Interesting, however I prefer the dielectric hysteresis curves of Steve Bench.
Although not provide numerical data, qualitatively better illustrates the linearity of the capacitors.
Maybe the work of Mr Bench not as rigorous, but it's the best I've seen so far.
That depends on what comes next. If the CCS is paralleled by the grid resistor of the next stage then you don't get infinite impedance so a cathode bypass might still be needed.GoatGuy said:
Likewise, the unbypassed cathode resistor increases the effective plate resistance, which highlights any shortcomings of the CCS. In many cases, I've experienced unexpectedly high distortion when the cathode resistor is unbypassed- I'm still uncertain why this is the case, but I have observed it pretty repeatedly. That's one reason I like to use LEDs for biasing as a partner to CCS- the combo really minimizes distortion.
When a triode has a purely linear anode load (such as a resistor) then a cathode resistor sampling the current provides a good proxy for anode voltage so it aids the natural triode feedback. When the load is non-linear (and the best CCS is likely to be less linear than even an average resistor, especially at higher frequencies) the cathode resistor is sampling a poorer proxy for output voltage. A mu which varies with signal might make things worse with a cathode resistor.
SY, the only "unbypassed cathode resistor increas[ing] the effective plate resistance" effect in a CCS loaded triode is imaginative. The whole point of the CCS is to make current flow "ideally constant". Therefore there is (somewhat surprisingly) - no A/C signal across Rk - at least if the CCS attains perfection. No A/C = infinite impedance for the cap. ((Zc = ∞) || Rk) = Rk. See?
DF96: Again, you have produced yet another realworld subtlety: the current draw of the grid resistor of the following stage tends to make the CCS "more perfect". A positive excursion of the grid (of stage 1) causes tube to "wish" to conduct more, pass more current. CCS adapts by increasing its voltage drop, lowering plate voltage. This corresponds to a negative excursion of the following grid, thus it "virtually sources" some A/C domain current back to the tube, and lessening the current change seen by the tube U1 (wow - for free, this). The opposite happens on negative source excursions. U1 wants to conduct less, CCS increases plate voltage on U1, Rg2 conducts more ... again "helping" the CCS. Wicked.
I do believe that I have come to the opposite conclusion - appropriately chosen (maybe Rk1*mu ?) ... Rg2 may have a significant "perfection enhancement" effect on the CCS. Further removing the requirement for having a Ck in parallel with Rk1 to "bypass" it.
GoatGuy
DF96: Again, you have produced yet another realworld subtlety: the current draw of the grid resistor of the following stage tends to make the CCS "more perfect". A positive excursion of the grid (of stage 1) causes tube to "wish" to conduct more, pass more current. CCS adapts by increasing its voltage drop, lowering plate voltage. This corresponds to a negative excursion of the following grid, thus it "virtually sources" some A/C domain current back to the tube, and lessening the current change seen by the tube U1 (wow - for free, this). The opposite happens on negative source excursions. U1 wants to conduct less, CCS increases plate voltage on U1, Rg2 conducts more ... again "helping" the CCS. Wicked.
I do believe that I have come to the opposite conclusion - appropriately chosen (maybe Rk1*mu ?) ... Rg2 may have a significant "perfection enhancement" effect on the CCS. Further removing the requirement for having a Ck in parallel with Rk1 to "bypass" it.
GoatGuy
That's the theory, yes, which is why I found the experimental results surprising.
The very best I've used were branded and sourced from Mouser or Digikey. The second best were cheap no-names bought surplus. I'd grab a few of your local ones and measure the AC impedance when forward biased- you could be pleasantly surprised.
Extra resistive loading may make the total load more 'perfect' in relative terms, but it also reduces its value. The latter more than removes the benefit of the former. There is no free lunch. Error terms are as likely to add as cancel.
A perfect CCS with no external load means that the cathode resistor provides no negative feedback so there is no benefit in not bypassing it. However, an unloaded stage provides no useful gain as you can't extract a signal from it.
Let us not forget that a CCS load deals with variation of ra(rp) with signal. It does not help with variation of mu with signal, although this is usually much smaller.
A perfect CCS with no external load means that the cathode resistor provides no negative feedback so there is no benefit in not bypassing it. However, an unloaded stage provides no useful gain as you can't extract a signal from it.
Let us not forget that a CCS load deals with variation of ra(rp) with signal. It does not help with variation of mu with signal, although this is usually much smaller.
Last edited:
However if there is little load (high impedance) you generally get less mu variation. This is not irrelevant when you want to swing large signals, especially using small valves. I see the advantage of plate CCS load from a different point of view. In a typical voltage amplifying stage your load with a plate CCS will essentially be that of the next stage and it can be much higher than you can get in any other way.
If you take for example a 6SL7 you can have 470K load with "the typical operation" configuration (i.e. 250V plate voltage and 2.3 mA current) rather than the typical configuration as resistance-couple amplifier. This allows the valve to work in extremely linear conditions ( the mu is essentially flat over a wide region and distortion really small and mostly related to the quality of the actual valve). Using resistors you should have 910K on the plate and 1M next stage, for example, which makes this impossible. It is impossible because even if you could waste some power and have (a rather compromised) 2343V supply voltage (!) you couldn't do it because the valve cannot withstand that initial high voltage.
I would agree that some people tend to use CCS's where they are not really needed or even desirable....
Mu may or may not be flat- remember, the curves shown on datasheets are almost always mu versus current, which is entirely irrelevant in this context. Mu versus voltage can only be picked off at two or three points and generally has to be measured directly.
I didn't mean those curves but the fact that if you pick your load and look at the amplified signal you will see that the flatter the loadline the lower the variation over a bigger scale (with less distortion). At lower anode current you won't get the same result even with higher load when asking for large output.
Last edited:
The intrinsic distortion in a triode can be seen as a non-linearity in ra. The less ra is loaded the lower the distortion until you reach the intrinsic distortion of the tube. A CCS plate load or a mu follower provide a very light load for ra and distortion approaches the intrinsic value. With an unbypassed cathode resistor, ra is increased by Rk*(u+1) to it is no longer so lightly loaded and distortion increases.
Cheers
Ian
That's basically what I said a few posts back. But that isn't clearly the answer for two reasons. First, the loading of the circuit in my measurements was about 10M (limited by the probe, not the CCS), which means that (aside from high rp tubes like 12AX7), the rp variation portion of the voltage divider equation is pretty small. Second, the "added" rp variation due to degeneration is small because the term is a function of mu, which varies less with voltage than does rp. (See, for example, page 10 of this unusually good datasheet http://tubedata.milbert.com/sheets/184/1/12AT7.pdf)
Gripping hand, the observed distortion increase wasn't subtle at all- it was easily visible on a scope trace, which implies several percent at least. I could attach a bypass or LED and watch the distortion come and go.
There's probably a very simple reason for this observation (and increased effective rp could still be it), but to be honest, I've been too lazy to chase it down.
Gripping hand, the observed distortion increase wasn't subtle at all- it was easily visible on a scope trace, which implies several percent at least. I could attach a bypass or LED and watch the distortion come and go.
There's probably a very simple reason for this observation (and increased effective rp could still be it), but to be honest, I've been too lazy to chase it down.
Not if mu is constant and the external load is linear. You have a potential divider consisting of ra(rp), Rk(mu+1) and RL - with output voltage taken across RL. Increasing Rk helps to further swamp the variation in ra, provided that mu is constant. If mu is not constant then the variation on the denominator helps to partially offset the variation in the overall multiplying mu, although this may be only a small effect with realistic values for Rk. If RL is non-linear then there will come a point where Rk(mu+1) dominates over ra and then further increases in Rk will increase distortion.ruffrecords said:
With constant mu, Rk gives a linear boost to ra so the valve distortion is reduced but at the cost of exposing CCS distortion. With variable mu (the real situation), Rk may help or hinder - depends on how mu varies compared with how ra varies. I think it normally hinders, as mu drops and ra rises at high voltages for most valves.
As SY says, data sheets usually give mu variation against current (or not at all) when we want it against voltage. However, it can be estimated from the compression of the mutual characteristic curves on the right of the plot.
I don't think so. Distortion could be lower. It all depends on what you are asking for and how you set it up.
Anyway a proper CCS can provide a much higher dynamical load. A simple depletion mode mosfet cascode CCS can do this and one can find now the right devices for any current (LND150, XPTP01N100D, DN2540 etc...). Actually the LND150, with currents up to 1 mA, will give the highest impedance because of its lower caps (it will be better than a 2 giga resistor for sure at any frequency in the audio range!) and that is coming exactly where this feature is more desirable (high mu signal triodes). In the worst case one has to do some selection.
Not bypassing the cathode resistor will also mean local feedback. So if you don't need the full gain it could be better both from the measurement and listening point of views. For me it is not correct judging before listening anyway.
"The use of an electrolytic cathode bypass capacitor makes no discernible difference to the sound other than the value affects the frequency response. In properly designed bypass and coupling circuits there is no evidence of any discernible difference. If you have such evidence I would be pleased to see it. On the one hand you ask me to accept that some people can hear differences yet you refuse to provide any evidence other than your opinion. Not surprisingly, I remain unconvinced."
I'm as sceptical as Vincent about this statement by Ian. It troubles me. Two reasons why:
1. No discernable difference compared to what? This is like the sound of one hand clapping. You can’t have a tube stage without any biasing, so in practice all you can do is compare different forms of bias. A tube stage without any biasing is inaudible.
2. What do you call evidence? Take the statement "many children are scared by dogs". To believe this do we need to locate the neural pathways in the brain that have been proven to be associated with fear reactions, and put a child in a hospital scanner when exposed to a canine to read out the digital picture of the brain? Or do we take on face value "the child states it is scared of the dog". User reports are widely used in statistics. They may be fuzzy in reliability but they do not constitute "no evidence at all".
These are just two things that trouble me. A third is the idea that if a circuit is theoretically correct you would listen to it on principle rather than to another circuit that was more subjectively satisfying.
I'm not making any dogmatics statements here, since there is a lot of obviously subjective material, but I remain troubled. The technical arguments are very interesting and I've been trying to follow them as well.
I'm as sceptical as Vincent about this statement by Ian. It troubles me. Two reasons why:
1. No discernable difference compared to what? This is like the sound of one hand clapping. You can’t have a tube stage without any biasing, so in practice all you can do is compare different forms of bias. A tube stage without any biasing is inaudible.
2. What do you call evidence? Take the statement "many children are scared by dogs". To believe this do we need to locate the neural pathways in the brain that have been proven to be associated with fear reactions, and put a child in a hospital scanner when exposed to a canine to read out the digital picture of the brain? Or do we take on face value "the child states it is scared of the dog". User reports are widely used in statistics. They may be fuzzy in reliability but they do not constitute "no evidence at all".
These are just two things that trouble me. A third is the idea that if a circuit is theoretically correct you would listen to it on principle rather than to another circuit that was more subjectively satisfying.
I'm not making any dogmatics statements here, since there is a lot of obviously subjective material, but I remain troubled. The technical arguments are very interesting and I've been trying to follow them as well.
Evidence: Well-documented and performed measurements and/or well-documented and performed listening tests.
Not Evidence: Bald assertion, anecdote, claims that something has to be true because so many people believe it.
Not Evidence: Bald assertion, anecdote, claims that something has to be true because so many people believe it.
If the wrong capacitor is put in there, then it is possible you will hear the problem, due to incorrect selection.
If the correct capacitor is in there, then it should be impossible to hear the difference when another correct capacitor is swapped in.
- Status
- Not open for further replies.