Lets assume a 3-tube phono amplifier, 2 stage amplification, all same triodes, same effective amplification factor, passive equalization, final stage buffer.
My question is: where to fit the equalizing network ?
There are two main possibilities for the network - lumped all-in-one, and split.
However there are also two places where to put the network components - after the 1st or after the 2nd stage.
Thats a total of four permutations.
a) lumped after the 1st stage
b) lumped after the 2nd stage
c) split high frequency roll-of after the 1st stage
d) split high frequency roll-of after the 2nd stage
Now, which one gives the lowest noise ?
My question is: where to fit the equalizing network ?
There are two main possibilities for the network - lumped all-in-one, and split.
However there are also two places where to put the network components - after the 1st or after the 2nd stage.
Thats a total of four permutations.
a) lumped after the 1st stage
b) lumped after the 2nd stage
c) split high frequency roll-of after the 1st stage
d) split high frequency roll-of after the 2nd stage
Now, which one gives the lowest noise ?
It depends on how you define noise, but you can't ignore the overload margins to get low noise. The two will trade off against each other.
Last edited:
Noise is largely not a consideration unless there is more loss between stage 1 and 2 than there is gain in stage 1. i.e. if the signal reaching stage 2 ends up smaller than when it left the cartridge! For all-in-one-go you have 20db loss at 1kHz, so you need >20dB gain from stage 1 for adequate noise (ideally 26dB or more). SNR will degrade at high frequencies but be better at low frequencies, but things tend to even out because high freq's are boosted on the vinyl. Split EQ tends to give better SNR so you can get away with as little as 20dB gain from stage 1, but since you attenuate less between stage 1 and 2, you need more headroom from the stages instead.
For split EQ you have to choose between using low-pass in both sections, or high-pass followed by low-pass (or in the opposite order). The latter has the advantage of attenuating low frequency signals before hitting the next stage, so you can get away with less headroom there (since most music exists at low frequencies), and the following 50Hz low-pass also provides good attenuation of noise from the previous stage (though this may not be significant).
For split EQ you have to choose between using low-pass in both sections, or high-pass followed by low-pass (or in the opposite order). The latter has the advantage of attenuating low frequency signals before hitting the next stage, so you can get away with less headroom there (since most music exists at low frequencies), and the following 50Hz low-pass also provides good attenuation of noise from the previous stage (though this may not be significant).
Last edited:
Thank you Merlinb,
however you confused me somewhat when you wrote:
in short, with a relatively low mu tube should the 75us section come first as in the attached block diagram or the 3180us section ?
however you confused me somewhat when you wrote:
because it is not clear to me to which case "the latter" refers to ...
in short, with a relatively low mu tube should the 75us section come first as in the attached block diagram or the 3180us section ?
Attachments
I was referring to this sort of EQ:
Usually you would do it as shown, because the Miller capacitance of stage 2 neatly blends into the 75us capacitor. Since stage 3 is presumably a buffer, it has very little capacitance to worry about.
Doing it the other way around has a headroom advantage because everything above 50Hz is attenuated by up to 20dB before hitting the second stage, but Miller capacitance then affects EQ accuracy.
Last edited:
It seems that the split EQ 3180/318us first and 75us last works best for me. Lowest noise and way lower distortion than the other way around.
The all-in-one EQ in my case is always worse, both noise and distortion, especially when positioned after the 2nd stage, where I put it foolishly in the beginning and which is why I started asking in the first place ...
The all-in-one EQ in my case is always worse, both noise and distortion, especially when positioned after the 2nd stage, where I put it foolishly in the beginning and which is why I started asking in the first place ...
From my experience, the best solution is not to split the passive RIAA.
Because when you split, you must have 3 gain blocks: if left with just 2, you must think on how to avoid the buffer stage (i.e. cathode follower) or just the pot of the line preamp which follows it - not to interfere with the precision of the equalisation.
If you choose to have 3 gain blocks, the basic problem is that whatever is introduced by the 1st tube will be amplified by all the others... you will need to select for noise and micro-phonics... which is sometimes impossible because you do not have enough tubes to start with - or the level required cannot be achieved with the tube type chosen.
Thus, prefering 2 gain stages - I opt for composite RIAA as well. Willing to reconsider?
So far topology has not been discussed and it has some bearing here. For example if the input circuit is a cascode then a composite RIAA circuit after than is a good move as most of your gain precedes it. So as you roll off highs you are rolling off noise too and no worries about overload. Contrary to popular opinion, the output of the cascode need not be particularly high impedance, although I would stay away from 12AX7s if you plan to run zero feedback. In this fashion with only 2 gain stages (a cascode functions as a single gain stage) you can work with low output moving coil cartridges.
I actually do have 3 gain stages, all rather low mu, thats why.
Since you mention cascode input ... I found this Allen Wright article on the web, promoting cascode plus split topology,
http://www.vacuumstate.com/fileupload/SP_15_Article.pdf
Since you mention cascode input ... I found this Allen Wright article on the web, promoting cascode plus split topology,
http://www.vacuumstate.com/fileupload/SP_15_Article.pdf
The trouble with the cascode is it is noisy. The lower device never achieves very much gain, so you have the gain from two devices contributing very early on, which worsens the SNR. It also has negligible PSRR.
Allen Wright's hybrid cascode might work if the source resistor was bypassed, but left unbypassed, it is unspeakably noisy. As far as SNR is concerned, it is very difficult to beat the simple resistor-load common-cathode gain stage. The mu-follower is also OK. The SRPP / half-mu is also too noisy for first stage use.
Allen Wright's hybrid cascode might work if the source resistor was bypassed, but left unbypassed, it is unspeakably noisy. As far as SNR is concerned, it is very difficult to beat the simple resistor-load common-cathode gain stage. The mu-follower is also OK. The SRPP / half-mu is also too noisy for first stage use.
I don't follow this logic. Whatever is introduced by the first stage is always amplifying by the following stages, however many stages there are. All-in-one or split EQ both use two stages of gain, in most cases.
Matter of transconductance in the lower device. If its strong, the contribution of the upper part is very weak since those noise sources are not correlated. Especially in a high Yfs JFET running high current using minimal Rs while topped with a low Rbb' common base BJT scenario, the difference to a straight common source can be minimal. Regarding PSRR in true life rails presence, folding the cascode to ground helps. I use such topology in my FSP phono.
Yes, Mu follower is good too. I use it in my Valve Itch phono. Both preamps never presented hiss problems to the many builders as they had reported in the relevant threads.
Since this is a valve forum we are primarily concerned with valve cascodes, which are noisier than simple common-cathode stages, and cannot be folded.
The PNP would represent a very low load impedance to the tube, so you would still get pathetic gain out of it. The result would still be more noise than if the triode were used alone with a resistive load.
Ultimately, a cascode involving a valve is almost never a low-noise solution, and a cascode with a tube for the bottom device is never a low-noise solution. A transistor followed by a triode can work, but it is usually better (and easier) to cascade rather than cascode them.
Have you listened to an RTP3 or SVP and found it too noisy? I have built both, and the only noise I can hear at all is a faint hum in the single-ended SVP, which I expect to be able to fix with minor alterations to the grounding scheme. Using a fairly standard MC (a 0.5 mV Lyra) in the RTP3, I use his recommendation of no source resistance for low-output cartridges, with no noise penalty.
Alex
Simply listening to noise in the final product is not a good way to gauge the absolute noise level. Whilst it may be acceptable in your listening environment, there are quieter circuits. Whether we should care about such specmanship is, admittedly, a matter of personal attitude. Noise is one of the things that I personally obsess over, perhaps more than some practitioners. For me, 'good enough' is not good enough and hum is absolutely unacceptable. YMMV.
Some cascodes are noisier than others, so some might be worse than pentodes (though this is probably rare). All cascodes involving tubes are noisier than a cascade using the same components. Since their linearity is worse too, I can see no justification for using a cascode in a phono stage.
It gets worse, but is minimised if the EQ is driven with a cathode follower (or mu-follower).
I never built one, nor do I currently intend to, so I don't have any stakes in it, but interestingly Allen Wright said quite the oppsite about his (hybrid) cascode:
"A cascode 1st stage brings several advantages:
... The high output Z means tube selection and/or aging doesn’t affect the 1st
part of the RIAA network—and with both the 3180 and 318µS points there
you don’t want too many changes …
... It is inherently quiet, linear and very fast—which is why Tek & HP used it. "
"A cascode 1st stage brings several advantages:
... The high output Z means tube selection and/or aging doesn’t affect the 1st
part of the RIAA network—and with both the 3180 and 318µS points there
you don’t want too many changes …
... It is inherently quiet, linear and very fast—which is why Tek & HP used it. "
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.