I became a believer on CCS after read this magnificent thread
http://www.diyaudio.com/forums/tube...ode-load-overcomes-3-2-law-non-linearity.html
But curiosity came first because of your designs, seriously.
Unfortunately DN2540 are unobtainable here.
What utter rot.
If the tube has a low Rp, then adding a CCS will make even less difference than the negligible diffrence it makes with a high Rp tube.
10 kohm (tube Rp) in parallel with 100 kohm (anode load reistor) = 9.1 kohm
10 kohm in parallel with 200 kohm = 9.52 kohm - a 5% increase.
10 kohm (tube Rp) in parallel with a perfect CCs = 10 kohm - another 5% increase.
You just can't hear changes that small.
There's nothing more simple and elegant than an RC or an RCRC filter. The heat from the series HT drop is the same. No heat is dissipated in a capacitor. An a simple RC or RCRC filter has no filament heat, control tube heat or gas tube heat. And a lot cheaper.
I became a believer on CCS after read this magnificent thread
http://www.diyaudio.com/forums/tube...ode-load-overcomes-3-2-law-non-linearity.html
But curiosity came first because of your designs, seriously.
Unfortunately DN2540 are unobtainable here.
The best thing in that thread is DF96's court reporting in Post #2.
Yes, if you miss the bigger picture, you might conclude that. Fortunately, not everyone looks at things so narrowly.
Some actual bench time with CCS circuits might prove useful to you in discovering the parts you're missing. But I do agree with you that tube regulators are more of a frippery, which is why I use solid state regs. (no offense intended, KK)
Blah.
CCS plate loads can be very good, and you might actually build one to compare with what you have and not all CCS are the same.
If we all dismissed things so quickly, we'd probably all be riding really fast horses with great fitting saddle gear.
LH
higher plate load--loadline is more horizontal. that means less distortion.
why chokes, ccs are so popular....
300H choke at 1khz is 1.9Mohm. imagine using such plate resistor, not very practical. dcr <3k will drop minimum plate voltage (and choke will swing above rail)
Last edited:
You are doing the wrong calculation, and hence drawing the wrong conclusions. Your sums are correct if you want to know output impedance, but that is not the issue. The issue is distortion.Keit said:
Ra/Rp is a nonlinear resistance, so as the signal swings it will vary over the cycle. Lets assume that it varies by 20%. If the valve feeds a short circuit then we get 20% variation in output signal current - which is 10% distortion (assuming mainly second-order - which it will mostly be). If the valve feeds 10x Rp then we get roughly 1%; for 20x Rp it is 0.5%. For a perfect CCS we get 0%. Now we can argue about how audible 1% distortion might be, but most would agree that 0% distortion is better.
The main advantage of a CCS load is that it doesn't require such high supply rail voltage as a suitable resistor value. Most people choose between CCS and an unsuitable resistor value fed from too low a supply rail. Given this (false) choice the CCS wins. Personally I am happy with the resistor and the high supply rail, but I don't diss those who choose the CCS.
Misleading example, based on bogus assumptions and operating conditions.
To illustrate requires a real example:
For a real triode to give Ra = 10K as you suggest means a mid to high mu triode.
Since we are talking about an audio amplifier, we will try some of the normal medium-high mu triodes with Ra = 10K, such as:
1. ECC81 = 12AT7
ECC81 gets Ra = 10.5K for Va=200V, 11.5mA from the typical characteristics section of the data sheet. A fairly high current is needed to keep it out of the curved (high distorting) portion of the triode curves.
Now we apply your 100K load resistor.
100K drops 1150V with 11.5mA flowing, so we need 1350V supply to satisfy your example, which is not only absurdly expensive, but violates Va(0) of the triode.
The load resistor burns 13.2W.
2. L63, 6SN7 etc
These are widely used in audio, and get Ra = 10K at about 250V and 5mA.
So in this case your 100K load resistor drops 500V and we need a mere 750V supply. The load resistor burns 2.5W.
Now compare these dismal results with a FET CCS load.
Even with cascoding, 50V is more than enough headroom, and with very little dependence on the actual current.
With a ECC81 we need a supply of 230V - 250V and the CCS load burns < 0.6W (no heatsink required).
With a 6SN7 the supply can range from 280-300V and the CCS burns < 0.25W (barely warm).
Then we must note that the power supply noise rejection will be hugely improved with a FET CCS (and can be improved even further if necessary, using cascoding).
And to round off the trio of advantages - in particular with the RIAA preamplifier mentioned - the FET CCS gives the option of a low-Z drive output - with no added parts. This is no small consideration when one remembers that the RIAA stage will drive an equaliser network, whose performance is vital to the overall performance of the amp.
To use either of these triodes with a load above 3x Ra (ie 33K or more) will require a compromise with the power supply voltage (415V or more) or compromise with the linearity by running Ia lower, and thus down into the curly region of the triode curves. And of course, the gain is reduced by much more than you claim.
This is neither "elegant" nor cost effective: it is futile skimping.
Actually, I was pointing out that the signal level was about the same. Perhaps I should have made that clear.
No, we don't get 0%. We get the distortion in the output voltage that is still there when anode current is held constant.
Have a look at the anode curves for a triode of your choice. For an example, I'll use the 12AU7 for which there are anode curves on Page C1 Mullard Technical Handbook Book 2 Part 1. Draw a horizontal line at a reasonable anode current, say 4 mA. You can read off the anode voltage resulting from any grid voltage. Here's the data:-
Vgk........ Vak....... del Vgk / del Vak (approx mu)
-25........ 428 ............-
-20........ 347.......... 16.2
-15........ 285.......... 12.4
-10........ 216.......... 13.8
-5.......... 135.......... 16.2
-0........... 46........... 17.8
Of course nobody would have a grid swing that large in a preamp, but you can see that the tube under constant current conditions certainly isn't linear.
Here' s the data for a 12AX7 at 4 mA:
Vgk........ Vak....... del Vgk / del Vak (approx mu)
-5.......... 466.............. -
-4.......... 373............. 93
-3.......... 283............. 90
-2.......... 192............. 91
-1........... 82............ 110
-0........... 14.............. 68
You can see from all this that there is an optimum anode voltage (which actually is a function of anode current) where distortion is theorectically zero, but non-zero for any finite signal swing. And the 12AX7 is a better choice is the higher mu is not inconvenient.
You probably are aware of the mathematical modelling of a triode. For an ideal planar triode mu is invariant with anode voltage. But for real triodes that isn't so, for a number of reasons, eg finite grid wire diameter, electron "leakage" around the grid instead of through it, etc.
This was brought home to me some years ago when I built an instrument for measuring voltages in the 10's of kilovolt range for client. The input is the anode of a colour TV shunt regulator triode. The grid is earthed, the cathode fedd from a CCS - a few uA as I recall. In theory the cathode voltage will then be the anode voltage divided by mu - a fixed ratio. It doesn't work that way in practice - a non-linear corrector built with op-amps was required to straighten it out.
I'll leave it as an exercise for you to plot anode voltage vs grid voltage for a triode of your choice, for various convenient constant anode currents, and anode currents changing slightly due to a resistor anode load.
You'll find you don't get 0% distortion at constant anode current, and the picture isn't much changed provided anode load is much greater than the tube Ra
On the other hand, you can see from the data that distortion for preamp level signals is pretty darn low - well below 1% even with low anode loads, comparable with Ra.
I would certainly agree that 0% will sound better than 1% too. But you pulled those figures out of nowhere. That's not like you, DF96. I have found your posts are always worth a carefull read.
It's certainly simpler to have a sufficiently high HT voltage. It's almost invariably available as its' needed for the power stage.
Last edited:
A reasonable choice would be the low mu 12AU7 which has Ra around 7 kohm. a 12SN7 is not much different. A 12AX7, a high mu triode, has an Ra about 75 but distorts a lot less with preamp level signals.
Your example, the 12AT7, is not designed for audio. It's meant to be a TV frequency changer - something that requires non-linearity. If you are going to use inappropriate tubes, the solution is not to add complexity. The solution is to choose a more appropriate tube.
The rest of your post is just so much woffle, because
a) the anode voltage can be a lot lower
b) A current of 3 to 4 mA is sufficient to get linearity about as good as it gets.
You don't need 11 mA.
After all, the 12AU7 and 12AX7 were designed for audio - with CCS loads.
For any given anode current, there is an optimum working point anode voltage, down to 2 mA or less. Take a look at some anode curves and see.
Last edited:
Agreed, but I was looking at the effect of varying the anode load - as I believe you were too. In real life there are always other things to think about as well.Keit said:
Yes. That is how you demonstrate an effect: pick plausible figures and show what they do. People who understand the effect can then use their own figures.
Yep. I was. And, yes, there are other things to think about.
However, so long as the anode load resistance is large compared to the tube Ra, the tube is operated close enough to constant current conditions that a CCS makes very little difference.
But your distortion figure were not plausible for a preamp. They are a couple of orders too high. 0.5% sounds better than 1%. But 0.05% does not sound better than 0.1%, as both are below what our ears can detect.
I'm glad you now agree that a CCS load does not result in 0% distortion.
It's not often you make an error, DF96.
The parameter Ra = 10K came from you. As did the RL = 100K - so don't blame me for the poor performance of your suggestions. I merely showed that such a stage is either impractical, or grossly compromised.
Trying to hide behind vague shifting of tube types or anode current does not rescue you, either - the same disadvantages will apply to a 12AU7 stage. If you don't think so, kindly present the full operating voltage/current/load and supply conditions, and I will demonstrate what you are missing.
Go away, Mr CCS seller.
Your example didn't stack up. There's no need for me to come up with anything more. You however can do so if you wish.
There's nothing vague about recommending a 12AX7 in place of the similar but much less linear 12AT7.
First, I correct your false allegation: I do NOT sell anode loads, whether CCS or any other kind.
You degrade the discussion into personal attacks - a sure sign that your technical argument has collapsed.
This is certain anyway, as there is no defence possible for your suggestion of a stage with parameter Ra = 10K and RL = 100K, without gross compromises compared to a CCS loaded stage. The 12AU7 that you suggested does not help you in the slightest.
12AX7 does not have Ra = 10K, or anything like it. You are merely trying to shift the goal-posts.
I don't see where he claimed that. However, the high rp of a 12AX7 makes things difficult. If one follows the "10 x rp" Rule of Dumb, we have to drop 600-800V across the anode resistor.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.