No mystery. Your circuit is wrong. No living amplifier have such response. Its a simple thing. What you need is proper compensation. At minimum you need one additional HF feedback. Can be from VAS (phase lead?), or across the feedback resistor.
There are variations between models (quantity of output MOSFET, voltage levels,...) But basically all they are same as schematic below,
I don't have 100k$ amp on my hand to confirm about a capacitor across feedback resistor. But following GM documents i have, simulation result is always oscillation.
I don't have 100k$ amp on my hand to confirm about a capacitor across feedback resistor. But following GM documents i have, simulation result is always oscillation.
Attachments
Will the ringing in your simulation be reduced if we increase the value of compensation capacitors (4.7pF) ?
Hi Kean,
GM Telos doesn't use MPSA transistors, it is zetex FZT857/FZT957 as SOT-223 package. My simulation use FZT models which downloaded from Zetex page. Maybe there was improvement from Mimesis models to Telos (and next gen) models.
You are correct, it is not yet oscillation, but it is sign of unstable. We usually need an amp which no sign of peaking/ringing/oscillation.
Hi Trung,
To overcome this ringing is easy. We can increase value of Cdom, or add more process method like (MIC, capacitors across feedback resistor...). However the price must pay is lower bandwidth, amp more slow. We know GM by they are fast amp, if they are not fast any more, so?!
But i mean we do not try to clone exactly same as GM schematic because it will conduct to unstableness.
Hi Veysel,
It is maybe you aren't understanding my idea. I don't ask about corrective methods. Just want to say don't make exactly same as schematic due stability issue.
Hi Walkalone,
Like you said, we can increase value of Cdom to damp this ringing at the expense of lower bandwidth. I suspect the manufacture of the GM amp will be like that. First they build an amp with the original Cdom (4.7pF). After that they put the circuit in oscilloscope or analyzer equipment, and change the Cdom value until there are no ringing. Using this method, you can keep the amp stable, and the Cdom will be low enough to not decrease the bandwidth too much. This kind of manufacture process is not feasible for normal industry level, but given the price of GM amp, they have both time, equipment and money available to do so.
Like you said, we can increase value of Cdom to damp this ringing at the expense of lower bandwidth. I suspect the manufacture of the GM amp will be like that. First they build an amp with the original Cdom (4.7pF). After that they put the circuit in oscilloscope or analyzer equipment, and change the Cdom value until there are no ringing. Using this method, you can keep the amp stable, and the Cdom will be low enough to not decrease the bandwidth too much. This kind of manufacture process is not feasible for normal industry level, but given the price of GM amp, they have both time, equipment and money available to do so.
Amp design is science. BTW, why do you think the power supply for the latfet is +/- 80V? What is the benefit running the output close to its maximum capability? There are many who have built the amp, you should be able to find the right circuit.
@ Trung,
Increasing Cdom to 12x original value (same as Veysel opinion), Bode plot looks more great. But BW is only ~550KHz. It means lower 5x spec. which GM has show (2.5MHz@-3dB). Using combination of compensation techniques will get higher BW. But never get numbers of Goldmund.
@ John,
As i said my schematic is general. +-80VDC power supply is applied for models that use 6 pairs LatFET per channel. 2 pairs models are lower (+-60VDC). However Goldmund Telos 590 next gen also uses 55VAC level (+-76VDC?) for 2 pairs LatFET.
6moons audioreviews: Goldmund Telos 590 NextGen
Increasing Cdom to 12x original value (same as Veysel opinion), Bode plot looks more great. But BW is only ~550KHz. It means lower 5x spec. which GM has show (2.5MHz@-3dB). Using combination of compensation techniques will get higher BW. But never get numbers of Goldmund.
@ John,
As i said my schematic is general. +-80VDC power supply is applied for models that use 6 pairs LatFET per channel. 2 pairs models are lower (+-60VDC). However Goldmund Telos 590 next gen also uses 55VAC level (+-76VDC?) for 2 pairs LatFET.
6moons audioreviews: Goldmund Telos 590 NextGen
Attachments
@Walkalone
I think with each PCB, due to variations of component, the Cdom could be smaller than the simulated one, hence the bandwidth could be approach 1MHz.
Even then, I think we don't need to concern about the marketing aspect of audio amplifier. Goldmund can report whatever they want about specs of their amplifier. Due to its price, no one can verify those, especially about bandwidth specs. Considering the fact that audio manufacturer tends to overrate their products's spec, I don't think the spec GM reported is accurate and represent the true spec of their amp.
For a diyer, I think we need to worry about how we can match those 6 pairs per channel of output LFET than concerning about the bandwidth of the amplifier, which is high enough even with high Cdom.
I think with each PCB, due to variations of component, the Cdom could be smaller than the simulated one, hence the bandwidth could be approach 1MHz.
Even then, I think we don't need to concern about the marketing aspect of audio amplifier. Goldmund can report whatever they want about specs of their amplifier. Due to its price, no one can verify those, especially about bandwidth specs. Considering the fact that audio manufacturer tends to overrate their products's spec, I don't think the spec GM reported is accurate and represent the true spec of their amp.
For a diyer, I think we need to worry about how we can match those 6 pairs per channel of output LFET than concerning about the bandwidth of the amplifier, which is high enough even with high Cdom.
Why do Goldmund, and these clones, use two pairs or transistors, Q9, Q10, Q11, Q12, to drive the MOSFETs, where Hafler / Rockford, and others, use one pair to drive all the MOSFETs, even with eight MOSFETs?
And why do you place the 470n across each resistor? Is that related to the charge / discharge of the gate capacitance?
And why do you place the 470n across each resistor? Is that related to the charge / discharge of the gate capacitance?
It's because the drivers have no heatsinks, just ambient convection cooling. 2 more transistors is cheaper than 2 heatsinks plus labor.
The 470n allows the gates to share the complimentary drivers during very fast transients and allows smoother operation when the drivers are briefly pushed into class B operation.
I think it is almost impossible. The people who have GM amp, especially the Telos ones, don't have the equipments to measure it or don't have any reason to measure it. The people who can measure the GM amp either don't buy it or can't afford it
But you cannot simulate a 'in general' schematic and think that the original one have the same response or instability?
I'm pretty sure that the circuit you simulated lack compensation for a reason that you cannot have good sound from this topology only relying on a single VAS compensation (like the one you simulated).
Hi John,
My simulation just draw 2 pairs output to save time, i don't want waste my time to redraw what i knew. Because as i said, i tried many times.
You wonder if six pairs output, does peaking happen? Yes, it still available.
Your confirmation of lack compensation which same as i think, but i can't affirmation due there is no any exhibit. If there are someone of GM team, please confirm.
Hi Trung,
1MHz target presented by single VAS trans is quite hard due Miller effect, at least in simulation. Try to use combination of compensation techniques to expand BW which easily lacking occurred. Needs to improve VAS linearity first (cascode, CPF, beta enhanced...). Hi speed/hi BW is quite important with amps as Goldmund, Spectral... It guarantees the precision of sound.
Matching is not problem now, Exicon has sold matched LMOS following color codes.
My simulation just draw 2 pairs output to save time, i don't want waste my time to redraw what i knew. Because as i said, i tried many times.
You wonder if six pairs output, does peaking happen? Yes, it still available.
Your confirmation of lack compensation which same as i think, but i can't affirmation due there is no any exhibit. If there are someone of GM team, please confirm.
Hi Trung,
1MHz target presented by single VAS trans is quite hard due Miller effect, at least in simulation. Try to use combination of compensation techniques to expand BW which easily lacking occurred. Needs to improve VAS linearity first (cascode, CPF, beta enhanced...). Hi speed/hi BW is quite important with amps as Goldmund, Spectral... It guarantees the precision of sound.
Matching is not problem now, Exicon has sold matched LMOS following color codes.