Slew rate is the wrong way to start . You'll need to start with the open loop bandwidth that is the gist of all. Next you'll have to look at stability without
any kind of overall feedback. That accomplished take care for sources and mechanisms of distortion.
This is just the path that I follow. Those thermaltrak power BJTs offer at least as a promise a stunningly linear ic/ib characteristic such linearity is unmatched by any FET. Thus the design objective is to make extensive use of this "innate" linearity and employ current drive throughout thus that the
exponential ic/Vbe never applies. If necessary combine current with current feedback.
Hi guys 
SSA BIGBT HP output resistance is rock solid 5,6 mohm at 20 kHz and it slowly goes down to 4 mohm at 100 Hz. So calculated damping factor is 715 at 4 ohm at 20 kHz, which is not bad at all.
Also measured a 40 Vpp square wave signal from 1 to 100 kHz and there's not a slight change to the signal form when 4 ohm load was connected. Since the amp has no output serial inductor I postponed parallel to load capacitor test. First have to make some 1 uH inductor ... till next time.
No part's values changed, measurements are too good to change anything so far.
P.S. Would I gain anything if I lower/adjust 100 ohm IRF's gate resistors?
P.S.S. On the secondary high level DC output of my DIY signal generator a 10 MHz ringing is present on squarewave. And the same copy of the amplified input signal was at the output - 10 MHz ringing went out gained.
SSA BIGBT HP output resistance is rock solid 5,6 mohm at 20 kHz and it slowly goes down to 4 mohm at 100 Hz. So calculated damping factor is 715 at 4 ohm at 20 kHz, which is not bad at all.
Also measured a 40 Vpp square wave signal from 1 to 100 kHz and there's not a slight change to the signal form when 4 ohm load was connected. Since the amp has no output serial inductor I postponed parallel to load capacitor test. First have to make some 1 uH inductor ... till next time.
No part's values changed, measurements are too good to change anything so far.
P.S. Would I gain anything if I lower/adjust 100 ohm IRF's gate resistors?
P.S.S. On the secondary high level DC output of my DIY signal generator a 10 MHz ringing is present on squarewave. And the same copy of the amplified input signal was at the output - 10 MHz ringing went out gained.
Last edited:
P.S. Would I gain anything if I lower/adjust 100 ohm IRF's gate resistors?
If you can get away with lowering that value you will increase the amps stability with capacitive load.
On the Laterals on my DC linked Fetzilla, with a very tight layout, I'm managing with 22R gate stoppers and it sounds great.
You just have to see what you can manage.
mike
If you can get away with lowering that value you will increase the amps stability with capacitive load.
On the Laterals on my DC linked Fetzilla, with a very tight layout, I'm managing with 22R gate stoppers and it sounds great.
You just have to see what you can manage.
mike
Thanks Mike
I will solder 200 ohm trimmers parallel to gate resistors and slowly decrease resistance by measuring 100 kHz squarewaves on 4 ohms on the output.
Is there any possibility to fall into bad uncontrolled full scale oscillations or they appear gradually?
I found that any complicated test conditions like rats nest wiring prevented best results.
In the end I just made the best possible layout with very short wires and tried different ( non inductive ) values.
My symptom was generally about 16Mhz @ about 1V - sometimes worse - but never full scale.
In the end I just made the best possible layout with very short wires and tried different ( non inductive ) values.
My symptom was generally about 16Mhz @ about 1V - sometimes worse - but never full scale.
This gets quite interesting. Thermal stability is given with thermal track diodes
contolling the bias of Q3 Q11. The op point of the power trans can be set as low as 10 mA. Difficulty in practice is to find complementary pair Q3 Q11 in terms of switching speed. While crossover distortion in the power stage cannot happen as this is always conducting, it can still occur in Q3 Q11.
Hi guys
SSA BIGBT HP output resistance is rock solid 5,6 mohm at 20 kHz and it slowly goes down to 4 mohm at 100 Hz. So calculated damping factor is 715 at 4 ohm at 20 kHz, which is not bad at all.
Also measured a 40 Vpp square wave signal from 1 to 100 kHz and there's not a slight change to the signal form when 4 ohm load was connected. Since the amp has no output serial inductor I postponed parallel to load capacitor test. First have to make some 1 uH inductor ... till next time.
No part's values changed, measurements are too good to change anything so far.
P.S. Would I gain anything if I lower/adjust 100 ohm IRF's gate resistors?
P.S.S. On the secondary high level DC output of my DIY signal generator a 10 MHz ringing is present on squarewave. And the same copy of the amplified input signal was at the output - 10 MHz ringing went out gained.
Be carefull with Extreme wide bandwidth. You Will pickup a lot of unwanted in your signal cable. In My experience ADD a lowpass filter at the input and dont change the circuit.
Regarding gate stoppers. I Will get back tomorrow.
Thanks Mike
I will solder 200 ohm trimmers parallel to gate resistors and slowly decrease resistance by measuring 100 kHz squarewaves on 4 ohms on the output.
Is there any possibility to fall into bad uncontrolled full scale oscillations or they appear gradually?
It depends on overall circuit layout. To small value Can make them oscillate with trace inductance
I have an input low-pass filter, formed by 47 ohm/100 pF as close to the bases as possible, from the early beginning. Without this filter it is really unsafe to run such fast amp cause there were a lot of RFI-s at the output present, one even as high as 138 MHz at mV levels.
Slew rate is a *direct* consequence of the open loop bandwidth and do not change with feedback value. Talking slew rate IS talking open loop bandwidth.
What i try is to make each pole of an amp a little faster than the previous one. And of course, never let the drivers running out of current ;-)
As says L.C., we alway need to low pass filter the incoming signal. To remove hf parasitics, high order distortion signals from previous stages and to make sure the slew rate of the incoming signal is alway far less than the one of the amp ( divided by his overall gain).
On the contrary, to avoid TIM, slower is you amp, more you have to filter the incoming signal. Faster is you amp less sensible it will be to hf low level incoming signals, as it will be able to amplify them adding no IM distortion with the audio signal.
Your speakers or your ears are not sensible to little hf as long as they had not deteriorated the audio in the active stages (assuming low hf level, of course).
My way is to linearise bandwith with CR comp and linearise ovesrshoots of square waves with the input filter. By habit it is enough to filter hf too. But you can go further (Try and listen).
It is too interesting to shield the loudspeakers cables and filter HF in the output lines, to avoid hf returns in the CR from there. The LC network on the output help all along with low impedance on that matter too.
PS to L.C: My protection is good too in detecting any dangerous hf signals in the input.
Last edited:
"Slew rate is a *direct* consequence of the open loop bandwidth and do not change with feedback value. Talking slew rate IS talking open loop bandwidth."
This is not correct you'll need to consider small signal and large signal behavior.
In the latter case FETs have a distinct advantage over BJTs.
This is not correct you'll need to consider small signal and large signal behavior.
In the latter case FETs have a distinct advantage over BJTs.
Hi hahfran
So far SSA was built by Nico Ras, Salas, Bigun (TGM5 all BJT SSA), Shaan, Esperado (Le crescendo revisité) and me. They all reported that it sounds correctly, very pleasant to listen to. It is not a big deal to make one channel according to your own version of sch and test it throughout.
So far SSA was built by Nico Ras, Salas, Bigun (TGM5 all BJT SSA), Shaan, Esperado (Le crescendo revisité) and me. They all reported that it sounds correctly, very pleasant to listen to. It is not a big deal to make one channel according to your own version of sch and test it throughout.
Hi hahfran
So far SSA was built by Nico Ras, Salas, Bigun ( It is not a big deal to make one channel according to your own version of sch and test it throughout.
Fear not! I am almost through with an all BJT version of a non switching class B
Btw. my experience with MFET driven by high impedance cascode is: it will oscillate.... especially under varying load impedance such as multi way speakers with weird passive crossovers...
I'll keep my hands off power MFets.
There is a lot more to say about current feedback and voltage feedback.
Driving passive multiway speakers with passive crossovers ... what designers never report is reactive distortion. That is distortion for a real load tells little about distortion on complex load varying with frequency.
The latter is much higher. In this case current fb is better but no feedback is best!
However another kind is dampened oscillation on transients also under complex load. here voltage fb is better than cfb.
Well amps are still subject of engineering sciences.
A few critical remarks on SSA
The original version with MFETS appears to deserve some criticisms...
the complementary cascodes are not biased in class A but rather at best AB.
On idle, the currents bias the FETs in AB. Then one cascode turns off pinching the FET off, too. Thus there are two crossover involved and one is unnecessary, the FETs should be kept conducting always.
Since FETS are slow in the transition from pinched off to turned on by their semiconductor physics "nature", and further highly non linear, leaving the FETS both always conducting makes the amp faster and a bit more linear.
That is the idea behind my first idea shown earlier with BJTs. If these are all current driven, the gain in speed and linearity is even better then for the same design with FETs, if the triple power cascade is always conducting.
The trouble I still have is to build in the thermal tracking. I have an idea,
that is to place the diodes out of the signal path in CS in order to keep the
always condcuting bias independent of junction temperature.
Stay tuned.
The original version with MFETS appears to deserve some criticisms...
the complementary cascodes are not biased in class A but rather at best AB.
On idle, the currents bias the FETs in AB. Then one cascode turns off pinching the FET off, too. Thus there are two crossover involved and one is unnecessary, the FETs should be kept conducting always.
Since FETS are slow in the transition from pinched off to turned on by their semiconductor physics "nature", and further highly non linear, leaving the FETS both always conducting makes the amp faster and a bit more linear.
That is the idea behind my first idea shown earlier with BJTs. If these are all current driven, the gain in speed and linearity is even better then for the same design with FETs, if the triple power cascade is always conducting.
The trouble I still have is to build in the thermal tracking. I have an idea,
that is to place the diodes out of the signal path in CS in order to keep the
always condcuting bias independent of junction temperature.
Stay tuned.
My experience with these two IRF-s as drivers are nothing but joy. No sign of that kind of behaviour you mentioned. On contrary I will go to no gate stoppers as I did in my MonoLith amp long time ago. Will do report later ...
VAS cascodes has 10 mA collector bias current, I don't think they ever go to off state ...
P.S. Vbe multiplier voltage is always present and quite constant, so no sign of FET crossover distortion. FET bias is independant from output stage and can be several hundred mA if you like. Voltage part of the circuit also has +/-10 V more PSU potential than the output stage.
Last edited:
If, by "correctly" you mean the best amp i ever listened to, i agree...
to hahfran: (Tired to argue), Slew rate is rail to rail, by definition.
If you read Hahfran's posts in the past, you will find out that he is one of those who favor or highly regard a "tube like" sound. For those who likes this kind of amp or music (i.e melow, tube like, etc), slew rate is not important, or it is never a part of what is required to achieve what they want to hear.
But if you are one of those who found that most tube amps lack something, and you like to hear "fast" music such as rock or certain classics, you will find out that slew rate is critical.
So it is more about one's priority and preference over music.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.