Member
Joined 2009
Paid Member
I know it's only spice but here's a comparison between a common cathode amplifier using 12AX7 and a common emitter amplifier using a typical BJT. The BJT has to biassed differently because it is not a depletion type device. It also has an unbypassed emitter resistor in order to match the gain of the two devices.
So - same B+, same signal, same gain and same bias current. From an FFT perspective the BJT beats out the triode.
So - same B+, same signal, same gain and same bias current. From an FFT perspective the BJT beats out the triode.
Attachments
Last edited:
What is it? Shouldn't the tube be running a much higher (relative to the transistor) voltage to be linear?
jeff
Member
Joined 2009
Paid Member
The voltage at the plate of the tube and collector of the BJT is just over 200V. hmmmm, that's actually too high for the BJT I've used.
Member
Joined 2009
Paid Member
I think it's better to remove the input capacitor in the BJT circuit so that the two circuits are DC coupled. Otherwise the triode one will have better performance at low frequencies.
Isn't the BJT going to have higher output impedance in this configuration ?
However more to the point, who is ever going to run the BJT in a way to achieve the slightly better performance ? I suggest no one.
Shoog
However more to the point, who is ever going to run the BJT in a way to achieve the slightly better performance ? I suggest no one.
Shoog
Hi,
A 175M bias resistor ? The bias is hopelessly inaccurate.
Bias should never depend on device hfe.
rgds, sreten.
A 175M bias resistor ? The bias is hopelessly inaccurate.
Bias should never depend on device hfe.
rgds, sreten.
Member
Joined 2009
Paid Member
Here's the thing, I've always seen a triode as the mose linear option for a VAS, something about the 3/2 law and all that compared with nasty transistor curves.
It's also said that a triode is linear partly because it operates at high voltage and works best when the swing at the plate is no more than 20% of the h.t. supply. Transistors often operate at low voltage.
I was curious how the transistor would perform if given similar operating conditions - lots of headroom and degenerated to same gain. I was pleasantly surprised - perhaps a single BJT could be used as a 'wimpy driver' for a 2A3.
It's also said that a triode is linear partly because it operates at high voltage and works best when the swing at the plate is no more than 20% of the h.t. supply. Transistors often operate at low voltage.
I was curious how the transistor would perform if given similar operating conditions - lots of headroom and degenerated to same gain. I was pleasantly surprised - perhaps a single BJT could be used as a 'wimpy driver' for a 2A3.
Just to add an anecdote which is not specifically relevant but interesting to me. I built a SLCF with a jfet input stage followed by FET cathode follower. I used a 1K cathode resistor on a negative rail to linearise the jfet, far more than the unbypassed cathode resistor on the equivilant triode. Comparing it to my valve based versions it was rolled off and lacking somewhat. I thought the degenerative feedback would make them interchangeable to the usual triode - but I was wrong. In the end I had to add a bit of gNFB to smooth the response out.
Lots of other parameters weren't equal but in all of the transistor amps I have recently built using jfet input they have consistently missed that top end sparkle.
Shoog
Lots of other parameters weren't equal but in all of the transistor amps I have recently built using jfet input they have consistently missed that top end sparkle.
Shoog
Nothing especially new or surprising there: for a long time now (at least 40yrs), transistor amps have soundly beaten tube ones on specs.
It is as difficult to find a solid state amp doing worst than 0.1% THD as it is finding a tube one doing better than that.
The reason is simple: transistors may be intrinsically more non-linear than tubes, but they also have a much higher transconductance, and can tolerate a much higher level of local degeneration for the same stage gain.
The result is a higher (apparent) linearity. To compare likes and likes, another configuration or tube type might be required.
If you use emitter degeneration to reduce gain, you are taking already high Zout and increasing it. This can create some bandwidth problems.
If you were to use collecter-base voltage feedback, you would end up with something that has much better linearity and much lower Zout than the triode. 😎
Smoking-amp started a thread recently "CFB and Schade feedback exactly equivalent" or something like that. Near the end of that thread, Gary Pimm posts some curves he traced of a Mosfet cascode with drain-gate feedback. Check out that graphic if you get a chance. The "curves" are simply perfectly-spaced near-vertical lines. Pretty awesome.
The fact is, these SS devices have far more transconductance than tubes. If you use that large transconductance in the right kind feedback loop to get equivalent gain to a tube, you will get better linearity and lower Zout than tubes.
If you were to use collecter-base voltage feedback, you would end up with something that has much better linearity and much lower Zout than the triode. 😎
Smoking-amp started a thread recently "CFB and Schade feedback exactly equivalent" or something like that. Near the end of that thread, Gary Pimm posts some curves he traced of a Mosfet cascode with drain-gate feedback. Check out that graphic if you get a chance. The "curves" are simply perfectly-spaced near-vertical lines. Pretty awesome.
The fact is, these SS devices have far more transconductance than tubes. If you use that large transconductance in the right kind feedback loop to get equivalent gain to a tube, you will get better linearity and lower Zout than tubes.
I would put triodes dead last for linearity. The different V to I power laws at grid1 and plate (or grid2), essentially locks in 2nd harmonic distortion with any kind of real load.
Even a pentode with resistive feedback is noticeably more linear than a triode.
Although, for tubes, I would take the resistive feedback to the previous stage cathodes, rather than the conventional Schade configuration. That keeps the gain functions high impedance.
Even a pentode with resistive feedback is noticeably more linear than a triode.
Although, for tubes, I would take the resistive feedback to the previous stage cathodes, rather than the conventional Schade configuration. That keeps the gain functions high impedance.
Last edited:
Member
Joined 2009
Paid Member
Well then, we have BJTs that can be set up to be readily more linear than triodes. So why do we see no feedback SE ended triode amps as being quite common yet BJT amps most commonly use lots of global feedback. And the few that don't are often quite complex. I thought perhaps it is because we accept low power from triodes but demand high power from BJTs so the playing field is very much not level ?
Lots of Pass amps with single ended and parallel FET's, but can't say I've seen any with BJT's.
jeff
You need to look at the higher harmonics, not just 2nd, 3rd & 4th. Low feedback around a BJT stage may sound bad. High feedback may be clinical sounding with the 2nd H gone too. FETs are very close to the best triodes (pentodes actually), just higher gm, so you can linearize them further as an option.
But 2nd harmonic is what 300B buyers are paying (dearly) for, more linearity would sound clinical.
But 2nd harmonic is what 300B buyers are paying (dearly) for, more linearity would sound clinical.
I noticed that in the original FFT plot the fourth and presumably higher order components of the BJT were higher than the Triode. This suggests that the BJT stage will sound worse than the Triode in this setup.
However the conclusion seem to be that if we convert almost all of the gain of the BJT into feedback we can get a better result. This enticing conclusion really does rest on the assumption that feedback comes without costs. If we use emitter degeneration then the impedance rises and the ability to drive the next stage falls. If we use drain to base feedback than the grid gets harder to drive and places more demands on the preceding stage (not great if the proceeding stage is another source over which we have little control). It seems there is no free lunch.
Shoog
However the conclusion seem to be that if we convert almost all of the gain of the BJT into feedback we can get a better result. This enticing conclusion really does rest on the assumption that feedback comes without costs. If we use emitter degeneration then the impedance rises and the ability to drive the next stage falls. If we use drain to base feedback than the grid gets harder to drive and places more demands on the preceding stage (not great if the proceeding stage is another source over which we have little control). It seems there is no free lunch.
Shoog
No, simply no: while it is true that the output conductance of a common emitter stage falls if the emitter degeneration is increased, it will always remain two orders of magnitude lower than the collector load resistance required for proper output biasing.
A Bjt is nothing like a triode, more like a pentode for the output characteristic at least.
This means that the output impedance will always be dominated by the physical collector resistor (unless an active load arrangement is used).
You can mimic triode operation by lowering the collector load if that particularly bugs you.
Otherwise, you can resort to typical semiconductor techniques and add a follower: additional transistors are almost free...
I did the same thing earlier this year with 140v and 12ax7 vs 2N5551.
I biased the transistor 470k and 7.5v zener supply.
Result of listening test of both through a SS amp is that the 12ax7 was smoother, better sounding.
I am stumped to as why but my next project is a 12ax7 pre-amp with loudness switch to drive the 250k volume pot in SX 3700 Reciever
I biased the transistor 470k and 7.5v zener supply.
Result of listening test of both through a SS amp is that the 12ax7 was smoother, better sounding.
I am stumped to as why but my next project is a 12ax7 pre-amp with loudness switch to drive the 250k volume pot in SX 3700 Reciever
Member
Joined 2009
Paid Member
Hey stocktrader200,
That's an interesting result, it is perhaps something to do with the higher order harmonics perhaps or non-linear capacitances of the BJT ?
That's an interesting result, it is perhaps something to do with the higher order harmonics perhaps or non-linear capacitances of the BJT ?
Hi Gareth,
Very nice of you to share this discovery.
If you are keen on trying an SS part to drive a 2A3, you may find the not so wimpy Ixys depletion mode IXTP01N100D a lot more interesting. Simulation shows higher gain and a lot less higher order harmonic components even compared to the 12AX7. Spice asc attached. I attenuate the input signal to get similar gain and op point.
Very nice of you to share this discovery.
If you are keen on trying an SS part to drive a 2A3, you may find the not so wimpy Ixys depletion mode IXTP01N100D a lot more interesting. Simulation shows higher gain and a lot less higher order harmonic components even compared to the 12AX7. Spice asc attached. I attenuate the input signal to get similar gain and op point.
- Status
- Not open for further replies.