Measuring noise and SINAD of MM phono preamps properly

[Mnyb]

The ubiquitous Ortofon OM series ? OM-10 was everyone's MM-cart back in the day or is the AT91's equivalent circuit middle of the road enough to represent the "MM cartridge" .


The OM series is a definite contender, but in sheer numbers and adoption, it's impossible to come up with a brand other than AT that dominated and set the standard more.

The beauty of AT, especially now, is their cartridge bodies exhibit an incredibly tight tolerance for L and DCR/R. Even more so now they are coming out of China. My father has perhaps a few hundred or more current ATs he's bought in the last few years (he has hundreds of MM carts) and reports they are all very close in DCR, particularly coil to coil and overall consistency.

Much better than the bad old days of Shure etc.

OT: Looks like they (AT) are going to be the survivor of this dwindling niche, all'l buy from them if I ever wanted vinyl playback again .

 

[Michael Fidler]

Shall I do a quick shootout between some common 8 pin DIP op-amps, then? I've definitely got the right kind of test fixtures

 

[pma]

As you wish, but it can be absolutely reliably told from datasheets. And it can be absolutely reliably simulated in Microcap or LTSpice. For MM, all you need is a JFET opamp with low input voltage noise. They have negligible input current noise. Opamps with bipolar inputs will always give worse results in noise because of the input current noise. Like that Fosi X5 with OPA1612. The best possible results are gained with discrete JFET inputs and also with paralleled low noise bipolar transistors. In case of bipolar input opamps, they have to be put in parallel as well to get low noise with MM. Testing with shorted input or 20 ohm AP generator gives false priority when considering real life operation.

 

[drmevo]

The ubiquitous Ortofon OM series ? OM-10 was everyone's MM-cart back in the day or is the AT91's equivalent circuit middle of the road enough to represent the "MM cartridge" .


The OM series is a definite contender, but in sheer numbers and adoption, it's impossible to come up with a brand other than AT that dominated and set the standard more.

The beauty of AT, especially now, is their cartridge bodies exhibit an incredibly tight tolerance for L and DCR/R. Even more so now they are coming out of China. My father has perhaps a few hundred or more current ATs he's bought in the last few years (he has hundreds of MM carts) and reports they are all very close in DCR, particularly coil to coil and overall consistency.

Much better than the bad old days of Shure etc.

I am dummy when it comes to everything else being discussed in this thread, but if you are going for current common cartridges used by the masses, I would suggest either the AT VM95 series (my preference) or Ortofon 2M series. Ortofon 2M red/blue are like the McAfee Antivirus of bundled cartridges, though AT may not be far behind in terms of numbers (and I think they are better-performing for the money).

 

[pma]

Based on some a little bit irrelevant comments on "hum" 50Hz components, which in fact almost does not affect the noise integrated flat over 22Hz - 22kHz range, I have improved screening of my test setup (mumetal above pre input, cartridge placed into shielded little box) and eliminated the hum components. The resulting noise with Shure M35X is almost unchanged

The difference between the bipolar ADA4898 and JFET OPA627 noise and SINAD remains unchanged and everything written in post #1 remains valid.

For those who wish to argue with SINAD results please study first the definitions that have also been provided in post #1. The conclusions are unchanged, measuring noise and SINAD from 20 ohm generator or with shorted preamp input gives valid results for such setup, however invalid results under real world conditions with MM cartridge regarding the noise.

Note: it is almost unimportant which MM cartridge is used for testing. The only important point is that it has inductance of several hundreds mH and this inductance is a key point in revealing contribution of amplifier current noise (and cartridge impedance // 47k thermal noise) to resulting noise and SINAD. Contribution that is hidden with usual 20 ohm AP generator testing. This has been known as long as I am involved in audio, and this is since late seventies of the previous century.

 

[pma]

More opamps, more noise curves:

Output voltage noise, flat 22Hz - 22kHz BW, preamp gain 40dB/1kHz

1: OPA627 (JFET) Vn = 71.73 uV
2: ADA4898 (bipolar) Vn = 210.3 uV (input current noise 2.4pA/rt(Hz), voltage noise 0.9nV/rt(Hz), very low)
3: LM6171 (bipolar) Vn = 157.5 uV (input current noise 1pA/rt(Hz), very good for bipolar. Voltage noise 12nV/rt(Hz), high)
4: 4: OPA134 (JFET) Vn = 92.02 uV
.... and surprise:
5: OPA177 (bipolar, slow) Vn = 96.73uV

The "best" opamp with respect to voltage noise, ADA4898, gave the worst noise record with Shure M35X. Both JFET opamps have better results than their bipolar counterparts, due to negligible input current noise. From bipolars, slow OPA177 had lowest noise. But was not tested for RIAA accuracy.
Sometimes people say that their "cartridge is noisy". It is not the cartridge, it is the preamp, due to design and components used.

 

[morillon]

rather "obsolete" in terms of availability, the aop 627 has become rather expensive... but this was indeed seen in quite a few prephono projects.... ;-)
very clearly with the impact of sites like here and the observation of Chinese DACs and the race for """sinad at 1khz".. the 1612 has become a highlighted element..."""very seller"""...
like the presence now of dip8 to be able to "play the opa's "...
;-)

 

[pma]

Yeah, unfortunately. OPA1612 would be probably good for MC but definitely not for MM preamp. OPA627 is obsolete and expensive, I agree. However, still technically excellent. I am sure that good replacement with same resulting parameters would be OPA165x series.

https://www.ti.com/lit/ds/symlink/opa1656.pdf

 

[DSJR]

The ubiquitous Ortofon OM series ? OM-10 was everyone's MM-cart back in the day or is the AT91's equivalent circuit middle of the road enough to represent the "MM cartridge" .


The OM series is a definite contender, but in sheer numbers and adoption, it's impossible to come up with a brand other than AT that dominated and set the standard more.

The beauty of AT, especially now, is their cartridge bodies exhibit an incredibly tight tolerance for L and DCR/R. Even more so now they are coming out of China. My father has perhaps a few hundred or more current ATs he's bought in the last few years (he has hundreds of MM carts) and reports they are all very close in DCR, particularly coil to coil and overall consistency.

Much better than the bad old days of Shure etc.

The OM10 used to be a very cheap cartridge and the 5e a giveaway with cheap decks (we routinely offered and sold an OM20 stylus at £25 to upgrade the sonics of the Dual 505 deck with Dual/Ortofon OM10 equivalent. The current 2M Red isn't, although the Technics SL1500C comes with one I believe (the last remaining OM in the UK is the Super OM5e, now at £70 or so (YIKES!!!), with slit-poles offering a slightly smoother top over the standard OM body.

The AT95E was a hugely popular cartridge (we sold them from OEM batch-trays rather than boxed and mine has a well glued )) but well finished elliptical tip which puts almost all the cheaper Shure 'nails' to shame

 

[lini]

(...) Yes the cartridges are different, but the lowest inductance I am aware of is 330mH (Shure V15) (...)

Actually there are/were quite a few with quite a good bit lower inductance - like for example in the range below 100 mH AT's AT22/23/24/25 & Signet TK9/10/100 family models and also several Technics models (e.g. the EPC-250C-IIL or the EPC-100C family models). All having a rather low output voltage, of course (*).

Greetings from Munich!

Manfred / lini

*) Not talking about the true low output MMs or MIs, that are/were meant to be used in combination with MC inputs, though - as offered by Elac and Stanton/Pickering back then or by Grado and Soundsmith today. Oh, and there also used to be a true LOMM by AT - namely the very rare AT15SaL.


edit: missing "." added.

 

[morillon]

Actually there are/were quite a few with quite a good bit lower inductance - like for example in the range below 100 mH AT's AT22/23/24/25 & Signet TK9/10/100 family models and also several Technics models (e.g. the EPC-250C-IIL or the EPC-100C family models). All having a rather low output voltage, of course (*).

Greetings from Munich!

Manfred / lini

*) Not talking about the true low output MMs or MIs, that are/were meant to be used in combination with MC inputs, though - as offered by Elac and Stanton/Pickering back then or by Grado and Soundsmith today. Oh, and there also used to be a true LOMM by AT - namely the very rare AT15SaL

you are pointing at very old cartdriges...
at 95 5**, om 10 etc, ortofon red/blue, nagaoka 110 etc are the best sellers of modern affordable mm... ;-)

 

[Mnyb]

Based on some a little bit irrelevant comments on "hum" 50Hz components, which in fact almost does not affect the noise integrated flat over 22Hz - 22kHz range, I have improved screening of my test setup (mumetal above pre input, cartridge placed into shielded little box) and eliminated the hum components. The resulting noise with Shure M35X is almost unchanged

View attachment 426633

The difference between the bipolar ADA4898 and JFET OPA627 noise and SINAD remains unchanged and everything written in post #1 remains valid.

View attachment 426640

For those who wish to argue with SINAD results please study first the definitions that have also been provided in post #1. The conclusions are unchanged, measuring noise and SINAD from 20 ohm generator or with shorted preamp input gives valid results for such setup, however invalid results under real world conditions with MM cartridge regarding the noise.

View attachment 426634


Note: it is almost unimportant which MM cartridge is used for testing. The only important point is that it has inductance of several hundreds mH and this inductance is a key point in revealing contribution of amplifier current noise (and cartridge impedance // 47k thermal noise) to resulting noise and SINAD. Contribution that is hidden with usual 20 ohm AP generator testing. This has been known as long as I am involved in audio, and this is since late seventies of the previous century.

Thanks then I know got it any MM cart has 100’s of mH so not particularly important, back on topic …

 

[pma]

As I repeatedly stated in the X5 Phono thread, mains hum dominate the noise and hence SINAD in actual testing. This is seen in your data:

Net, net, you have not shown why SINAD changes by using a cartridge instead of AP output impedance. Answer is that it doesn't due to rising noise floor at low frequencies and existence of mains hum in almost all cases.

Oh no. SINAD changes by using a cartridge because of total noise integrated through the 20kHz or 22kHz audio band. I hope that you will get it. In previous posts here above the mains hum lines were eliminated by better shielding and the total noise change was only of 1 - 2%.

After better shielding of my setup I have also prepared a SINAD measurement by your method (low generator impedance) and use 32k FFT as you do. The result is below

85 dB SINAD would take bronze medal in your SINAD phono pre chart. Unfortunately, with the cartridge, SINAD drops to 67.5dB, 17.5dB less than with the generator, due to 210uV noise with cartridge in 22Hz - 22kHz range. The elevation of LF noise is not the "hum", but the result of RIAA EQ gain curve of the amp.

-----

Please, for once, try to understand the difference between noise voltage density (the points you see in the noise floor with amplitude for 1Hz bandwidth) and the noise over some frequency range like 22Hz - 22kHz, which we get by integrating the noise density from 22Hz to 22kHz. This integrated noise is then used to calculate SINAD as the N component. Equations in post #1 of the thread. Understanding this will prevent many misunderstandings in future. Please also take into account that frequency axis has logarithmic scale. The seemingly wide low frequency portion has in fact little impact to the total noise integrated over 22kHz range. It is the part from 1kHz to 20kHz that has much higher impact to integrated noise than the 20Hz - 1kHz part. And this "hill" between 2kHz and 10kHz due to ADA4898 current noise through cartridge//47k makes that high total noise. Total noise is 210uV, so only if LF spectrum components would exceed high above 100uV they would affect the total noise. And they do not, they are below 10uV. Integration of frequency dependent noise density curve makes the total noise with cartridge. And you cannot see it with the usual generator method.

One more image - @amirm I deeply hope that this would explain the issue with the bipolar input opamp and MM cartridge. Green plot is 1kHz measurement from 5mV/50ohm generator. Violet plot is noise at amplifier output with cartridge in the input. As a night and day, with the cartridge the noise above 1kHz is highly elevated and SINAD is thus degraded from 85dB with the generator to 67dB with the cartridge. Both plots are in dBV and same settings of measurement.

 

[lini]

you are pointing at very old cartdriges... (...)

Yes, I'm aware of that. But the V15 (type V), which pma (presumably) referred to as the MM with the lowest inductance, that he was aware of, is anything but a new cartridge either.

And of course my examples weren't supposed to mean, that I'd vote for a lower inductance to be applied for MM input SNR measurement purposes. In fact I'd rather vote for using just the same input termination dummy values as already suggested in the good, old IHF-A-202 standard.

Greetings from Munich!

Manfred / lini

 

[pma]

In fact I'd rather vote for using just the same input termination dummy values as already suggested in the good, old IHF-A-202 standard.

That's it. To make a standard R + L dummy load, put it into well shielded box, and use it for noise measurements or even in series with a low impedance generator. As per the IHF standard that you have mentioned and that we used at least 30 years ago. The standard originated in 1966, AFAIK. It is sad to see how analog audio knowledge is diminishing and becomes a lost knowledge.

New IHF Amp Standard (June 1978)

Furthermore, the S/N rating called for in the new standard is based upon a terminated rather than upon a shorted input.

The terminations are as specified before except that moving magnet phono inputs will be terminated with an R-L-C circuit that simulates the source impedance of fixed-coil cartridges.

 

[DVDdoug]

If I cared about vinyl my main concern would simply be consistency (which we have with Amir's measurements) so we can know if one preamp is better than another, and the noise should be referenced to a signal or gain (which we also get from Amir) so that a lower gain preamp doesn't automatically "measure better".

 

[pma]

If I cared about vinyl my main concern would simply be consistency (which we have with Amir's measurements) so we can know if one preamp is better than another, and the noise should be referenced to a signal or gain (which we also get from Amir) so that a lower gain preamp doesn't automatically "measure better".

This is exactly what we do not have from the MM phono preamp measurements published on the homepage here. Respective we have consistency with low impedance generator as a source but no consistency with a cartridge as a source which is a real life situation. It is not about lower or higher gain preamp. It is about preamp input noise current that can make phono preamp that measures better with the generator in fact much worse with the cartridge, than another amp that is worse with the generator. This is what I am trying to explain through the whole thread and I am afraid this effort would be often hopeless.

 

[AnalogSteph]

Actually, CEA-490-A-R2008, the successor to IHF-A-202, only states

4.6.2 Each MM-Phono Input
The input termination for each MM-phono input shall consist of a 1 k ohm ± 10% resistor.
4.6.3 Each MC-Phone Input
The input termination for each MC-phono input shall consist of a 10 ohm ± 10% resistor.

1k is better than nothing but realistically still inadequate.

My go-to values for simulation have been 750R + 450 mH based on an Ortofon OMP10. In any case, the order of magnitude should be clear.

 

[amirm]

Please, for once, try to understand the difference between noise voltage density (the points you see in the noise floor with amplitude for 1Hz bandwidth) and the noise over some frequency range like 22Hz - 22kHz, which we get by integrating the noise density from 22Hz to 22kHz.

It is you that needs to understand the argument. Once more, it makes no difference if the noise*bandwidth goes up when mains noise, is the dominant factor. The mains noise will set the overall noise floor against which the SINAD is computed relative to signal. You can trivially test this by varying the mains noise and watching the massive impact it makes on SINAD. In contrast, high frequency noise does not since it is so low in level to begin with, and much lower than mains noise. Here is an example:

Notice how SINAD is is 60 to 68 dB which quite closely matches the mains peaks. All that noise in treble region including the many spikes are literally in the noise and ignored in computation of SINAD.

You also keep missing the point that you can't measure SINAD per above using a cartridge on the input. You can make noise measurements but not SINAD. SINAD requires injection of that 1 kHz noise. If your source has < 50 Ohm impedance, it would dominate and it doesn't matter what you do with the cartridge in the middle. You can attempt to manually compute it but that requires understanding the above factor which you still don't get. And at any rate, would not be presentable.

 

[amirm]

Yes of course they are! You do understand that a large FFT is required if you want accuracy in the lowest frequencies?

First, we are not looking for accuracy in low frequencies in this discussion. Second, 32K points gives you near 1 Hz resolution already.

But these are not the main issue. The issue is that you seem to be saying, "look at how low the noise floor is" but such a statement cannot be made without looking at impact of FFT gain. Your 128K FFT divides the noise buckets by four, resulting in low numbers on the graph when in reality that doesn't happen.

Just because you decided it was a time saver for testing, doesn't mean it becomes an official world 'gold standard' does it?

Again, it has nothing to do with that. I have shown my measurements with a set of parameters that are clear on the graph. If you want to brag about another phono stage being better, you must use the same measurement parameters. If you can't remeasure, you need to note and compensate for differing measurement parameters which have a direct impact on the point you are trying to make.

And no, 32K is not used because it is a "time saver." I have picked that because at 44kHz sampling, you get an update in less than one second. This is a balance between needing real time updates in the dashboard and having sufficient resolution to see all the interesting signals in the noise floor. In other words, there is wisdom behind the 32K selection.

In non-dashboard cases such as FFT and Multitone, I use far larger FFTS and more averaging, but not here.