Measuring the Noise Floors of my ART Tube Preamps

[quesne]

Hi Folks:

I have three ART tube preamps in my rack (a Digital MPA II plus two Dual MPs), as well as a Zoom R16 interface with eight unimpressive preamp channels. I do also have a PreSonus FireStudio Project, but I lent it to a friend last year and I'm waiting to get it back (yes, I am very much a "budget" home recording hobbyist....but this is the gear I've got, and I'm actually pretty happy with it

I have been interested for a while in getting some reliable measurements of the comparative noise floors of my pres, especially to evaluate the various tube performances. I have been developing a testing protocol, and this is basically my baseline test (i.e. no input signal, measuring only the "delta" between the various channels). I thought I would share the data here in the hope of improving my methods, as well as prompting people to post their own noise floor measurements.

I tested a total of eight channels: Two in my MPA II, four in the Dual MPs, and two in the Zoom. The two MPA II channels (#3 and #4 ) have matching Sylvania NOS 12AX7 tubes. The two DMP units have the following tubes: Ch #5 RCA 12AU7 (NOS), Ch #6 TungSol 12AX7 (New), Ch #7 Dynaco Telefunken 12AX7 (NOS) and Ch #8 JJ 5751 (NOS).

The test was pretty simple. For each of the channels, I did the following:

1. Connected an XLR cable to the preamp input, with a Shure SM-58 mic attached and switched off (pins terminated with no signal from the capsule).

2. Added a new track to my Reaper project, and mapped the appropriate input channel.

3. Turned all of the preamp knobs and buttons down to zero (Input, Output and Gain Pad). I also turned the HPF knobs on both MPA II channels to the lowest setting.

4. Armed the track, and recorded for 40 seconds while making the following changes:

a. At the 8 second mark, I turned the output knob to 100%
b. At the 16 second mark, I turned the input knob to 100%
c. At the 24 second mark, I turned the input knob to zero, and I engaged the gain pad switch
d. At the 32 second mark, I left the gain pad engaged and turned the input knob to 100%
e. The two ZOOM R16 channels (Ch #1 and Ch #2 ) have only one gain knob and no gain pad. As such, I made only one change on those two tracks: At the 8 second mark, I turned the gain knob up to 100%

So, for the tube channels, there are five noise floor modes recorded:

1. Base measurement, all settings at zero (0 to 8 sec)
2. Gain Pad Off, Input 0%, Output 100% (8 to 16 sec)
3. Gain Pad Off, Input 100%, Output 100% (16 to 24 sec)
4. Gain Pad On, Input 0%, Output 100% (24 to 32 sec)
5. Gain Pad On, Input 100%, Output 100% (32 to 40 sec)

For the ZOOM R16 channels, there are only two mode:
1. Base measurement, all settings at zero (0 to 8 sec)
2. Gain Input 100% (8 to 16 sec)

I noticed some pretty interesting results here, but I will hold back and let others analyze the files before I chime in. BTW - This was just a first attempt at gathering some good solid data to analyze objectively. If you feel I've made a mistake -- or that my test protocol can be improved -- please let me know

FYI - Since I can't post WAV files, I had to link to files in my Dropbox (below):

CH3 MPA2 SYLX
CH4 MPA2 SYLX
CH5 ADMP RCAU
CH6 ADMP TNGX
CH7 ADMP DTFX
CH8 ADMP JJ51

CH1 ZR16 XLR
CH2 ZR16 XLR

 

[TalismanRich]

My first comment is that all of the ART preamps had 60cycle hum which got progressively worse as you went through the selections. All the the preamps start out with noise levels between -110 and -120dB.

So I looked at the last 8 seconds of each preamp. That's where the total system noise is most apparent.

The MPA 2 had lower hiss than the DMP in all cases. What isn't known is the available gain. If the DMP gets 20dB more gain, then it would offset much of the advantage.

RE the tubes themselves, I rate them 6, 4, 3, 5.. Which would be JJ, TngS, RCA, TF. The JJ and TungSol both had similar levels of apparent "hum". The Telefunken had lots of buzz with lots of higher order harmonics which makes it even more intrusive.

The Zoom preamps were pretty much consistently at -110dB across the spectrum with little if any 60cycle hum.

Here are four spectrum captures. The MPA is the first sample. Then the JJ and Telefunken tubes in the DMP (best and worst), and finally the R-16.

Again, these numbers are a bit meaningless unless you can compare the amount of gain this represents. If the Zoom does 40dB of gain and the DMP does 80dB, then that hum drops from -70 to -110dB in direct comparison. However, of the tubes, I would be choosing the JJ. The Telefunken has roughly 10-12dB more harmonic at 240Hz and 4 or 5dB more at 180Hz. Those harmonics are very apparent where the 60 cycle might be masked by speakers with bass roll off although all would be apparent with a sub.

 

[ecc83]

I agree with Rich, you must know the gain of each pre amp but as I said before, that is not easy to do unless you have some pretty specialized kit.
Also not fair to have a shorted input. The noise performance of an amplifier is sensitive to the terminating impedance and each pre should be terminated with the manufacturer's specified load. Read up on "noise models that deal with voltage and current noise" Different circuits need a certain termination for optimum noise performance. I shall see what I can find.

Save.

 

[TalismanRich]

Just checking the specs on the Digital MPA II, it specifies up to 70dB gain and .005% THD. The Dual MP specifies 70dB gain (XLR to XLR) but 0.1% THD. Both claim -129dB EIN for the XLR inputs. So they are spec'd roughly the same. That doesn't explain the differences between the two. Only the distortion spec seems significantly different.

Zoom specifies -50dBm to +4dBm (dB referenced to 1milliwatt). So it's appears to have a total of 54dB of total gain.

If you're feeding either preamp into the R-16, you'll need to keep the trim at minimum. Both Art pres can put out something like +27dBu. Keep input and output levels down otherwise you'll overdrive the inputs of the Zoom.

 

[quesne]

Thanks very much for your thoughtfulness and contributions to the discussion, Rich. I greatly appreciate your insights!

Recall the considerable gain difference between the tubes, which may account for the disparity. 12AX7 = 100mu, JJ-5751 = 60mu, 12AU7 = 17mu. The gain specs you provided would presume the stock 12AX7 tube.

 

[ecc83]

Thanks very much for your thoughtfulness and contributions to the discussion, Rich. I greatly appreciate your insights!

Recall the considerable gain difference between the tubes, which may account for the disparity. 12AX7 = 100mu, JJ-5751 = 60mu, 12AU7 = 17mu. The gain specs you provided would presume the stock 12AX7 tube.

Ah! Valves. The gain of an individual valve stage is very dependent on the circuit design and "u" is only a part of the picture.

Stage gain is given by A= uRa/Ra+ra. Where Ra is the anode (plate) load and ra is the anode slope resistance of the valve*. But the ACTUAL load on the stage can be complex...there might be a pot in circuit or the stage might be driving a cathode follower and DC coupled and thus the load is close to infinite. Then, what happens in the cathode circuit matters greatly. Normally the cathode R is bypassed with a capacitor but not always and then the stage gain will be MUCH lower.

You never get more than about 60% of the amplification indicated by 'u' An exception to that is a DC coupled pair where the first stage Ra is split and boot strapped.

But then valve mic preamps virtually all use an input, step up transformer giving 'free' noiseless gain. You need to know its ratio. In fact the 12au7 with its much lower ra plus a traff can give good enough gain but the much lower circuit resistance (1/10th the 12aX7) gives about 10dB less noise. Downside is the valve draws 10mA as against ~1mA for the X7.

*A parameter which varies with Va and Ia!

Dave.

 

[quesne]

Thanks very much, Dave! I greatly appreciate the input Based on what you have said, given the plate voltage of each particular unit, is there a reliable way to compute the gain contribution of a particular valve? (I think the starved Dual MP plate voltage is about 48v, whereas the Digital MPA II is about 150v....but I can find out for sure).

 

[ecc83]

Thanks very much, Dave! I greatly appreciate the input Based on what you have said, given the plate voltage of each particular unit, is there a reliable way to compute the gain contribution of a particular valve? (I think the starved Dual MP plate voltage is about 48v, whereas the Digital MPA II is about 150v....but I can find out for sure).

If you have the exact schematics of each stage you could make a fair stab at the gain but it won't be very accurate.

In the absence of the right kit to do it I think I would "measure" gain using a sound source. A speaker reproducing pink noise (Audacity can generate that) and set a specific microphone a standard distance with a phone having an SPL app' ( C weighting) That way you get the same signal voltage into each pre amp*. Then measure the output voltage.

*be slightly affected by the varying input impedances but what can yer do!?

Dave.