then why did you state "one pair"?That's why an amp with low count (read one pair only) of output pairs will always (soundwisely) outperform PA like power amplifiers.
I've suggested that a 10pair (20 devices) of high performance "driver" devices be adopted for an optimally biased ClassAB stage with a maximum output of around 100W into 8ohms and ClassA output up to 50W (3.5Apk).
I would never accept that this is a PA style amplifier that is crippled by adopting too many output devices.
I would never accept that this is a PA style amplifier that is crippled by adopting too many output devices.
Small signal square wave tests OR why a zobel is always needed
Just to get back to the essence. The amplifier is pretty stable and playing fine! The latest measurements THD and IMD in combination with hearing test emphasize, that this is an ultra low distortion amplifier even on high output power levels.
What I and many others (which have contributed to this thread) want:
Just to get back to the essence. The amplifier is pretty stable and playing fine! The latest measurements THD and IMD in combination with hearing test emphasize, that this is an ultra low distortion amplifier even on high output power levels.
What I and many others (which have contributed to this thread) want:
- get the 2stageEF simulation as close as possible to real world results
- demystify effects like too less gain margin and so on
- learning in depth
- and very important: have fun
Just to get back to the essence. The amplifier is pretty stable and playing fine! The latest measurements THD and IMD in combination with hearing test emphasize, that this is an ultra low distortion amplifier even on high output power levels.
What I and many others (which have contributed to this thread) want:
- get the 2stageEF simulation as close as possible to real world results
- demystify effects like too less gain margin and so on
- learning in depth
- and very important: have fun
So as it can be understood from your input, this amp is best what it can be produced in DIY world?
No.So as it can be understood from your input, this amp is best what it can be produced in DIY world?
That is not a summary of what astx said.
Using TTC5200 and TTA1943 providing such output power levels we need so much devices to fulfill SOA requirements under any type of reactive load.
Using other and less output devices is possible but the amplifier may behave completely different. E.g.: I have a duplicate of the output stage using MJW3281A and MJW1302A. This one is harder to get stable and simulation showed me why...
To get high Class A output power levels is also very easy with this design. Just tune the bias to your needs ... and to the size of your heatsinks.
Using other and less output devices is possible but the amplifier may behave completely different. E.g.: I have a duplicate of the output stage using MJW3281A and MJW1302A. This one is harder to get stable and simulation showed me why...
To get high Class A output power levels is also very easy with this design. Just tune the bias to your needs ... and to the size of your heatsinks.
Hi Toni
I have had a think about your results.
You say that the little unevenness on the square wave is about 9 MHz and this is close to the simulation at 8.5 MHz.
That concurrence probably doesn't tell us much about the cause of the effect, it does not prove your capacitance measurement is correct, or incorrect.
Those "wiggles" are the tell-tale of a GM that has started to become small, just as a small PM has a characteristic look.
So they will be close to the GM frequency, whatever the cause of the reduction in GM, I expect.
As to the Zobel, I have simulated a distributed version where a resistor and capacitor attach directly to each output transistor emitter, even before the emitter resistor. This allows absolute minimum stray inductance, well under 1 nH.
Finally, did you ever show the final PCB layout for this amp? I can't remember any details.
A careful double sided layout can reduce the trace inductance to much less than your simulated 12 nH, even with 8 pairs of outputs.
I have just started simulations on a 6 pair amp, 12 transistors is coincidentally the same as the number of cylinders in my car - but that doesn't mean it's a truck
Best wishes
David
I have had a think about your results.
You say that the little unevenness on the square wave is about 9 MHz and this is close to the simulation at 8.5 MHz.
That concurrence probably doesn't tell us much about the cause of the effect, it does not prove your capacitance measurement is correct, or incorrect.
Those "wiggles" are the tell-tale of a GM that has started to become small, just as a small PM has a characteristic look.
So they will be close to the GM frequency, whatever the cause of the reduction in GM, I expect.
As to the Zobel, I have simulated a distributed version where a resistor and capacitor attach directly to each output transistor emitter, even before the emitter resistor. This allows absolute minimum stray inductance, well under 1 nH.
Finally, did you ever show the final PCB layout for this amp? I can't remember any details.
A careful double sided layout can reduce the trace inductance to much less than your simulated 12 nH, even with 8 pairs of outputs.
I have just started simulations on a 6 pair amp, 12 transistors is coincidentally the same as the number of cylinders in my car - but that doesn't mean it's a truck
Best wishes
David
Last edited:
Hi Toni
I have had a think about your results.
You say that the little unevenness on the square wave is about 9 MHz and this is close to the simulation at 8.5 MHz.
That concurrence probably doesn't tell us much about the cause of the effect, it does not prove your capacitance measurement is correct, or incorrect.
Those "wiggles" are usually the tell-tale of a GM that has started to become small, just as a small PM has a characteristic look.
So they will be close to the GM frequency, whatever the cause of the reduction in GM, I expect.
As to the Zobel, I have simulated a distributed version where a resistor and capacitor attach directly to each output transistor emitter, even before the emitter resistor. This allows absolute minimum stray inductance, well under 1 nH.
Finally, did you ever show the final PCB layout for this amp? I can't remember if it was double sided, or any details.
A careful double sided layout can reduce the trace inductance to much less than your simulated 12 nH, even with 8 pairs of outputs.
I have just started simulations on a 6 pair amp, 12 transistors is coincidentally the same as the number of cylinders in my car - but that doesn't mean it's a truck
Best wishes
David
Dear David,
cool idea to add "mini"-zobels. Have you a starting recommendation of the used values?
Btw.: would be good to know how a low GM looks in comparison to low PM using the osci. Using TMC here I'm missing practical information on the net (screenshots). With TPC compensation it was easier to detect low GM/PM as there exists some information using the square wave overshoot as indicator.
The used 12nH is from a internet accessible calculator using 25mm of 9mm wide 70µ copper - unfortunately can't access my bookmarks where I found the calculator.
The currently used doublesided 70µ copper PCB layout of the output stage can be downloaded and viewed in post # 674.
So as next project I have to design a small 3 pair amp ... if the number of power bjt's have to match the number of cylinders in my car.
I just start at about 10 ohm + 10 nF in total. So for 2 pairs use 20 ohm + 5 nF, and so on.
"Saturn PCB" have a decent downloadable one. For 4.8 mm trace on 1.6 mm board over a plane I have about 1.8 nH / 10 mm
So about 1 nH / 10 mm for 9 mm wide. Seems inconsistent with your estimate
Thanks
So as next project I have to design a small 3 pair amp ... if the number of power bjt's have to match the number of cylinders in my car.
No, you need to buy a V16 car to match your amp
Best wishes
David
Last edited:
Thanks. Will try this application - hopefully running under wine/linux
Link found: flat_wire_inductance_calculator
2.5cm length / 0.9cm wide / 0.07mm thickness = 11,44nH
But this calculator does not take the ground plane into account...
V16: which car would you recommend to buy?
Another nonsense about this amp.But this calculator does not take the ground plane into account...
Do you mean the layout achieves low stray inductance in the Zobel?
Or that the circuit is more sensitive to stray inductance?
Another 'Zobel' solution worth investigating is the one favoured by Cherry. He puts it AFTER the Inductor/Resistor and uses a pure cap without any series resistance.
I pooh poohed it when I first saw it but it actually has some very nice features ... in particular, the amp is immune to instability with small pure capacitance, which is the usual Achilles heel of other amps.
The CON is that you really need something AT the output devices .. hence Bob Cordell's liking for Zobels at each O/P device.
Toni's Output PCB is double sided with Ground Plane which increases stray capacitance. This may reduce 'effective' stray inductance but the whole thing needs to be looked at carefully.
My $0.02 is not to have a Ground Plane but to run high current tracks on either side in thick parallel tracks in an attempt to achieve twisted pairs.
I'm still trying to get my head around Toni's latest sims vs his 'real life'.
I know the inductance of his ribbon cable is essential to simulating some (but not all) of his overload & slew behaviour.
This amp is full of nonsenses coming from wrong presumption to be better than others. Ground plane, 16 outputs are just some of them.
Yes.
Yes. My idea is basically the same as Bob's pi network double Zobel.
Probably a Cherry style after the inductor, but with a Zobel at each O/P device to minimize stray inductance.
The difference is that I connect them directly to the devices before the emitter resistors.
Also helps spread power dissipation in the Zobel if you really want to run 100 kHz square wave tests to prove... whatever.
Very carefully.
Yes, This is my tactic too. I don't think a plane is optimal when you only have a two layer board, each current needs to be considered and cancelled.
But I used the plane calculation because the answer is very close to the inter-track inductance for a parallel pair.
Best wishes
David
V16: which car would you recommend to buy?
Cizeta 16 would be suitable.
Fast, powerful, excellent performance and hand built, just like your amp
Best wishes
David
Last edited:
...
Cizeta 16 would be suitable.
Fast, powerful, excellent performance and hand built, just like your amp
Best wishes
David
Cool car and thanks,
BTW: maybe they have a Cizeta 4 which performs better as the Cizeta 16 and is much cheaper?