europe
You can't conclude anything about noise from these measurements, because the noise floor is dependant from the amount of samples (fft gain).
For the rest, the fundamentals of the IMD are supposed to be at - 6dB (division by 2), and a table with all your findings would make all this much more readable :)
 
OP
C

capslock

Senior Member
Joined
Jun 19, 2020
Messages
316
Likes
145
Yes, I have run into this issue before that FFT binning and averaging can change the apparent picture. But there should be a point when you really measure the noise spectral density of the DUT (if your software lets you do it).

There seem to be different habits as to how to scale the fundamentals, but as long as we agree to look at voltage, it is just relative. Choosing the height of one fundamental is maybe not the most logical way but ok as long as it is consistent.

I have started a spreadsheet but I found it even more confusing so I decided to highlight the important differences in writing. Maybe I'll try to streamline my spreadsheet at some point and post it here.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,375
Likes
234,488
Location
Seattle Area
I have started a spreadsheet but I found it even more confusing so I decided to highlight the important differences in writing. Maybe I'll try to streamline my spreadsheet at some point and post it here.
Please do. I am too impatient to read all that text in your post as is. :)

Just a random comment then: my IMD tests use SMPTE dual tones of 60 and 7 kHz. They are not 19 and 20 khz tones.

Second, exact comparison of distortion is challenging because they can be load dependent. And even the quality of connectors for speakers matters! In general my results follow the trends as published by hypex, but not in the minor detail.
 

boXem

Major Contributor
Audio Company
Joined
Jun 19, 2019
Messages
2,014
Likes
4,852
Location
Europe
Yes, I have run into this issue before that FFT binning and averaging can change the apparent picture. But there should be a point when you really measure the noise spectral density of the DUT (if your software lets you do it).

There seem to be different habits as to how to scale the fundamentals, but as long as we agree to look at voltage, it is just relative. Choosing the height of one fundamental is maybe not the most logical way but ok as long as it is consistent.
If you know the amount of samples, you can compute the noise from the graph. Proper software directly extracts it from the samples and displays it's value.
For scaling the fundamental, my opinion is that 0dB should correspond to the sum of all fundamental tones. Then the distortion products scale relatively to the total amplitude of the signal. Others think differently.
 
OP
C

capslock

Senior Member
Joined
Jun 19, 2020
Messages
316
Likes
145
Please do. I am too impatient to read all that text in your post as is. :)

Just a random comment then: my IMD tests use SMPTE dual tones of 60 and 7 kHz. They are not 19 and 20 khz tones.

Second, exact comparison of distortion is challenging because they can be load dependent. And even the quality of connectors for speakers matters! In general my results follow the trends as published by hypex, but not in the minor detail.

Yes, I will clean up and post the table.

You had a wideband 19 + 20 kHz two tone IMD test in some of your reviews that is similar to what can be found in the hypex data sheets.

I realize a lot depends on the load, but fortunately, some hypex data sheets have IMD data for 2, 4, sometimes 8 R load, or Purifi has them for 1, 10, 100 W. Things change a little but not that much as long as we are not driving something into clipping or saturation.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,375
Likes
234,488
Location
Seattle Area
There seem to be different habits as to how to scale the fundamentals, but as long as we agree to look at voltage, it is just relative.
I almost never show voltages on my graphs. Instead, I have standardized references I use to set to 0 dBrA. It becomes very difficult to look at results that keep changing based on what voltage is in use.
 
Top Bottom