Amplifier cannot be heard. You can hear how the amplifier works in conjunction with specific speakers or headphones. With a specific source. With specific connecting cables and connectors. With certain conditions of electromagnetic interference and mains interference. etc.
Sometimes you can hear a transformer humming or an SMPS beeping
That appears where your experience differs substantially from mine.
Personally I agree with Bob Carver when he said that 90% of making a ss amp sound like a tube amp is to drastically reduce the DF.
Also why I have never heard a conventional passive multiway speaker produce an accurate bass in over 40 years.
Ebers-Moll is definitely a large-signal model. Small signal parameters are derived from it by linearizing (assuming the slope is linear) at the operating point of collector current and Vce.
Of course, it is very simplified model that doesn't account for a lot of a transistor's behavior. More complete models are Gummel-Poon and Mextram, which are used by SPICE and simular simulators. The simulators do the same thing. Transient and DC performance is determined from the large signal model. Small-signal "AC" performance is determined from a linearization of parameters (including capacitance) at the DC operating point.
Is it possible to link to such an amplifier, with such a topology ? Any similar one . I'm not familiar with this abbreviation very well.
ss = solid state
df = damping factor
No particular topology is implied, but assume a standard topology such as used by Doug Self in his "Blameless class-B amplifier"
Distortion In Power Amplifiers
df = damping factor
No particular topology is implied, but assume a standard topology such as used by Doug Self in his "Blameless class-B amplifier"
Distortion In Power Amplifiers
To paraphrase Bob Carver: How to make a transistor amplifier sound like a tube amplifier: Degrade its performance to match that of a tube amp.
Step 1) Drop the output impedance to mimic the output transformer. (90% of the mod)
Step 2) Increase the distortion, especially even order, because tube amps have a lot of that. Tubes are horribly non-linear, just like transistors, as are output transformers.
To reduce only the damping factor, and not change other amplifier parameters, you would have to add a low value resistor
in series with the amplifier's output. This will certainly change the sound with most speakers, due to the voltage divider action
with the variable speaker impedance. An example of the effect with a low damping factor amplifier, with a simulated speaker load:
Air Tight ATM-211 tube monoblock power amplifier Measurements | Stereophile.com
Last edited:
Lol, you could use that frequency response to plot loudspeaker impedance curves.
Yes, in fact you can plot speaker impedance curves by directly driving the speaker with a signal generator using a series 1k (or higher) resistor.
That forms a quasi-constant current source, so the voltage across the speaker is approximately equal to the speaker's impedance
at a given frequency times (Vgenerator/1k).
That forms a quasi-constant current source, so the voltage across the speaker is approximately equal to the speaker's impedance
at a given frequency times (Vgenerator/1k).
if you lower the Beta to match Mu transistors are more linear then tubes.
A KSC1845 set up for a gain of 10 outperforms most tubes for linearity.
it is the simplest form of negative feedback
if you are going to compare output levels, make V+ = B+ and adjust the circuit as nesscessary.
A KSC1845 set up for a gain of 10 outperforms most tubes for linearity.
it is the simplest form of negative feedback
if you are going to compare output levels, make V+ = B+ and adjust the circuit as nesscessary.
Last edited:
You can use any value resistor if you use the impedance voltage divider equation to calculate the speaker's impedance. When I had an Audio Precision, I would just use the 600 ohm series setting on the source and plot directly.
Z spkr = Zsrc*Vspkr / (Vsrc - Vspkr)
This yields magnitude only, but many speaker CAD tools will infer the phase from the magnitude vs freq.
But I was only making a joke, which you clearly understood.
Seems Bob Carver found a rather simple way to emulate a tube amp with ss. However, compensating any system for issues found from measurements of the acoustic output is not that simple, a lot of people simply choose to ignore the elephant in the room for one reason or another.
If you are wondering how with nonlinear elements as BJT can end with so low distortion, the topology that the designers ended up with is differential IPS, Miller VAS, and triple EF OS. The triple emitter followers bring the load impedance to near 1 Mohms to load the VAS, the VAS due to its local Miller feedback provides not only low distortion but also low impedance to drive this 1Mohms load providing very high voltage gain. The differential stage has inherent even order harmonic cancelation generated due to exponential character of each unit. The overall feedback due to high voltage gain with little distortion obtained by the VAS, vanishes the remaining nonlinearities.
Last edited:
Hi edbarx,
I have a new spreadsheet for beta testing here (download to run, use OpenOffice)
PAK402-spreadsheet-PP-CC-wingspread.ods - Google Drive
It uses the Schotky equation to model basic BJT's and plots the wingspread curves for the basic Follower power stage (no driver stage yet). It is (almost) the large signal solution for Common Collector (but without the Vce saturation of the Ebers-Moll model). It is based on the W-function which is evaluated to an accuracy of 0.3% and has been verified against LTspice using the exact same basic BJT model.
You can change the bias voltage and emitter resistance and the Schotky 'IS' parameter and see the plots change. At this stage it is only for matched PNP/NPN but a later version may allow plots with mismatch.
Another thread here Installing and using LTspice IV (now including LTXVII). From beginner to advanced.
has a similar interest in wingspread plots. I mentioned that temperature variations in power amplifiers mean it is not practical to maintain the right bias point - the bias point wonder all over the place with real music. Simulations and bench testing doesn't show us what happens in reality.
.
Last edited:
"... distortions arise from the improper use of the transistor as an amplifying device. The fact is that the transistor is a current amplifier, and it is forced to perform the functions of a voltage amplifier unusual for it.
In voltage amplifiers, the signal to the transistor is supplied from a source with a low internal resistance, i.e. from a voltage generator. As a result, the entire signal voltage drops at the input resistance of the transistor and its base current is entirely determined by the value of the input resistance. And since this value at the initial section of the input characteristic is very large, the base current is extremely small. Only when the value of the input signal exceeds the turn-on threshold of the transistor (approximately 0.6 V for silicon transistors) and the input resistance decreases to a few kΩ or less, the current in the base circuit begins to increase.
In current amplifiers, the signal to the transistor is supplied from a source with a high internal resistance, i.e. from the current generator. In this case, the current in the base circuit of the transistor depends little on the input resistance and is mainly determined by the internal resistance of the current source. The curve of the dependence of the collector current on the base current passes through the origin and is almost linear in the initial section. Much has been written about this at one time:
J. Budinsky. Amplifiers of low frequency on transistors. - M .: Energiya, 1963.
K. Kachurin. Current control of the final stage of LF amplifiers. - Radio, 1967, No. 9, p. 32, 33.
V. Demyanov. Broadband amplifiers on triodes. Radio, 1966, No. 10, p. 50-53.
V. Demyanov, I. Akulinichev. Resonant amplifiers on tubes and transistors. - M .: Energiya, 1970.
S. Biryukov. Low frequency power amplifiers. USSR author's certificate No. 315267, class H03F3 / 18. - Bulletin "Discoveries, inventions, trademarks", 1971, No. 28.
Author: Dorofeev M., based on materials: magazine "Radio" No. 3, 1991
In voltage amplifiers, the signal to the transistor is supplied from a source with a low internal resistance, i.e. from a voltage generator. As a result, the entire signal voltage drops at the input resistance of the transistor and its base current is entirely determined by the value of the input resistance. And since this value at the initial section of the input characteristic is very large, the base current is extremely small. Only when the value of the input signal exceeds the turn-on threshold of the transistor (approximately 0.6 V for silicon transistors) and the input resistance decreases to a few kΩ or less, the current in the base circuit begins to increase.
In current amplifiers, the signal to the transistor is supplied from a source with a high internal resistance, i.e. from the current generator. In this case, the current in the base circuit of the transistor depends little on the input resistance and is mainly determined by the internal resistance of the current source. The curve of the dependence of the collector current on the base current passes through the origin and is almost linear in the initial section. Much has been written about this at one time:
J. Budinsky. Amplifiers of low frequency on transistors. - M .: Energiya, 1963.
K. Kachurin. Current control of the final stage of LF amplifiers. - Radio, 1967, No. 9, p. 32, 33.
V. Demyanov. Broadband amplifiers on triodes. Radio, 1966, No. 10, p. 50-53.
V. Demyanov, I. Akulinichev. Resonant amplifiers on tubes and transistors. - M .: Energiya, 1970.
S. Biryukov. Low frequency power amplifiers. USSR author's certificate No. 315267, class H03F3 / 18. - Bulletin "Discoveries, inventions, trademarks", 1971, No. 28.
Author: Dorofeev M., based on materials: magazine "Radio" No. 3, 1991
Charge storage is a big part of BJT distortion in many early designs and badly designed ones now. They just don't turn off nicely.
I have owned poor sounding amplifiers in the distant past and they suffered badly from high order distortion at the frequency extremes.
Valve amplifiers tend to have far worse measured specs at low frequencies, they have nothing like constant DF and low order harmonic distortion with frequency
I have owned poor sounding amplifiers in the distant past and they suffered badly from high order distortion at the frequency extremes.
Valve amplifiers tend to have far worse measured specs at low frequencies, they have nothing like constant DF and low order harmonic distortion with frequency
If the CB junction is allowed to forward bias, it takes a substantial amount of time to pull it out of saturation. This is usually only seen when amplifiers clip.
The BE junction is similar. It is usually forward biased (of course) and in class (A)B outputs, the depletion layer needs to be charged for turn off. Just as tricky is turning the BJT back on.
A low impedance drive is usually sufficient. MOSFETs also store a lot of charge that needs removed to turn off the gate. Slow turn-off means increasing output stage current with frequency. Slow turn-on means increasing crossover distortion with frequency.
The BE junction is similar. It is usually forward biased (of course) and in class (A)B outputs, the depletion layer needs to be charged for turn off. Just as tricky is turning the BJT back on.
A low impedance drive is usually sufficient. MOSFETs also store a lot of charge that needs removed to turn off the gate. Slow turn-off means increasing output stage current with frequency. Slow turn-on means increasing crossover distortion with frequency.
Mark Tillotson has laid it out quite succinctly.
BJTs have good linearity if you can
Emitter follower stages and current-driven common-emitter VAS stages both operate on principle (a).
I would add that replacing resistor loads with current sources linearizes each stage, by raising the output impedance by orders of magnitude. The output voltage of a VAS can then swing almost rail to rail while the current stays nearly constant. Triple-EF output stages provide the same benefit to the VAS, by providing a nearly constant load current.
Why is constant current important? Here's the gain of a common emitter amp or diff pair with resistor loading:
If RL is a resistor, then the gain is proportional to Ic, which changes exponentially with the input voltage, ie, it is nonlinear.
If ro is an active load then the gain becomes
The actual equations are a little more complicated, but these are good 1st-order approximations that show that the dependency of gain on signal swing goes away with good current-source loading. Gain is just the ratio between the Early voltage and Vt. Current mirrors also force balance on the input pair, linearizing those better.
That's just the low frequency picture. In reality, the current does vary quite a lot to charge & discharge the compensation cap, but only at high frequencies. This is an argument for having way more slew-rate than you need, so that not much of the available current is needed to slew the VAS output. Or for not using Miller compensation around the VAS.
BJTs have good linearity if you can
(a) Amplify current (current input), or
(b) Amplify voltage with the device at a constant current.
(b) Amplify voltage with the device at a constant current.
Emitter follower stages and current-driven common-emitter VAS stages both operate on principle (a).
I would add that replacing resistor loads with current sources linearizes each stage, by raising the output impedance by orders of magnitude. The output voltage of a VAS can then swing almost rail to rail while the current stays nearly constant. Triple-EF output stages provide the same benefit to the VAS, by providing a nearly constant load current.
Why is constant current important? Here's the gain of a common emitter amp or diff pair with resistor loading:
gm*ro = (Ic/Vt)*RL.
where Vt is the thermal voltage = 26mV * Tkelvin/300
where Vt is the thermal voltage = 26mV * Tkelvin/300
If RL is a resistor, then the gain is proportional to Ic, which changes exponentially with the input voltage, ie, it is nonlinear.
If ro is an active load then the gain becomes
Av = gm * ro_npn||ro_pnp
Av = (Ic/Vt) * (Va_npn/Ic)||(Va_pnp/Ic)
Av = (Va_npn||Va_pnp)/Vt
whereAv = (Ic/Vt) * (Va_npn/Ic)||(Va_pnp/Ic)
Av = (Va_npn||Va_pnp)/Vt
ro_npn = output impedance of the NPN
ro_pnp = output impedance of the PNP
Va_npn = NPN Early voltage
Va_pnp = PNP Early voltage
and || means "in parallel" or a*b/(a+b)
ro_pnp = output impedance of the PNP
Va_npn = NPN Early voltage
Va_pnp = PNP Early voltage
and || means "in parallel" or a*b/(a+b)
The actual equations are a little more complicated, but these are good 1st-order approximations that show that the dependency of gain on signal swing goes away with good current-source loading. Gain is just the ratio between the Early voltage and Vt. Current mirrors also force balance on the input pair, linearizing those better.
That's just the low frequency picture. In reality, the current does vary quite a lot to charge & discharge the compensation cap, but only at high frequencies. This is an argument for having way more slew-rate than you need, so that not much of the available current is needed to slew the VAS output. Or for not using Miller compensation around the VAS.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.