I think you took this diagram and the equations from here (page 16). I hoped this document would provide a bit more context on what Ale Moglia means by "totem-pole current balance", and why gfs = gm. Also, I am not sure if Gm (tube transconductance) and gm are supposed to be the same thing, or if that's a typo. Same questions for Gfs (FET transconductance) and gfs.
- What is "totem-pole current balance" supposed to mean?
- What are gm and gfs in Ales equations?
- Why gm = gfs?
Last edited:
Ok, here goes my take on the mu-stage, and how to choose the Rmu value. I mainly post it here so I can find it again, and maybe someone else even thinks it's useful.
First of all, read Allan Kimmels article on the mu-stage. His description starts out based on a all-tube SRPP circuit, but that may not be necessary here. Consider the hybrid mu-follower circuit as in this diagram:
With no audio/AC signal applied to the input, the gate of FET Q1 is grounded, and so the anode voltage of the tube is biased to a slightly negative DC voltage below GND (as determined by the Vgs value of the FET and the slight voltage drop across Rmu). The DC current flowing through the tube, Rmu and the FET is controlled by the bias volage applied to the grid of the tube.
Once an audio/AC voltage is applied to the input, any dynamic change (dVA) in the voltage at the tube anode results in a similar voltage change (dVS) at the FET source pin according to the voltage gain (A) of the source follower (typically A = 0.99...):
dVS = A x dVA
The corresponding change of the current through the mu resistor (Rmu) is:
dI = (dVA-dVS) / Rmu = (1-A) x dVA / Rmu
The dynamic load resistance of the active load as seen by the tube anode is:
r0 = dVA / dI = dVA / [ (1-A) x dVA / Rmu ] = Rmu / (1-A)
With gfs the transconductance of the FET and A ≈ gfs x Rmu / (gfs x Rmu + 1) and Rmu >> 1/gfs:
1-A ≈ 1 / (gfs x Rmu + 1) ≈ 1 / gfs / Rmu
Therefore:
r0 ≈ Rmu^2 x gfs or Rmu ≈ sqrt(r0 / gfs)
The goal is to make the load line for the tube as flat as reasonably possible. In other words, r0 must be much larger than the plate resistance rp of the tube. r0 = 50 x rp seems like a reasonable point of diminishing returns. For example, consider a mu-stage using a tube with rp = 1 kΩ and a FET with gfs = 1 A/V:
Rmu ≈ sqrt(50 x 1 kΩ / 1 A/V) ≈ 220 Ω
A larger Rmu value would also work, but wouldn't provide any substantial improvements of the audio performance of the mu-stage. A larger resistor means higher voltage drop, more heat, and also more feedback in the CCS, which may give rise to oscillation in the mu-stage.
First of all, read Allan Kimmels article on the mu-stage. His description starts out based on a all-tube SRPP circuit, but that may not be necessary here. Consider the hybrid mu-follower circuit as in this diagram:
With no audio/AC signal applied to the input, the gate of FET Q1 is grounded, and so the anode voltage of the tube is biased to a slightly negative DC voltage below GND (as determined by the Vgs value of the FET and the slight voltage drop across Rmu). The DC current flowing through the tube, Rmu and the FET is controlled by the bias volage applied to the grid of the tube.
Once an audio/AC voltage is applied to the input, any dynamic change (dVA) in the voltage at the tube anode results in a similar voltage change (dVS) at the FET source pin according to the voltage gain (A) of the source follower (typically A = 0.99...):
dVS = A x dVA
The corresponding change of the current through the mu resistor (Rmu) is:
dI = (dVA-dVS) / Rmu = (1-A) x dVA / Rmu
The dynamic load resistance of the active load as seen by the tube anode is:
r0 = dVA / dI = dVA / [ (1-A) x dVA / Rmu ] = Rmu / (1-A)
With gfs the transconductance of the FET and A ≈ gfs x Rmu / (gfs x Rmu + 1) and Rmu >> 1/gfs:
1-A ≈ 1 / (gfs x Rmu + 1) ≈ 1 / gfs / Rmu
Therefore:
r0 ≈ Rmu^2 x gfs or Rmu ≈ sqrt(r0 / gfs)
The goal is to make the load line for the tube as flat as reasonably possible. In other words, r0 must be much larger than the plate resistance rp of the tube. r0 = 50 x rp seems like a reasonable point of diminishing returns. For example, consider a mu-stage using a tube with rp = 1 kΩ and a FET with gfs = 1 A/V:
Rmu ≈ sqrt(50 x 1 kΩ / 1 A/V) ≈ 220 Ω
A larger Rmu value would also work, but wouldn't provide any substantial improvements of the audio performance of the mu-stage. A larger resistor means higher voltage drop, more heat, and also more feedback in the CCS, which may give rise to oscillation in the mu-stage.
Last edited:
So as per your maths for 5K Rp & FET gfs 35mS
Rmu ≈ sqrt(r0 / gfs)
Rmu = sqrt (50x5000/35) = 14,28 Ω????
Rmu ≈ sqrt(r0 / gfs)
Rmu = sqrt (50x5000/35) = 14,28 Ω????
Nah. Pay attention to the unit prefixes:
sqrt( 50 x 5 kOhm / 35 mS) = 2.7 kOhm
sqrt( 50 x 5 kOhm / 35 mS) = 2.7 kOhm
From Ale's earlier postings (2018):
Then:
and:
The above can be read in this post: https://www.diyaudio.com/community/threads/45-amp-build-direct-coupled.320418/page-6
Then:
and:
The above can be read in this post: https://www.diyaudio.com/community/threads/45-amp-build-direct-coupled.320418/page-6
That sounds like as good an explanation as any. At some point thought for many FET's (taking the above example) the ≈ 220 Ω would be a minimum value. 1k is not significantly different, and won't lead to oscillation in the mu-stage in any case.
My experience has been that the resistor quality is critical. I used some fancy carbon ones but ended up swapping them out for some cheap metal film ones that were less noisy (and didn't drift).
...targets the output impedance of the mu-stage (i.e., the audio output). In my post #42 (yes...) I looked at the impedance presented to the tube.
The mu-stage in my amp did oscillate if the Rmu value was too high. See here.
Can you provide some details on the resistor part? How about Vishay/Dale CMF55 types?
Interesting. Ale was also using AOT1N60 and posted something about oscillation here:
https://www.bartola.co.uk/valves/tag/aot1n60/
But he was using it for source follower duty. I never tried this FET. Seriously, 1k Ohm gate stopper for source follower should not make a difference, and I would not expect it to make any difference for the Rmu too. Ale went down to 100 ohm and claimed it cured his fuzz on his source follower. Maybe something else was going on? This FET has decent looking specs.
The metal films I was using were from our local electronics shop in Zurich - Pusterla. They are 2W and have an attactive size. But I discovered that certain values are noisy. They are slighlty more noizy than Kiwame metal films... I am now just using them for power supply ground reference, stuff where their noise can't make it to my speakers...
Vishay/Dale CMF55 are generally pretty good. I have been using CMF60 type for gridstopper/gatestopper for some time and they have been fine. Not as expensive as fancy parts such as the Takman's I like to use. BUT if you are working with tiny signals, then the Amtrans carbon film are also great choice. They are not as expensive as other boutique resistors (like tantalum) and even more quiet than the Vishay/Dale.
TDK's are also very good and super low on noise - but they are kind of big. I always try to avoid using carbon film due to the potential for the values to drift. But sometimes its unavoidable.
Matching (especially for precision phono) is always a pain...
https://www.bartola.co.uk/valves/tag/aot1n60/
But he was using it for source follower duty. I never tried this FET. Seriously, 1k Ohm gate stopper for source follower should not make a difference, and I would not expect it to make any difference for the Rmu too. Ale went down to 100 ohm and claimed it cured his fuzz on his source follower. Maybe something else was going on? This FET has decent looking specs.
The metal films I was using were from our local electronics shop in Zurich - Pusterla. They are 2W and have an attactive size. But I discovered that certain values are noisy. They are slighlty more noizy than Kiwame metal films... I am now just using them for power supply ground reference, stuff where their noise can't make it to my speakers...
Vishay/Dale CMF55 are generally pretty good. I have been using CMF60 type for gridstopper/gatestopper for some time and they have been fine. Not as expensive as fancy parts such as the Takman's I like to use. BUT if you are working with tiny signals, then the Amtrans carbon film are also great choice. They are not as expensive as other boutique resistors (like tantalum) and even more quiet than the Vishay/Dale.
TDK's are also very good and super low on noise - but they are kind of big. I always try to avoid using carbon film due to the potential for the values to drift. But sometimes its unavoidable.
Matching (especially for precision phono) is always a pain...
Last edited:
Ahhhhh!!! Haven't been in the temple for too long! Time to get one more of those blue plastic bags... 😉
Thanks @mbrennwa , I have missed the thread has continued.
I checked with an E188CC and a J113 as low fet and seems your formula taking 1K for Rp with that tube, gave me with the Gfs of this fet: 1k65 Rmu . I took 1k5 1W metal film. Seems to work, hopping not gross error !
Btw, is there any interrest to take a low fet or mosfet with a mA output near the one of the tube as the circuitry is a follower ? The more mA of the low fet the better to drive RCA capacitive cables for instance ?
I checked with an E188CC and a J113 as low fet and seems your formula taking 1K for Rp with that tube, gave me with the Gfs of this fet: 1k65 Rmu . I took 1k5 1W metal film. Seems to work, hopping not gross error !
Btw, is there any interrest to take a low fet or mosfet with a mA output near the one of the tube as the circuitry is a follower ? The more mA of the low fet the better to drive RCA capacitive cables for instance ?
Sure, why not. It all depends on the application. For example I am using FET followers to drive the grids of the output tubes.
Not sure I understood Ales point for driving cables, tipically his préamps output.
Balancing the mA of the tube with follower fet mA, he wrote, but this is not a push pull circuitry, no?
Then inputed he liked very high curent low mosfet (BH11 etc serie) cause sounds better???
What are your expérience guys with that circuitry, have you à prefered still sourciller fet préférence? Better to use a higher current and lower noise than à j113 before à preamp, like the j310 for illustration?
Do you like better cathode follower than Mu follower or else to drive à pre?
Logic says the low impedance output winns, but sometimes ears prefer something else... ?
Maybe fets are not the more linear device while they cope well with tubes...
Hummm, i am off topic of my own op question ....
Balancing the mA of the tube with follower fet mA, he wrote, but this is not a push pull circuitry, no?
Then inputed he liked very high curent low mosfet (BH11 etc serie) cause sounds better???
What are your expérience guys with that circuitry, have you à prefered still sourciller fet préférence? Better to use a higher current and lower noise than à j113 before à preamp, like the j310 for illustration?
Do you like better cathode follower than Mu follower or else to drive à pre?
Logic says the low impedance output winns, but sometimes ears prefer something else... ?
Maybe fets are not the more linear device while they cope well with tubes...
Hummm, i am off topic of my own op question ....
Last edited: