An interesting article from the AES on noise in triodes:
AES E-Library Noise in Triodes with Particular Reference to Phono Preamplifiers
Cheers
Ian
AES E-Library Noise in Triodes with Particular Reference to Phono Preamplifiers
Cheers
Ian
Speaking to the 2nd from last paragraph in the article, in which active CCS is dismissed as noisy enough to effect SNR, I examined the noise of the 2 Depletion MOSFET CCS in SY's Impasse preamplifier http://www.diyaudio.com/forums/pass-labs/136835-impasse-preamplifier.html
I would like to see the author's statement backed up with data.
I would like to see the author's statement backed up with data.
Last edited:
I think SY might confirm that comment in some specific instances depending on application. I don't use them in phono stages for this reason. (Truthfully I don't use them much otherwise either, I likes me distortion.. jk 
)
Hmm, I might have to join AES so that I can get access to the articles relevant to my own interests.
)Hmm, I might have to join AES so that I can get access to the articles relevant to my own interests.
Yes, they're problematic with high plate resistance tubes. I've got three preamps in house with low rp triodes and the noise contribution from the CCS is completely dominated by the tubes' own voltage noise.
I've packed up one of the cascode DN2540 for Merlin to measure for noise- I'll try to overcome my activation barrier and get to the post office if we don't get the promised blizzard tomorrow...
So you have an activation barrier too, huh?
Our next "blizzard" is scheduled for Tuesday, wondering whether it will be as lame as yesterday's and the preceding one.
I couldn't remember the specific instance where CCS noise became dominant, but high rp tubes make sense since the CCS noise current appears across the larger rp and becomes dominant over the tube's inherent voltage noise. The only high rp device in my system is the pentode in the front end of one of my two phono stages, lousy psrr, very quiet B+ supply.
Our next "blizzard" is scheduled for Tuesday, wondering whether it will be as lame as yesterday's and the preceding one.
I couldn't remember the specific instance where CCS noise became dominant, but high rp tubes make sense since the CCS noise current appears across the larger rp and becomes dominant over the tube's inherent voltage noise. The only high rp device in my system is the pentode in the front end of one of my two phono stages, lousy psrr, very quiet B+ supply.
Thinking of the simplest CCS, consisting of a single transistor and a resistor in the emitter circuit, it can be shown that provided the gm is not puny (and it never is with transistors), the noise current generated in the resistor is reproduced at in the collector with almost no attenuation. In other words, a CCS is at least as noisy as the resistor it amplifies, and that's NOT yet including any noise generated by the transistor itself.
Now, suppose you're using a 100R resistor. The CCS will produce 1.8nA with a noiseless transistor. Obviously it will be more in reality.
Now suppose that the tube has an equivalent input noise voltage of 0.7uV (realistic) and a whopping gm of 2.5mA/V (yeah, yeah, you can do better, but not trivially). The noise voltage is translated to an anode current of 1uV*gm = 1.8nA. Hence the tube noise is the same as the CCS noise, so the CCS is making the noise factor 3dB worse. Not insignificant in a phono amp.
(Plate resistance is not a factor, since it is common to both sources).
OK, this was a contrived example. But when you factor in transistor noise too, I find it hard to believe that the CCS would be blameless very often. But I am willing to be proved wrong!
There are usually a half dozen articles every year that are of interest to me, and for a fee you get access to the library. You do get invited to a lot of local meetings, at least here in the NYC area. I would imagine that there are a lot of activities in the Boston area as well.
There was a follow-up to Walt's current source articles of 4/07 two years later in AX 4/09. which he crafted a bias multiplier for the depletion MOSFET CCS:
An externally hosted image should be here but it was not working when we last tested it.
He also changed his measurement setup a bit, increasing Rload to 10R and the voltage source to 2V RMS.
It's probably a topic for another thread. I really enjoyed the article Merlin! (The next trick is to incorporate sigma and kappa into the SPICE models.)
I took the 6SN7GT input stage with the cascode depletion MOSFET, Jung's enhanced depletion MOSFET CCS, a cookbook 470K/7K5 input stage and a 470K/2 red LED input stage, adjusted for the gain and created this chart of noise (nV/RtHz) vs frequency for the various schemes. Simplicity is wonderful as Merlin suggests:
An externally hosted image should be here but it was not working when we last tested it.
This is obviously not a standards grade measurement setup as there's a lot of bleed from the filament supply of my Heath power supply. Clearly the enhanced depletion MOSFET CCS has some impedance issues when the cathode LEDS are bypassed.
Does it strike you as odd that the setup with the highest effective plate resistance shows the lowest noise?
Is the vertical axis V/rt Hz rather than nV/rt Hz? I would guess that the current through the tube is quite low- perhaps it might be worthwhile to shunt some current through the LEDs to see the effect of noise. At under 1mA, one would expect the LED dynamic impedance (and presumably noise) to be rather high.
Is the vertical axis V/rt Hz rather than nV/rt Hz? I would guess that the current through the tube is quite low- perhaps it might be worthwhile to shunt some current through the LEDs to see the effect of noise. At under 1mA, one would expect the LED dynamic impedance (and presumably noise) to be rather high.
Mea culpa --
Should have mentioned that I built two depletion mosfet ccs set for 8mA, one per impasse, one as Jung suggested. These used 350 V in and a 12K5 3W resistor. The 470K/7K5 used 300V as suggested in the GE datasheet.
The measurement amplifier has a 10Meg input impedance, but it still loads the output.
I normalized the output for the gain of each amplifier -- for the depletion mode CCS it's 26.2dB, for the 470K/7K5 it's 23.9, substituting 2 red leds for the 7K5 its 23.7dB. There's a 40dB low noise amplifier in the chain. Will repeat over the weekend. I think the enhanced depletion mode should do better than what is illustrated.
The filter factor is SQRT(0.2316) for the analog bandpass filter.
The filter factor is SQRT(0.2316) for the analog bandpass filter.
Took out a lot of the clutter -- only two schemes -- this time I used the Heath HV supply to drive the Maida variant which is part of the Impasse design. Housed everything in a Scots' biscuit tin:
Method:
An externally hosted image should be here but it was not working when we last tested it.
Method:
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
I too was surprised by this assertion, not least because any CCS in the plate would have its noise reduced by the stage gain when referred to the input.
I have also measured the noise of lots of triodes both in CC and mu follower connections and not been able to measure and noticeable difference between them for a given tube.
Cheers
Ian
It helps if you think about the output noise current of the CCS, and tube, rather than noise voltages. That way you don't have to consider the effect of loading, since both currents flow in the same impedance. It then comes down to a simple argument: if the CCS produces more noise current than an anode resistor, then it is making the SNR worse.
(Indeed, since the CCS is a current source it makes little sense to try and analyse it in terms of noise voltages, whatever the situation).
Consider, the voltage gain is:
A = gm*R
If the valve generates a noise current i1, and the CCS or load resistor a noise current i2, then the noise voltage at anode is:
N = (i12 + i22)0.5R
Therefore the SNR is:
A/N = gm/((i12 + i22)0.5)
And it is my assumption that a CCS must always be noisier than an equivalent anode resistor...
Last edited:
Hi Merlin,
Any practical current source that I know of is noisier than an anode resistor, but that doesn't necessarily mean that it is noisier than the valve. The example that you give of a current source with only 100 ohm degradation resistance is simply a poor design, I would tend to make the resistance >> 1/gm,valve and use an active device with high current gain.
I don't understand why you put "with particular reference to phono preamplifiers" in the title. The article doesn't say anything about the effect of RIAA weighting, and its results are useable for any amplifier that is sensitive to the valve's voltage noise.
Anyway, thanks for the nice article! It is nice to have low-frequency noise parameters available for all of the most common triodes, as well as an indication of sample-to-sample spread.
Best regards,
Marcel van de Gevel
Any practical current source that I know of is noisier than an anode resistor, but that doesn't necessarily mean that it is noisier than the valve. The example that you give of a current source with only 100 ohm degradation resistance is simply a poor design, I would tend to make the resistance >> 1/gm,valve and use an active device with high current gain.
I don't understand why you put "with particular reference to phono preamplifiers" in the title. The article doesn't say anything about the effect of RIAA weighting, and its results are useable for any amplifier that is sensitive to the valve's voltage noise.
Anyway, thanks for the nice article! It is nice to have low-frequency noise parameters available for all of the most common triodes, as well as an indication of sample-to-sample spread.
Best regards,
Marcel van de Gevel
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.