G’day all.
Here is a design for a high power (500W RMS per channel) class A amplifier I’ve come up with. I’m planning to start building this amp up rather soon as the “ultimate” amplifier for my lounge room audio system.
The design is based on a differential op-amp topology with a common mode voltage servo to drive the speaker differentially (or bridged) so as to take advantage of even harmonic cancellation and slew rate doubling.
The most notable thing about it though is the use of two pairs of parallel connected Class B / Class A output stages to drive the speaker. The Class B output stages drive floating power supplies for the Class A output stages.
This makes a 500W RMS pure class A amplifier which only dissipates 320W of idle current.
I’ve scribbled a block diagram of the idea down and attached it below.
Just interested in some feedback on what other’s think of my idea, of if anyone out there has seen or come up with and/or built something along similar lines.
Cheers,
Glen
Here is a design for a high power (500W RMS per channel) class A amplifier I’ve come up with. I’m planning to start building this amp up rather soon as the “ultimate” amplifier for my lounge room audio system.
The design is based on a differential op-amp topology with a common mode voltage servo to drive the speaker differentially (or bridged) so as to take advantage of even harmonic cancellation and slew rate doubling.
The most notable thing about it though is the use of two pairs of parallel connected Class B / Class A output stages to drive the speaker. The Class B output stages drive floating power supplies for the Class A output stages.
This makes a 500W RMS pure class A amplifier which only dissipates 320W of idle current.
I’ve scribbled a block diagram of the idea down and attached it below.
Just interested in some feedback on what other’s think of my idea, of if anyone out there has seen or come up with and/or built something along similar lines.
Cheers,
Glen
An externally hosted image should be here but it was not working when we last tested it.
theChris said:
G'day.
I actually considered using a class D amp for the supply rails, but decided aqainst it on grounds of added complexity. The extra watts dissipated by a class B stages isnt much, compared with the 320W being pumped out by the class A stages!
And a question for anyone who may know the answer - what's the typical primary to secondary winding capacitance of a typical 150VA 240V-9V+9V toroidal transformer??????????
Cheers,
Glen
I would go with a full bridge instead of couple of amps with a common diffamp with symmetrical outputs and strange common feedback. And I would not ever think about noises of the amp with +4 dB (or even 0 dB) input level.
12ax7 may give you more of equal gain on wider band with less audible distortions. But anyway, I would use something more powerful to drive transistors (less of them will be needed).
And... I don't think that class A and B amps driving output in series with each other (instead of when class D controls supply voltage) is a good idea.
12ax7 may give you more of equal gain on wider band with less audible distortions. But anyway, I would use something more powerful to drive transistors (less of them will be needed).
And... I don't think that class A and B amps driving output in series with each other (instead of when class D controls supply voltage) is a good idea.
G'day
I would go with a full bridge instead of couple of amps with a common diffamp with symmetrical outputs and strange common feedback.
I’ve used this topology before in a bridged class AB amp with very good results. And I like to do something a bit different. The common mode feedback servo plays no role in the audio amplification, it just keeps the nominal DC potential between the outputs near zero volts, just like a DC servo in a normal amp. In a normal amp the servo is optional, but in a fully differential op-amp it is a necessity for setting the DC operating point, as the global feedback acts differentially, not with respect to a single output and ground.
Texas Instruments have a really good explanation of the differential op-amp topology here:
http://focus.ti.com/lit/an/slyt165/slyt165.pdf
And I would not ever think about noises of the amp with +4 dB (or even 0 dB) input level.
OK, I’m still interested in the noise characteristics of a 12AX7 out of curiosity though.
12ax7 may give you more of equal gain on wider band with less audible distortions. But anyway, I would use something more powerful to drive transistors (less of them will be needed).
I’m only planning on using a pair of 12AX7’s at the input – one as the differential input pair and the other as a cascode for the other one. They wont be driving the transistors – the circuit I’ve drawn up is a LOT more complex than that.
And... I don't think that class A and B amps driving output in series with each other (instead of when class D controls supply voltage) is a good idea.
Why??????
Properly implemented, I think it will work sweet. I’ve since made a significant topological change to my diagram above though – I’ll be driving the Class B stages from the output of the Class A stages instead of parallelling the inputs so as to prevent the class B stages loading the VAS stages.
Cheers,
Glen
I would go with a full bridge instead of couple of amps with a common diffamp with symmetrical outputs and strange common feedback.
I’ve used this topology before in a bridged class AB amp with very good results. And I like to do something a bit different. The common mode feedback servo plays no role in the audio amplification, it just keeps the nominal DC potential between the outputs near zero volts, just like a DC servo in a normal amp. In a normal amp the servo is optional, but in a fully differential op-amp it is a necessity for setting the DC operating point, as the global feedback acts differentially, not with respect to a single output and ground.
Texas Instruments have a really good explanation of the differential op-amp topology here:
http://focus.ti.com/lit/an/slyt165/slyt165.pdf
And I would not ever think about noises of the amp with +4 dB (or even 0 dB) input level.
OK, I’m still interested in the noise characteristics of a 12AX7 out of curiosity though.
12ax7 may give you more of equal gain on wider band with less audible distortions. But anyway, I would use something more powerful to drive transistors (less of them will be needed).
I’m only planning on using a pair of 12AX7’s at the input – one as the differential input pair and the other as a cascode for the other one. They wont be driving the transistors – the circuit I’ve drawn up is a LOT more complex than that.
And... I don't think that class A and B amps driving output in series with each other (instead of when class D controls supply voltage) is a good idea.
Why??????
Properly implemented, I think it will work sweet. I’ve since made a significant topological change to my diagram above though – I’ll be driving the Class B stages from the output of the Class A stages instead of parallelling the inputs so as to prevent the class B stages loading the VAS stages.
Cheers,
Glen
G.Kleinschmidt said:
Hmmm... You compensate a common mode drift; who will compensate a differential one?
I still don't understand.
In such case I'd better used transistors, in my mind it is waste of tubes.
Because they are in series, i.e. load between A class amp (without crossover distortions) and class B amp (with crossover distortions), the result is all crossover distortions on the load.
Hmmmm... Power supply with crossover distortions... If will never sound sweet...
Look at your idea from a slightly different perspective.
It will actually split voltages between output transistors. You don't need your "powering" transistors work in class b, instead connect your power sources in series (they really are), and use transistors as emitter followers that regulate voltage that power your class A amp.
The same topology, slightly different arrangement. Used long time ago to produce more output voltage from opamps. However, such phase-shifted positive feedback is another story...
A picture says a thousand words……….
I’ve just finished drawing a preliminary schematic for one channel, so here it is. This has just come fresh from my brain to the computer easel and it is not entirely complete – there are various bypass caps and other stuff still to add and there may be some errors.
The two Class B output stages and floating power supplies can be seen at the top. Immediately below in centre is the hybrid vacuum tube / BJT differential amplifier and VAS stages, either side of which are the Class A output stages comprised of multiple complementary feedback pairs. The common mode voltage and DC offset correction servo’s can be seen just off the bottom of the page
This is only about half the circuitry for one channel. It does not include the preamplifier section (which will also be a full differential opamp topology), the +/-40V Class B output stage unregulated power supply, or the +/- 50V, +230V and +/-15V regulated power supplies which will be implemented with discrete circuitry.
Cheers,
Glen
I’ve just finished drawing a preliminary schematic for one channel, so here it is. This has just come fresh from my brain to the computer easel and it is not entirely complete – there are various bypass caps and other stuff still to add and there may be some errors.
The two Class B output stages and floating power supplies can be seen at the top. Immediately below in centre is the hybrid vacuum tube / BJT differential amplifier and VAS stages, either side of which are the Class A output stages comprised of multiple complementary feedback pairs. The common mode voltage and DC offset correction servo’s can be seen just off the bottom of the page
This is only about half the circuitry for one channel. It does not include the preamplifier section (which will also be a full differential opamp topology), the +/-40V Class B output stage unregulated power supply, or the +/- 50V, +230V and +/-15V regulated power supplies which will be implemented with discrete circuitry.
An externally hosted image should be here but it was not working when we last tested it.
Cheers,
Glen
Wavebourne:
Hmmm... You compensate a common mode drift; who will compensate a differential one?
I still don't understand.
The differential offset is dictated by the input differential pair's offset, just like the offset voltage with respect to groung is determined in a normal amplifier.
I'm using a servo to correct for that too!
Wavebourne:
Hmmmm... Power supply with crossover distortions... If will never sound sweet...
I disagree. The class A stages will exhibit a high degree of power supply rail rejection. Very little, if any measurable crossover distortion will make it from the rails to the load.
Hmmm... You compensate a common mode drift; who will compensate a differential one?
I still don't understand.
The differential offset is dictated by the input differential pair's offset, just like the offset voltage with respect to groung is determined in a normal amplifier.
I'm using a servo to correct for that too!
Wavebourne:
Hmmmm... Power supply with crossover distortions... If will never sound sweet...
I disagree. The class A stages will exhibit a high degree of power supply rail rejection. Very little, if any measurable crossover distortion will make it from the rails to the load.
G.Kleinschmidt said:
In such case it have to be very fast.
Wait.. Your class B has non-linear input impedance, right?
Make an error signal from this fact, and feed it to your class A amp.
But anyway... If your amp is so fast to correct such horrible PS errors you don't need class A for the same results.
G.Kleinschmidt said:
It depends on character of what to reject, right? Nearly pure 100 (120 in America) Herz is nothing against sharp crossover peaks.
I still don't understand what for to dissipate so much of precious power... If you already go with floating power supply, use it to power opamps between bases (output of opamp) and emitters (inputs around small equal resistors) of your transistors, so they will be perfectly linear current amplifiers... And drive them by triodes directly, if you want them... At least, you will get very fast and extremely short crossover distortions and nearly an absolute minimum of necessary dissipation.
It depends on character of what to reject, right? Nearly pure 100 (120 in America) Herz is nothing against sharp crossover peaks.
Well of course it does. It also depends on the amplifiers topology too. The Class B driven floating supplies only supply the Class A output stages. All the low level circuitry has independent regulated rails. The commutation of HF distortion on the rails of the Class A output stages to the load will be practically negligible.
I still don't understand what for to dissipate so much of precious power... If you already go with floating power supply, use it to power opamps between bases (output of opamp) and emitters (inputs around small equal resistors) of your transistors, so they will be perfectly linear current amplifiers... And drive them by triodes directly, if you want them... At least, you will get very fast and extremely short crossover distortions and nearly an absolute minimum of necessary dissipation.
Yes, but with Class A I won’t get any crossover distortion. That’s the whole point of Class A. Also, in this design, I’m not interested in an absolute minimum of power dissipation – I’m only interested in lowering it to manageable levels.
The net gm of the many BJT complementary feedback pairs connected in parallel will be so high linearity wont be an issue either.
Well of course it does. It also depends on the amplifiers topology too. The Class B driven floating supplies only supply the Class A output stages. All the low level circuitry has independent regulated rails. The commutation of HF distortion on the rails of the Class A output stages to the load will be practically negligible.
I still don't understand what for to dissipate so much of precious power... If you already go with floating power supply, use it to power opamps between bases (output of opamp) and emitters (inputs around small equal resistors) of your transistors, so they will be perfectly linear current amplifiers... And drive them by triodes directly, if you want them... At least, you will get very fast and extremely short crossover distortions and nearly an absolute minimum of necessary dissipation.
Yes, but with Class A I won’t get any crossover distortion. That’s the whole point of Class A. Also, in this design, I’m not interested in an absolute minimum of power dissipation – I’m only interested in lowering it to manageable levels.
The net gm of the many BJT complementary feedback pairs connected in parallel will be so high linearity wont be an issue either.
Bit of an update.......
I got a bit carried away while drawing the schematic; upped the power specification to 1kW per channel. That's probably a bit much for my loungeroom, so I've gone back to the original, much more sensible spec of 500W RMS per channel into 4 ohms.
Decided to use heftier output devices too, so as to keep the required number down, as this thing was just getting a bit too complicated.
Here's a full 360 deg (45 deg I phase shift) load line plot for the Class A stages with the SOA for 2 pairs of paralleled output devices. This plot probably wont make much sense until you realise that the supply voltage for the Class A stages track the output voltage - in other words, the collector-emitter voltage is always 10V.
Here's the 180 deg (45 deg I phase shift) load line plot appliciable for each Class B output stage which fits inside the SOA of 3 paralleled output pairs:
All, up there's now a total of 20 output devices per channel for a rated 512W RMS into 4 ohms (4 in each Class A stage and 6 in each Class B stage). These plots were generated by the following program I quickly typed up for this project, which runs in Visual Basic V3 (I really hate doing a million calculations manually and plotting the results on graph paper)
Cheers,
Glen
I got a bit carried away while drawing the schematic; upped the power specification to 1kW per channel. That's probably a bit much for my loungeroom, so I've gone back to the original, much more sensible spec of 500W RMS per channel into 4 ohms.
Decided to use heftier output devices too, so as to keep the required number down, as this thing was just getting a bit too complicated.
Here's a full 360 deg (45 deg I phase shift) load line plot for the Class A stages with the SOA for 2 pairs of paralleled output devices. This plot probably wont make much sense until you realise that the supply voltage for the Class A stages track the output voltage - in other words, the collector-emitter voltage is always 10V.
An externally hosted image should be here but it was not working when we last tested it.
Here's the 180 deg (45 deg I phase shift) load line plot appliciable for each Class B output stage which fits inside the SOA of 3 paralleled output pairs:
An externally hosted image should be here but it was not working when we last tested it.
All, up there's now a total of 20 output devices per channel for a rated 512W RMS into 4 ohms (4 in each Class A stage and 6 in each Class B stage). These plots were generated by the following program I quickly typed up for this project, which runs in Visual Basic V3 (I really hate doing a million calculations manually and plotting the results on graph paper)
Sub Form_Paint ()
'REM BASIC MJL21193/MJL21194 OUTPUT STAGE CALCULATOR
'REM GLEN KLEINSCHMIDT 2006
vsupply = 42 'REM SUPPLY VOLTAGE (volts)
Trackingsupply = 0 'REM SUPPLY VOLTAGE TRACKING? (1=yes, 0=no)
vpeakload = 32 'REM PEAK LOAD VOLTAGE (volts)
loadimp = 2 'REM LOAD RES (ohms)
iphaselag = 45 'REM CURRENT PHASE SHIFT (deg)
emitterohms = 0 'REM EMITTER BALLAST (ohms)
transpairs = 3 'REM NUMBER OF PARALLEL PAIRS
bias = 0 'REM BIAS CURRENT (Amps)
cycles = .5 'REM CYCLES
vscalefactor = 10
hscalefactor = 4
'REM PLOT LOAD LINE
Line (180, 10)-(780, 410), RGB(100, 100, 100), BF
imaxload = vpeakload / loadimp
For iphase = 0 To (360 * cycles) Step .1
iloadinst = Sin(iphase * (3.1415927 / 180)) * imaxload
IEinst = bias + (Sin(iphase * (3.1415927 / 180)) * (imaxload - bias))
vloadinst = Sin((iphase + iphaselag) * (3.1415927 / 180)) * vpeakload
vemitterohmsinst = (iloadinst / transpairs) * emitterohms
If Trackingsupply = 0 Then
vce = vsupply - (vloadinst + vemitterohmsinst)
Else
vce = vsupply
End If
PSet (180 + (vce * hscalefactor), 410 - (IEinst * vscalefactor)), RGB(255, 255, 0)
Next
'REM PLOT MJL21193/4 * n 1 SEC SOA TC=25degC
For v = 200 To 0 Step -5
If v = 0 Then a = 16
If v = 5 Then a = 16
If v = 10 Then a = 16
If v = 15 Then a = 12.7
If v = 20 Then a = 10.1
If v = 25 Then a = 8.1
If v = 30 Then a = 7
If v = 35 Then a = 6
If v = 40 Then a = 5.2
If v = 45 Then a = 4.5
If v = 50 Then a = 4
If v = 55 Then a = 3.6
If v = 60 Then a = 3.27
If v = 65 Then a = 2.95
If v = 70 Then a = 2.65
If v = 75 Then a = 2.4
If v = 80 Then a = 2.2
If v = 85 Then a = 2
If v = 90 Then a = 1.863
If v = 95 Then a = 1.73
If v = 100 Then a = 1.6
If v = 105 Then a = 1.5
If v = 110 Then a = 1.383
If v = 115 Then a = 1.29
If v = 120 Then a = 1.217
If v = 125 Then a = 1.143
If v = 130 Then a = 1.06
If v = 135 Then a = .995
If v = 140 Then a = .9315
If v = 145 Then a = .885
If v = 150 Then a = .8285
If v = 155 Then a = .783
If v = 160 Then a = .738
If v = 165 Then a = .705
If v = 170 Then a = .668
If v = 175 Then a = .641
If v = 180 Then a = .61
If v = 185 Then a = .59
If v = 190 Then a = .5731
If v = 195 Then a = .566
If v = 200 Then a = .566
a = a * transpairs
If v = 200 Then PSet (180 + (hscalefactor * v), 410 - (vscalefactor * a)), RGB(0, 255, 0)
Line -(180 + (hscalefactor * v), 410 - (vscalefactor * a)), RGB(0, 255, 0)
Next
End Sub
Cheers,
Glen
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.