Hi
Everyone says that triode tubes have more 2nd harmonic THD than anything else and I wanted to see this in LTspice. I made a CC gain stage with typical values and ran it with a low-level sine input. The output was generally clean looking and THD was pretty typical from what I've read.
The FFT of the output showed odd and even harmonics with the 3rd not being much lower than the 2nd. A few higher harmonics were present in both FFTs. I know the curve for a triode follows a 2/3 exponent, which maybe does produce both even and odd harmonics?
I wondered what the input signal looked like and its FFT was similar AND its THD baseline was only a little below the baseline for the output FFT. This made me wonder if the tube was simply amplifying the dirty oscillator harmonics?
I went to a different file for a super-low-THD solid-state amp and checked its input and output FFT. In this case, the baseline for the output FFT was around -180dB and 10dB lower for the input.
So, it seems like LTspice scales the THD of the sine generator depending on the range of THD the circuit under test generates?
Does this seem normal?
Thanks
Everyone says that triode tubes have more 2nd harmonic THD than anything else and I wanted to see this in LTspice. I made a CC gain stage with typical values and ran it with a low-level sine input. The output was generally clean looking and THD was pretty typical from what I've read.
The FFT of the output showed odd and even harmonics with the 3rd not being much lower than the 2nd. A few higher harmonics were present in both FFTs. I know the curve for a triode follows a 2/3 exponent, which maybe does produce both even and odd harmonics?
I wondered what the input signal looked like and its FFT was similar AND its THD baseline was only a little below the baseline for the output FFT. This made me wonder if the tube was simply amplifying the dirty oscillator harmonics?
I went to a different file for a super-low-THD solid-state amp and checked its input and output FFT. In this case, the baseline for the output FFT was around -180dB and 10dB lower for the input.
So, it seems like LTspice scales the THD of the sine generator depending on the range of THD the circuit under test generates?
Does this seem normal?
Thanks
No. Beware that SPICE is not an accurate prediction for distortion for anything but BJTs, but it doesn't do anything like that. If SPICE is showing higher distortion at the output of the generator, that's because it thinks it's there. Post the circuit - it's very easy to get odd results in SPICE.
LTspice THD will max out unless you set certain directives:
.options plotwinsize =0
and set the transient simulation Maximum timestep something like
.tran 0 {10/f} {2/f} {1e-4/f}
where
.par f=1k (etc)
Others have done more but the above are what counts.
.options plotwinsize =0
and set the transient simulation Maximum timestep something like
.tran 0 {10/f} {2/f} {1e-4/f}
where
.par f=1k (etc)
Others have done more but the above are what counts.
It depends on idle current. If the stage is biases such a way so signals are distorted a-symmetrically, there will be 2'nd harmonic dominant, if it is biased in the middle, so distortions are symmetrical, 3'rd order will be dominant. But in any case increasing signal level you will see wider and wider spectrum of distortions.
Hi
Thanks for the replies!
EC81010: Your reply prompted me to try a different BJT circuit, a simple EF that has fairly poor performance. With a 100mVpk 20kHz sine input the THD out is 0.63%. The THD of the generator shows as 30ppb (five zeroes and a 3).
I tried an intermediate performing circuit whose output THD is 27ppm at 20kHz and Vin shows as <10ppb (six zeroes).
I tried a different BJT circuit where THD out was 0.09% and for Vin is 0.01%. I reduced the input level and had Vout 0.038% and Vin 0.009. I reduced Vin much further and Vout THD was slightly lower than Vin, 0.13853 vs 0.013859. This circuit is a power follower with jfet input and mosfet outputs, and a DC servo.
So, I believe you are correct about LTspice being optimised for BJTs, but it seems dumbfounding that the generator THD would vary with the different active devices. And there still seems to be some relationship to the circuit under test, where as in the third example the devices are all fets and performance is not so great.
Steveu: I only know how to do things in LTspice that were detailed in Bob Cordell's book, so the .plotwinsize=0 and other parameters are all there, as are the settings for 1kHz THD and 20kHz THD tests.
One thing I have noticed with BJT circuits is that the THD will be higher if the DC offset at the output is high. For some circuits with very low available gain, I added a DC servo to take care of that and THD goes down.
Waveburn: Thanks for that. I recall that the basic parameters of the tube are all based on using a tiny portion of the transfer curve, which is probably miniscule compared to the usual signal sizes on the plate?
I went back to the original tube circuit which is simply a 12AX7 with 100k on the plate and 10k for the cathode. The load through the plate coupling cap is 250k. I set Vin to 30mV and got 180mVpk out on the positive half and slightly less negative. THD out is 0.065734% and for Vin is 0.051850%. I went down to 3mVin and THD was worse: 0.072 out and 0.056 in. I increased Vin to where there was very noticeable output distortion (37%) and THD of Vin went down (0.045)! I find this kind of confusing.
LTspice may not be the best choice for getting THD results from tube circuits, but hopefully it can provide relative performance indication if it is consistent?
Thanks for the replies!
EC81010: Your reply prompted me to try a different BJT circuit, a simple EF that has fairly poor performance. With a 100mVpk 20kHz sine input the THD out is 0.63%. The THD of the generator shows as 30ppb (five zeroes and a 3).
I tried an intermediate performing circuit whose output THD is 27ppm at 20kHz and Vin shows as <10ppb (six zeroes).
I tried a different BJT circuit where THD out was 0.09% and for Vin is 0.01%. I reduced the input level and had Vout 0.038% and Vin 0.009. I reduced Vin much further and Vout THD was slightly lower than Vin, 0.13853 vs 0.013859. This circuit is a power follower with jfet input and mosfet outputs, and a DC servo.
So, I believe you are correct about LTspice being optimised for BJTs, but it seems dumbfounding that the generator THD would vary with the different active devices. And there still seems to be some relationship to the circuit under test, where as in the third example the devices are all fets and performance is not so great.
Steveu: I only know how to do things in LTspice that were detailed in Bob Cordell's book, so the .plotwinsize=0 and other parameters are all there, as are the settings for 1kHz THD and 20kHz THD tests.
One thing I have noticed with BJT circuits is that the THD will be higher if the DC offset at the output is high. For some circuits with very low available gain, I added a DC servo to take care of that and THD goes down.
Waveburn: Thanks for that. I recall that the basic parameters of the tube are all based on using a tiny portion of the transfer curve, which is probably miniscule compared to the usual signal sizes on the plate?
I went back to the original tube circuit which is simply a 12AX7 with 100k on the plate and 10k for the cathode. The load through the plate coupling cap is 250k. I set Vin to 30mV and got 180mVpk out on the positive half and slightly less negative. THD out is 0.065734% and for Vin is 0.051850%. I went down to 3mVin and THD was worse: 0.072 out and 0.056 in. I increased Vin to where there was very noticeable output distortion (37%) and THD of Vin went down (0.045)! I find this kind of confusing.
LTspice may not be the best choice for getting THD results from tube circuits, but hopefully it can provide relative performance indication if it is consistent?
Check generator source resistance; the way distortion can change is if the load draws a distorted current that drops a distorted voltage across generator source resistance. There's nothing magic about SPICE, but I think you're looking at it in the wrong way. All devices obey physics, but BJTs follow physics models with far less device variation than anything else, so SPICE models them well. FETs have huge device variation in VTO, and valves are mechanical structures with associated mechanical tolerances, so they have significant device variation, which means distortion predictions are guesses. All of that sounds as if SPICE isn't useful, but it is; it's useful for checking to see whether a circuit is worth testing in hardware. Having said that, I have made things work in hardware that didn't work in SPICE. SPICE is a useful tool but needs to be used carefully.
Hi
I assume that the sine generator defaults to zero series resistance and zero parallel capacitance if those windows are left blank? That is how BC did it. Just to verify, I set those as zero and rechecked the THD numbers with no change to what I reported above.
The net labeled "Vin" is the generator output.
Checking other all-BJT circuits, the generator THD is down in the ppb or off the scale; if there are jfets, mosfets or tubes in the circuit, gen THD is much higher. This still seems strange but falls in line with the optimisation comments above.
I also know that LTspice was originally designed to help with SMPS design and used to be called a power supply simulator 🙂
Thanks for all the input
I assume that the sine generator defaults to zero series resistance and zero parallel capacitance if those windows are left blank? That is how BC did it. Just to verify, I set those as zero and rechecked the THD numbers with no change to what I reported above.
The net labeled "Vin" is the generator output.
Checking other all-BJT circuits, the generator THD is down in the ppb or off the scale; if there are jfets, mosfets or tubes in the circuit, gen THD is much higher. This still seems strange but falls in line with the optimisation comments above.
I also know that LTspice was originally designed to help with SMPS design and used to be called a power supply simulator 🙂
Thanks for all the input