Hi Woody,
That's an excellent question but one I can't answer properly.
One thing though is that the tube is still doing the voltage amplification.
The other is that some find that it does cause a solid state type of sound and would rather use a resistor or plate choke (or a Mu-stage).
Another thing apparently is that the shunt capacitance if one of the most important variables when it comes to subjective appreciation of the gain stage. And the lower the better.
The added advantage of a ss ccs (assuming ofcourse that it works properly) is ofcourse the ripple rejection is so huge that the advantage of using one might offset any possible disadvantages..
I think the first point comes closest to an anwer.
That's an excellent question but one I can't answer properly.
One thing though is that the tube is still doing the voltage amplification.
The other is that some find that it does cause a solid state type of sound and would rather use a resistor or plate choke (or a Mu-stage).
Another thing apparently is that the shunt capacitance if one of the most important variables when it comes to subjective appreciation of the gain stage. And the lower the better.
The added advantage of a ss ccs (assuming ofcourse that it works properly) is ofcourse the ripple rejection is so huge that the advantage of using one might offset any possible disadvantages..
I think the first point comes closest to an anwer.
In a SS CCS the transistor is being used in constant current mode and therefore avoiding the worst SS issues of beta changing with collector current.
Of course you still have to cope with Early Effect issues which with a single transistor CCS will limit the effective load impedance to say 300K and will roll off highs at say 200 - 300kHz due to device capacitance. Often this is good enough. Another issue (well part of the same issue really) is that the capacitance (Ccb) changes with Vcb changes (i.e. the collector to base depletion region width and hence capacitance varies with signal voltage developed across the transistor). This introduces variable high frequency feedback in the CCS.
Both of the latter issues can be addressed by using a Cascode SS Current Source.
Bas has it right - the CCS is only providing a load (albiet a good high impedance load with good {linear} frequency characteristics) for the tube voltage amplifier and so sonically you are getting the signature of the tube itself, improved somewhat by the tube being operated in constant current mode also.
This maximizes gain and minimizes distortion from the tube itself.
Hope this makes sense to you.
Cheers,
Ian
Of course you still have to cope with Early Effect issues which with a single transistor CCS will limit the effective load impedance to say 300K and will roll off highs at say 200 - 300kHz due to device capacitance. Often this is good enough. Another issue (well part of the same issue really) is that the capacitance (Ccb) changes with Vcb changes (i.e. the collector to base depletion region width and hence capacitance varies with signal voltage developed across the transistor). This introduces variable high frequency feedback in the CCS.
Both of the latter issues can be addressed by using a Cascode SS Current Source.
Bas has it right - the CCS is only providing a load (albiet a good high impedance load with good {linear} frequency characteristics) for the tube voltage amplifier and so sonically you are getting the signature of the tube itself, improved somewhat by the tube being operated in constant current mode also.
This maximizes gain and minimizes distortion from the tube itself.
Hope this makes sense to you.
Cheers,
Ian
Why does a CCS not sound like sand?
Whoa, this question could be the subject of a PHD dissertation. The answer would involve the discussion of why transistors sound the way that they do. There are many sides to this subject but two points are important here.
A transistor (fet or bipolar) is an inherently non-linear voltage amplifier when operated at large signal. Cascading many transistors to build an amplifier generally creates a non-linear amplifier.
To make up for the non-linearities, a large amount of negative feedback is applied. The feedback is largely responsible for the solid state sound, and most tube amplifiers do not use much feedback.
For a better explanation see:
http://www.tubelab.com/philosophy.htm section #7
The CCS does not operate the active device as a voltage amplifier. In fact it is not used to amplify the audio signal at all. An ideal CCS would pass a constant amount of current to the amplifying tube, while presenting an infinite impedance at ALL frequencies of importance to the amplifying stage. It should be noted that this is the ideal load for a triode. A CCS load should not be used for a tetrode or pentode.
As with everything else in life nothing is perfect. All CCS designs both tube and solid state have a finite impedance which changes with frequency. There are many good CCS designs available on the web.
This guy seems to have made a career out of them:
http://www.pacifier.com/~gpimm/
I have tried several different CCS's and have settled on an IC based current source that works excellent. In fact I have used it in every amp design that I have made since I found out about it.
http://www.tubelab.com/Simple45.htm
I have used this IC (IXYS 10M45) to build a driver that is flat out to 300 KHz while driving an 833A into A2. There are mosfets and IC's in the design, but no sandy sound.
http://www.tubelab.com/powerdrive.htm
I have even used one of these to load a 45 for chokeless parafeed. You need a 500 volt power supply to do this though.
Whoa, this question could be the subject of a PHD dissertation. The answer would involve the discussion of why transistors sound the way that they do. There are many sides to this subject but two points are important here.
A transistor (fet or bipolar) is an inherently non-linear voltage amplifier when operated at large signal. Cascading many transistors to build an amplifier generally creates a non-linear amplifier.
To make up for the non-linearities, a large amount of negative feedback is applied. The feedback is largely responsible for the solid state sound, and most tube amplifiers do not use much feedback.
For a better explanation see:
http://www.tubelab.com/philosophy.htm section #7
The CCS does not operate the active device as a voltage amplifier. In fact it is not used to amplify the audio signal at all. An ideal CCS would pass a constant amount of current to the amplifying tube, while presenting an infinite impedance at ALL frequencies of importance to the amplifying stage. It should be noted that this is the ideal load for a triode. A CCS load should not be used for a tetrode or pentode.
As with everything else in life nothing is perfect. All CCS designs both tube and solid state have a finite impedance which changes with frequency. There are many good CCS designs available on the web.
This guy seems to have made a career out of them:
http://www.pacifier.com/~gpimm/
I have tried several different CCS's and have settled on an IC based current source that works excellent. In fact I have used it in every amp design that I have made since I found out about it.
http://www.tubelab.com/Simple45.htm
I have used this IC (IXYS 10M45) to build a driver that is flat out to 300 KHz while driving an 833A into A2. There are mosfets and IC's in the design, but no sandy sound.
http://www.tubelab.com/powerdrive.htm
I have even used one of these to load a 45 for chokeless parafeed. You need a 500 volt power supply to do this though.
I agree with the above responses but just wanted to add that I think that a solid state CCS can sound like "sand" even if is "just a plate load". Nor should we take any false comfort from knowing that the tube is the voltage amplifying element. Amplification fidelity is a function of the environment too, plate loads, cathode loads, supplies, etc. A badly conceived SS plate load can make a tube gain stage sound plenty “sandy”. Only when a CCS has extremely high dynamic resistance, and extremely low capacitance (arguably <<1pF) then I think the tube parameters will dominate. The key, to me, is just how much shunt capacitance exists across this CCS, assuming that dynamic resistance is already very high (which is pretty easy). As mentioned above, capacitance associated with SS devices can be enormously variable with signal swing, creating a nasty non-linearity that causes phase intermodulation distortion. I think this is worse than Early Effect in aural destructiveness. Gary Pimm's work has driven the CCS dynamic resistance very high and also reduced the shunt C values, especially in some of his later versions, to vanishingly low levels. I see that he cascodes some of his designs with a pentode, which happily isolates the semiconductor capacitance even more. But too often I see well-intentioned, but overly simple, solid state current sources as plate loads. Sometimes you'll see just a dangling PNP handing off B+ with collector to plate, or a simple FET current source. These I would avoid! And MOSFET capacitances can be enormous and very distorting.
Here’s a thought: use a series choke with a CCS to further isolate it. It would have to be wound with very low shunt C itself though.
Here’s a thought: use a series choke with a CCS to further isolate it. It would have to be wound with very low shunt C itself though.
CCS loads for triodes.
I also agree that the capacitance of the CCS and the following stage is very important. I like to consider the DC and AC components of the triode's load seperately. Ideally they should both be infinite. Since this is impossible, they should be as high as possible. This is what the Mu - follower, SRPP and other similar circuits attempt to do.
I have tried most of those circuits with varying degrees of success. During the testing I found circuits that definitely sounded bad. I don't know if they sounded "sandy" but the "simple fet" circuits sounded dull and lifeless. A simple pentode (hard to find one with really constant current) CCS's didn't sound much better. As expected the complex circuits provided better sound AND measurements. Many of these circuits also provide buffering to increase the AC impedance that the triode sees. I have also seen the circuits with a fet CCS followed by a cathode follower.
While testing various CCS circuits, I found the IXYS 10M45 IC. This surprised me with good performance. Ok this solves the DC load problem, but what about AC. Just hanging the grid of an 845 off of the triode / CCS won't work. Cathode followers help, but I wasn't satisfied. I experimented with bipolar emitter followers but ran into SOA problems. I am currently using a mosfet source follower with good results. The gate to source capacitance may be high, but these two terminals are in phase at very similar voltages, so the actual capacitance seen by the triode is low. Still the choice of fet is important. I got some IGBT's but haven't tried them yet.
Another parameter that is often overlooked is the triode's current. Too often the designer chooses a 6SN7 or a 12A*7 and biases it to 1mA of current. I have found that more current is needed to charge and discharge the AC load capacitance. I found that the 5842 tube likes 10 to 14 mA of current. This is what I used in the TubelabSE, but since I designed that amp, those tubes have become trendy and the price has gone way up.
All of this science led to the development of the PowerDrive circuit. This circuit has ruler flat phase response out to 245KHz, while driving an 833A into A2. It sounds good too.
I see that your avatar is an 833A, have you built an 833A amp yet. If so, what do you use for an output transformer.
I also agree that the capacitance of the CCS and the following stage is very important. I like to consider the DC and AC components of the triode's load seperately. Ideally they should both be infinite. Since this is impossible, they should be as high as possible. This is what the Mu - follower, SRPP and other similar circuits attempt to do.
I have tried most of those circuits with varying degrees of success. During the testing I found circuits that definitely sounded bad. I don't know if they sounded "sandy" but the "simple fet" circuits sounded dull and lifeless. A simple pentode (hard to find one with really constant current) CCS's didn't sound much better. As expected the complex circuits provided better sound AND measurements. Many of these circuits also provide buffering to increase the AC impedance that the triode sees. I have also seen the circuits with a fet CCS followed by a cathode follower.
While testing various CCS circuits, I found the IXYS 10M45 IC. This surprised me with good performance. Ok this solves the DC load problem, but what about AC. Just hanging the grid of an 845 off of the triode / CCS won't work. Cathode followers help, but I wasn't satisfied. I experimented with bipolar emitter followers but ran into SOA problems. I am currently using a mosfet source follower with good results. The gate to source capacitance may be high, but these two terminals are in phase at very similar voltages, so the actual capacitance seen by the triode is low. Still the choice of fet is important. I got some IGBT's but haven't tried them yet.
Another parameter that is often overlooked is the triode's current. Too often the designer chooses a 6SN7 or a 12A*7 and biases it to 1mA of current. I have found that more current is needed to charge and discharge the AC load capacitance. I found that the 5842 tube likes 10 to 14 mA of current. This is what I used in the TubelabSE, but since I designed that amp, those tubes have become trendy and the price has gone way up.
All of this science led to the development of the PowerDrive circuit. This circuit has ruler flat phase response out to 245KHz, while driving an 833A into A2. It sounds good too.
I see that your avatar is an 833A, have you built an 833A amp yet. If so, what do you use for an output transformer.
Gluca,
You will hear arguments from both choke advocates and the CCS advocates. Not to sound too pedantic, but the real answer is that it depends. My short answer is that if you are willing to build a complex design, one of Gary Pimm’s more recent CCS designs seem to come pretty close to perfection on technical grounds. I have not built his exact designs so I am not able to comment on their sound, but others, including some people I regard highly, are very impressed. But I do know first hand that a good CCS can make a triode sing. A good choke is also hard to beat and needs less B+ voltage. You have to size the inductance to get deep enough bass, and the shunt C must be low for extended highs. As I said above, a really dedicated person could combine a CCS with a smaller choke. The CCS would give high impedance down to low bass (possibly even to DC), while a small choke with very low C in series would make sure that the shunt C of the CCS is isolated at high frequencies. Of all these choices, probably just buying a really good choke is the easiest.
You will hear arguments from both choke advocates and the CCS advocates. Not to sound too pedantic, but the real answer is that it depends. My short answer is that if you are willing to build a complex design, one of Gary Pimm’s more recent CCS designs seem to come pretty close to perfection on technical grounds. I have not built his exact designs so I am not able to comment on their sound, but others, including some people I regard highly, are very impressed. But I do know first hand that a good CCS can make a triode sing. A good choke is also hard to beat and needs less B+ voltage. You have to size the inductance to get deep enough bass, and the shunt C must be low for extended highs. As I said above, a really dedicated person could combine a CCS with a smaller choke. The CCS would give high impedance down to low bass (possibly even to DC), while a small choke with very low C in series would make sure that the shunt C of the CCS is isolated at high frequencies. Of all these choices, probably just buying a really good choke is the easiest.
Tubelabs,
Your point about inadequate plate current bias is so true. Running triodes at richer current levels increases transconductance almost proportionately, lowers plate resistance and lowers noise. Plus, as you say, there is more current to charge capacitances as well as lower resistances to create shorter time constants. The down sides include the possibility of grid current becoming a factor at low plate voltages combined with high current as Vgk shrinks, and that B+ needs to be higher if using plate resistor loads. Morgan Jones takes the latter approach in his fine book “Valve Amplifiers”. He uses large value wire-wound plate resistors to a high B+. A lot of wasted power but simpler than a CCS.
I forget that my avatar follows me around. 833s yes, output transformer, no. Here’s a description (scroll to bottom):
Direct drive ESL amp using 833s
Your point about inadequate plate current bias is so true. Running triodes at richer current levels increases transconductance almost proportionately, lowers plate resistance and lowers noise. Plus, as you say, there is more current to charge capacitances as well as lower resistances to create shorter time constants. The down sides include the possibility of grid current becoming a factor at low plate voltages combined with high current as Vgk shrinks, and that B+ needs to be higher if using plate resistor loads. Morgan Jones takes the latter approach in his fine book “Valve Amplifiers”. He uses large value wire-wound plate resistors to a high B+. A lot of wasted power but simpler than a CCS.
I forget that my avatar follows me around. 833s yes, output transformer, no. Here’s a description (scroll to bottom):
Direct drive ESL amp using 833s
Brian, re the choke in series with a CCS, might the series resonance with the CCS capacitance be an issue?
For me, I hit the wall of diminishing returns with a cascode bipolar CCS. The nice thing about current and voltage sources is that their performance is completely objective- either the current is constant or it isn't. So measurement can be very effective in sorting the wheat from the chaff.
For me, I hit the wall of diminishing returns with a cascode bipolar CCS. The nice thing about current and voltage sources is that their performance is completely objective- either the current is constant or it isn't. So measurement can be very effective in sorting the wheat from the chaff.
SY,
It could be a problem if the Q is too high at resonance. You got me thinking, so I ran a quick simulation assuming a CCS with 5pF of capacitance shunted across a 500K dynamic resistance, a reasonable set of values for a simple SS CCS. With a 10Henry choke having small internal capacitance (that’s really the hard part), the choke usefully improves (raises) Z in the upper audio octaves and beyond. With extremely high values of CCS resistance (in the hundreds of meg), the series resonance Q increases and we see a Z dip somewhere at the top of the audio band or just above, depending on the internal C of the choke. There is no visible trace of resonance with 500K in the CCS as damping though. Like everything else, it depends on lots of factors and assumptions. I was just brainstorming out loud - and that’s a risky thing to do!
It could be a problem if the Q is too high at resonance. You got me thinking, so I ran a quick simulation assuming a CCS with 5pF of capacitance shunted across a 500K dynamic resistance, a reasonable set of values for a simple SS CCS. With a 10Henry choke having small internal capacitance (that’s really the hard part), the choke usefully improves (raises) Z in the upper audio octaves and beyond. With extremely high values of CCS resistance (in the hundreds of meg), the series resonance Q increases and we see a Z dip somewhere at the top of the audio band or just above, depending on the internal C of the choke. There is no visible trace of resonance with 500K in the CCS as damping though. Like everything else, it depends on lots of factors and assumptions. I was just brainstorming out loud - and that’s a risky thing to do!
CCS
Brian Beck:
+ and - 4000 volts, and I thought I was crazy. Actually I have no experience with electrostatics, and no room for them ( I have a small house), so I never thought much about how to drive them. I followed that thread, and thought about it, and it does make perfect sense, an OTL that could actually work right.
I just got Morgan Jones new books (now a two book set). The books have been reviewed as "hard to understand", but I find lots of good information here, although there does seem to be a bias against SE amplifiers.
SY:
I have tried several fets. The one I currently use is the 2SK2700 from Toshiba. It is available from DigiKey. It was chosen based on a 900 volt breakdown voltage and a low gate capacitance. The driver has no phase shift out to 245KHz, and the -3DB point is 350KHz, while driving an 833A at +20 volts bias (A2). I have ordered some newer fets and some IGBT's but won't get to try them until I return from vacation. I leave tomorrow morning.
How does screen drive work with the 6LF6. I have a box full of 6LW6's (same tube with octal base) and would like to try it. Are you using SE or P-P? What voltage on G1 and plate? What voltage do you set G2 at and how much drive voltage do you feed it?
Brian Beck:
+ and - 4000 volts, and I thought I was crazy. Actually I have no experience with electrostatics, and no room for them ( I have a small house), so I never thought much about how to drive them. I followed that thread, and thought about it, and it does make perfect sense, an OTL that could actually work right.
I just got Morgan Jones new books (now a two book set). The books have been reviewed as "hard to understand", but I find lots of good information here, although there does seem to be a bias against SE amplifiers.
SY:
I have tried several fets. The one I currently use is the 2SK2700 from Toshiba. It is available from DigiKey. It was chosen based on a 900 volt breakdown voltage and a low gate capacitance. The driver has no phase shift out to 245KHz, and the -3DB point is 350KHz, while driving an 833A at +20 volts bias (A2). I have ordered some newer fets and some IGBT's but won't get to try them until I return from vacation. I leave tomorrow morning.
How does screen drive work with the 6LF6. I have a box full of 6LW6's (same tube with octal base) and would like to try it. Are you using SE or P-P? What voltage on G1 and plate? What voltage do you set G2 at and how much drive voltage do you feed it?
SY, me either, other than as a dumb user. Anybody else know what's achievable? I'm sure there are winding interleaving techniques to lower shunt C, but how low can you go?
If I can get a chance, I will measure a few I have in my collection. Should be easy, series R, measure for a dip in level at resonance, back calculate C.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.