Edmond,
Looks like Crown uses the kind of output network you recommend on some of their amps. The Microtech 600&1200 has 47nF after 2.7R//1.8uH. Microtech 1000 has 47nF after 2.7R//0.5uH. Macrotech 2400: L = 1.3uH, R and C same.
Most models seem to have 10n caps across both polarity output banks for the output side. On the ground side there are zobels over both polarity banks.
Macrotech 5000VZ has 2.5uH//2.35R and 22nF on output. It also has 5.6R + 47n zobels on both sides of output network. It also has 22nF + 12R zobels over all four banks of output devices.
Crown seems to use a parallell RLC circuit between predrivers and drivers on a lot of their amps instead of just base stopper resistors. Any idea why?
http://www.crownaudio.com/gen_htm/legacy/legacamp.htm
Looks like Crown uses the kind of output network you recommend on some of their amps. The Microtech 600&1200 has 47nF after 2.7R//1.8uH. Microtech 1000 has 47nF after 2.7R//0.5uH. Macrotech 2400: L = 1.3uH, R and C same.
Most models seem to have 10n caps across both polarity output banks for the output side. On the ground side there are zobels over both polarity banks.
Macrotech 5000VZ has 2.5uH//2.35R and 22nF on output. It also has 5.6R + 47n zobels on both sides of output network. It also has 22nF + 12R zobels over all four banks of output devices.
Crown seems to use a parallell RLC circuit between predrivers and drivers on a lot of their amps instead of just base stopper resistors. Any idea why?
http://www.crownaudio.com/gen_htm/legacy/legacamp.htm
The question has come up a couple of times where to put the Zobel network, before or after the coil. Also, someone mentioned the use of a shunt capacitor at the speaker terminals.
First, I believe Bryston uses a shunt capacitor across the loudspeaker terminals. This has some interesting possibilities, but it does seriously mis-terminate a transmission line and possibly invites a resonance.
I like to put a Zobel on both sides of the coil. I think Self frowned on this, and could not understanbd why anyone would do this.
Here is the reasoning. The first Zobel is placed right at the output stage and assures that the output stage will be loaded resistively at high frequencies no matter what, up to quite high frequencies. This Zobel is the one that works to prevent HF oscillations of an un-loaded emitter follower. With high-ft output transistors, it can be important that it be close by, especially if there is inductance in the output lead before it gets to the coil and second zobel.
The second zobel is also a series-RC and it is right at the loudspeaker terminals. Its most important job is the high-frequency termination of the speaker cable transmission line. It wants to have that resistor in it so that the termination looks resistive to very high frequencies, and will damp any attempts at resonance.
Notice that all three resistors in the complete output network work together to damp resonances.
With the dual-zobel approach, I have found that relatively light zobel networks can be used. This also works to spread out any power dissipation that the Zobel might have to handle in 20 kHz full-power testing. The bottom line is two more components than traditional, but they can be smaller in size.
Depending on the design, dual shunt zobels as light as 0.01 uF and 33 ohms might be fine. RF ingress will likely be a bit lower if the resistor in the second zobel is made smaller, however, maybe on the order of 10 ohms. For the coil, I like to keep it between 0.5 u and 1 uH, with a shunt resistor between 1 ohm and 5 ohms. I tend to prefer smaller shunt resistances across the coil.
There are many combinations of values that will work, and I am honestly not sure there is a right and wrong combination of values as long as the complete output network does its intended jobs and does minimal damage to the pulse response.
Cheers,
Bob
First, I believe Bryston uses a shunt capacitor across the loudspeaker terminals. This has some interesting possibilities, but it does seriously mis-terminate a transmission line and possibly invites a resonance.
I like to put a Zobel on both sides of the coil. I think Self frowned on this, and could not understanbd why anyone would do this.
Here is the reasoning. The first Zobel is placed right at the output stage and assures that the output stage will be loaded resistively at high frequencies no matter what, up to quite high frequencies. This Zobel is the one that works to prevent HF oscillations of an un-loaded emitter follower. With high-ft output transistors, it can be important that it be close by, especially if there is inductance in the output lead before it gets to the coil and second zobel.
The second zobel is also a series-RC and it is right at the loudspeaker terminals. Its most important job is the high-frequency termination of the speaker cable transmission line. It wants to have that resistor in it so that the termination looks resistive to very high frequencies, and will damp any attempts at resonance.
Notice that all three resistors in the complete output network work together to damp resonances.
With the dual-zobel approach, I have found that relatively light zobel networks can be used. This also works to spread out any power dissipation that the Zobel might have to handle in 20 kHz full-power testing. The bottom line is two more components than traditional, but they can be smaller in size.
Depending on the design, dual shunt zobels as light as 0.01 uF and 33 ohms might be fine. RF ingress will likely be a bit lower if the resistor in the second zobel is made smaller, however, maybe on the order of 10 ohms. For the coil, I like to keep it between 0.5 u and 1 uH, with a shunt resistor between 1 ohm and 5 ohms. I tend to prefer smaller shunt resistances across the coil.
There are many combinations of values that will work, and I am honestly not sure there is a right and wrong combination of values as long as the complete output network does its intended jobs and does minimal damage to the pulse response.
Cheers,
Bob
I have a question for all those people intentionally skipping the RC and/or RL output networks...
How else do you achieve predictable, repeatable, load-independent gain/phase characteristics in the critical RF unity-gain crossing region?
Do you perform detailed stability tests? I doubt...
How else do you achieve predictable, repeatable, load-independent gain/phase characteristics in the critical RF unity-gain crossing region?
Do you perform detailed stability tests? I doubt...
For commercial designs, I suspect a motiver is the designer is trying to be extra cautious with regard to what might end up attached to his unit. Both zobel and an output coil may be intended a a sort of "firewall" between the amplifier and infinite possible permutations of cables and loudspeaker designs.
A potential benefit of DIY is that you know, presumably, exactly what will be connected to your amplifier and can use either, both or neither a zobel or output coil as appropriate.
A potential benefit of DIY is that you know, presumably, exactly what will be connected to your amplifier and can use either, both or neither a zobel or output coil as appropriate.
Hi, EVA,
Many of gainclone builder intentionally skip RC (the chip's datasheet have them), they say without RC the gainclone sounds better.
Many of gainclone builder intentionally skip RC (the chip's datasheet have them), they say without RC the gainclone sounds better.
sam9 said:
A potential benefit of professional designer is presence of deep knowledge and measurement tools, so all infinite possible permutations of capacitive, resistive, and inductive loads may be tested.
Why plain standard production always have lot of odd things?
Because it is safer to copy datasheets and fashion designs as is in terms of job security...
Nice to finally see an industry schematic with this. My take is (I've been fiddling with that RLC "base stoppers", too), the RC, when choosen optimally, gives an almost purely resistive output impedance of the driver EF (when 2*pi*f_beta*R*C = 1, f_beta being the -3dB corner of beta vs. freq), and the inductor shunts this RC for audio freqs to have low drive impedance (low distortion) there. So they could get rid of base stoppers for the main outputs.megajocke said:
The RLC values from the MA600 which I looked at now nicely fit these assumptions, the LR corner freq is 25kHz and the f_beta would be at about 500kHz which seems to be quite reasonable.
- Klaus
Do you mean LC in base of that emitter follower that can lead to oscillations due to base to emitter capacitance plus capacitance from emitter to ground? Otherwise an emitter follower itself can't oscillate, additional stages and feedback are needed.Bob Cordell said:
Do you mean termination of 300 Ohm line for frequencies above 3 MHz? I assume an ordinary speaker cable of 20 meter (approximately 60 feet) length.
Bob Cordell said:
I can understand your reasoning but from a practical point of view this assumes that the characteristic impedance of the speaker cable is constant and known in advance. It also assumes that the speaker cable is consistent throughout it's length. I doubt if speaker cable manufacturers have any quality control measures regarding the characteristic impedance so maybe this extra series RC network at the speaker terminals might be a bit of a mismatch.
However this begs the question should we be designing speaker cables with a certain characteristic impedance and terminating them at the speaker end with a series R-C circuit ??
Has anyone tried to used some heavy duty 50 ohm coax cable with their speakers ?? Results anyone ??
snoopy said:
With 50 Ohm speakers on frequencies of several MHz? I did! 😀
Also, I had a problem with 300 kilometers of 1200 Ohm line (air, on poles). It was a railway special service line for high impedance phones with tonal coded calls. Some phone did not answer, and I've found that on that station they had a notch on 2700 Hz, exactly one frequency of the sequence assigned to the station. Investigation revealed that there was a 150 Ohm underground cable in that line, but on a different railway station.
But there were 300 kilometers of 1200 Ohm line, enough for reflections in audio band!
lumanauw said:
And even gainclone kits are sold without Zobels. I have it from a source I consider reliable that quite a number of those have ended up in repair shops.
Most of transistor amps sound better with a resistor between output and a speaker. Who said that an electrical damping is more significant than linearity of output resistance?
Oscillation results in compression and "smooth" distortion of the output waveform. In most cases, oscillation only arises for certain ranges of output voltage and current in an asymmetrical way. This is similar to tube distortion and some people actually enjoys it.
Checking that an amplifier is unconditionally stable is not a trivial task. The RLC base stoppers are really the kind of design that I like. In fact, there are a lot of things happening well above 20Khz in audio amplifiers, particularly in class AB circuits, but there are few designers taking care about that.
I think oscillation is frequent among DIY and audiophile-style amplifiers. Circuits put together "by listening" often exhibit terrible waveform performance.
People builds gainclones without RL or even RC networks and then they complain about a too high noise floor or about heatsinks reaching 50ºC without any signal...
Checking that an amplifier is unconditionally stable is not a trivial task. The RLC base stoppers are really the kind of design that I like. In fact, there are a lot of things happening well above 20Khz in audio amplifiers, particularly in class AB circuits, but there are few designers taking care about that.
I think oscillation is frequent among DIY and audiophile-style amplifiers. Circuits put together "by listening" often exhibit terrible waveform performance.
People builds gainclones without RL or even RC networks and then they complain about a too high noise floor or about heatsinks reaching 50ºC without any signal...
Wavebourn said:
Amplifiers don't sound.
Most SPEAKERS exhibit great deviations in their frequency response, like -6dB
, depending on the impedance from which they are driven.In some cases, a high output impedance results in attenuation of frequencies that the subject does not like or in an effective boost of frequencies that the subject does like.
The subject gets more complex when drive impedance is not uniform across the audio band, or even worse, when output impedance is dependent on signal level (both things happen in almost every piece of tube gear).
A high drive impedance to a 2-way speaker often results in effective boost applied to three frequency ranges: The resonance of the bass driver, the usual 2Khz crossover region and the high trebble range above 10Khz (where the tweeter becomes inductive). Actually, the rest of the frequencies are the ones becoming attenuated but the result is the same.
sam9 said:
This is a very good point. Indeed, in the active loudspeakers I built I did not use inductors because the loudspeaker drivers were less than 18 inches from the amplifiers that were built into the cabinet. However, I DID use an RC Zobel because that is prudent in any case, since the impedance seen looking into the driver can go very high at high frequencies due to its inductive nature.
Cheers,
Bob
Yes, see attached circuit detail from the MA600/1200. Note that the two RC's on the "Low Side" on the right are output stage snubbers, as this amp is a bridge design where the right output is grounded and the rails and center tap are floating, shifted with the output voltage. Also we see a combined output L//R with an output snubber/zobel, for the high side, with the 2.7R as the shared element.Christer said:
As for the chip amps, I've seen output RC's there which are quite ineffective because of poor layout/wiring. With quasi comp outputs, the best place to put the RC is directly from the output pin to the neg supply pin.
- Klaus
Attachments
Wavebourn said:
Emitter followers configured in output stages like Darlington or Triple can in some situations get into oscillation in a no-load condition even if there is no discrete LC in the base. I've even seen a Triple get unstable in a SPICE simulation when unloaded where there were no real world parasitic inductances involved. For this reason, it is prudent that an R-C Zobel shunt network to ground is there to make sure that the EF output stage sees at least some resistive load at high frequencies.
When I refer to terminating the speaker line at high frequencies I am not referring to terminating it in its characteristic impedance. I'm mainly saying that it would like to be terminated in a lossy resistance at high frequencies, something typically in the 2-20 ohms region. This, of course, does not terminate the typical speaker cable in its characteristic impedance, which often lies between 50 and 300 ohms.
Cheers,
Bob
snoopy said:
Hi snoopy,
As I mentioned above, it was not my intention to suggest that the speaker cable be terminated in its characteristic impedance. I apologize for any confusion I created. A lossy resistive termination at high frequencies on at least one end is what is desirable, even if that termination resistance is quite a bit lower than the characteristic impedance of the line. A pure open or a pure short at either end of a transmission line is where there is sometimes the potential for reflection and resonance troubles. A lossy termination reduces reflections and damps potential resonances.
Cheers,
Bob
Wavebourn said:
Hi wavebourn,
This is a very astute observation, although I am not sure why it is the case. It may be just a subjective preference for the frequency response deviations caused by the series resistance working against the highly variable impedance of the loudspeaker.
Looking at it another way, I think that the lower damping factor of tube amplifiers is partly the reason for the sound that they have, which some prefer. If I want a solid state amplifier to sound more like a tube amplifier, I put some series resistance in its output to drop the damping factor down to about 20.
Cheers,
Bob