Bob Cordell said:
I should also be cautious.
Barring nonlinear disturbances, and considering only linear L-R effects, it is true that for example 2uHy/8 ohms are mostrly inocuous within the audio band. But the rub is in the "mostly". A back of the envelope check yields for example differences on the order of -60dB for harmonics of a 1 KHz square wave **within the audio range**.
For lower R, perturbation is higher still.
Whether this is significant taking into account we customarliy seek other distortion products to be 80-100 or less dB down ....
Rodolfo
[xpost]
I remembered that Doug Self at one time also researched this coil thing, and found yesterday the paragraph in his book 'Audio Power Amplifier Design Handbook' . For those who don't have it, he addresses three issues:
1 - Mutual coupling between coils. Short story: if you keep them 3-4 inches apart, and don't make a point of aligning them exactly, crosstalk will probably be below -100dB;
2 - Depending on the wire length and thickness of the coil material, even with thick wires a coil's resistance could be enough to cut the amp damping factor in half and cause appreciable power loss in loads less than 4 ohms.
I wonder whether this could also cause audible differences between coiled and coilless amps?
3 - Influence on ringing etc. He notes that ringing and overshoot with a coil is observed at the coil-speaker node, NOT at the actual amp output (before the coil), so in his opinion such ringing and overshoot says nothing about amp stability but just illustrates what happens when you excite a reactive network.
He notes that the most important factor determining the ringing and overshoot is the rise-time of the input test signal. I parafrase: the transient response measured in this way depends critically on the input signal rise time which can be manipulated to give the results wanted.
From the graphs he provides, it looks that with an input signal rise time down to 20uS there is almost no overshoot and ringing. I wonder how that translates into audio bandwidth. 20kHz has a period of 50uS so with a very, very rough estimate one could say that a 20kHz wave has a 25uS rise time, but it is faster at some parts of the wave because the rise time at zero crossing is faster than at the amplitude extremes. Nevertheless, testing with fast signals may make it look bad but doesn't seem to be relevant for audio.
Bottom line (for me):
- It is clear that the test signal has an major impact on the results and that audio-band limited test signals are appropriate;
- Simple resistive effects of the coil can cause audible differences due to lower damping and power loss in freq areas where the load dips to a low value.
Jan Didden
I remembered that Doug Self at one time also researched this coil thing, and found yesterday the paragraph in his book 'Audio Power Amplifier Design Handbook' . For those who don't have it, he addresses three issues:
1 - Mutual coupling between coils. Short story: if you keep them 3-4 inches apart, and don't make a point of aligning them exactly, crosstalk will probably be below -100dB;
2 - Depending on the wire length and thickness of the coil material, even with thick wires a coil's resistance could be enough to cut the amp damping factor in half and cause appreciable power loss in loads less than 4 ohms.
I wonder whether this could also cause audible differences between coiled and coilless amps?
3 - Influence on ringing etc. He notes that ringing and overshoot with a coil is observed at the coil-speaker node, NOT at the actual amp output (before the coil), so in his opinion such ringing and overshoot says nothing about amp stability but just illustrates what happens when you excite a reactive network.
He notes that the most important factor determining the ringing and overshoot is the rise-time of the input test signal. I parafrase: the transient response measured in this way depends critically on the input signal rise time which can be manipulated to give the results wanted.
From the graphs he provides, it looks that with an input signal rise time down to 20uS there is almost no overshoot and ringing. I wonder how that translates into audio bandwidth. 20kHz has a period of 50uS so with a very, very rough estimate one could say that a 20kHz wave has a 25uS rise time, but it is faster at some parts of the wave because the rise time at zero crossing is faster than at the amplitude extremes. Nevertheless, testing with fast signals may make it look bad but doesn't seem to be relevant for audio.
Bottom line (for me):
- It is clear that the test signal has an major impact on the results and that audio-band limited test signals are appropriate;
- Simple resistive effects of the coil can cause audible differences due to lower damping and power loss in freq areas where the load dips to a low value.
Jan Didden
[xpost]
Apologies for quoting myself...
The last point above suggests an interesting test:
One amp has a coil of, say 5uH. Another amp has a dummy coil which has the same resistance as the first one but not the inductance. Then A/B the two amps.
The dummy coill is contructed as follows: take the same wire and same length as the coil and fold it through the middle. You now have a twin wire connected at one point. Now wind it into a coil starting with the connected end. This will give you a bi-filar wound coil with close to zero inductance, or at any rate orders of magnitude less than the first coil, but the same resistance. And same heating effects etc.
Jan Didden
janneman said:
Apologies for quoting myself...
The last point above suggests an interesting test:
One amp has a coil of, say 5uH. Another amp has a dummy coil which has the same resistance as the first one but not the inductance. Then A/B the two amps.
The dummy coill is contructed as follows: take the same wire and same length as the coil and fold it through the middle. You now have a twin wire connected at one point. Now wind it into a coil starting with the connected end. This will give you a bi-filar wound coil with close to zero inductance, or at any rate orders of magnitude less than the first coil, but the same resistance. And same heating effects etc.
Jan Didden
The magnitude of the maximum rate of change of a voltage sine wave seems to be:
slew rate max of sine (in volts per microsecond) =
[(2 x Pi) x (freq in Hz) x (amplitude in volts)] / 1,000,000
So, for example, for a 10V 0-to-Peak 20 kHz sine, the maximum slew rate is about 1.257 V/us. For 20V 0-P it would be about 2.513 V/us.
So, for the rise and fall times of squarewaves used for amplifier testing or simulation, for example, IF we assume that we don't need to have slew rates at an amplifier's output that are outside of some maxfreq range, then the maximum slew rate for the input would be 2 x Pi x peak output voltage x maxfreq / gain / 1000000, in V/usec.
That would mean that for a gain of 20, and a 40v p-p squarewave output, with maxfreq = 22kHz, the input would be a 2v p-p squarewave that would only need rise and fall slew rates of <= 0.138 V/us, i.e. about 14.47 usec risetime and falltime, minimum, giving a maximum slew rate of about 2.764 V/us for a -20v to +20v squarewave transition at the output.
Does that sound right?
If so, I guess I might have wasted some LTspice time by shooting for more like 7.5 V/us output edge-times, with no overshoot or ringing into <= 2.2uF || 8 Ohms. On the bright side, those designs should now look _really_ great with slower-edged squarewaves, and all of my designs from now on should be much easier to do. ;-)
Does anyone have any idea about how much slewrate "headroom" might be reasonable, to design-in?
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
-
slew rate max of sine (in volts per microsecond) =
[(2 x Pi) x (freq in Hz) x (amplitude in volts)] / 1,000,000
So, for example, for a 10V 0-to-Peak 20 kHz sine, the maximum slew rate is about 1.257 V/us. For 20V 0-P it would be about 2.513 V/us.
So, for the rise and fall times of squarewaves used for amplifier testing or simulation, for example, IF we assume that we don't need to have slew rates at an amplifier's output that are outside of some maxfreq range, then the maximum slew rate for the input would be 2 x Pi x peak output voltage x maxfreq / gain / 1000000, in V/usec.
That would mean that for a gain of 20, and a 40v p-p squarewave output, with maxfreq = 22kHz, the input would be a 2v p-p squarewave that would only need rise and fall slew rates of <= 0.138 V/us, i.e. about 14.47 usec risetime and falltime, minimum, giving a maximum slew rate of about 2.764 V/us for a -20v to +20v squarewave transition at the output.
Does that sound right?
If so, I guess I might have wasted some LTspice time by shooting for more like 7.5 V/us output edge-times, with no overshoot or ringing into <= 2.2uF || 8 Ohms. On the bright side, those designs should now look _really_ great with slower-edged squarewaves, and all of my designs from now on should be much easier to do. ;-)
Does anyone have any idea about how much slewrate "headroom" might be reasonable, to design-in?
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
-
gootee said:
Hi Tom,
I think you are largely right, but perhaps with a couple of caveats.
I would not try to artificially limit the square wave edge slope by turning a knob on the generator. Depending on what I was trying to test, I might bandlimit the squarewave to, say, 100 kHz or so.
In any case, the square wave should be small-signal and not cause slew rate limiting. If you are wanting to analyze stability, you probably don't want to significantly bandlimit the squarewave, as you want to retain the harmonics in it so as to excite regions in the spectrum where instability may occur.
As far as amplifier design for adequate slew rate headroom, I usually like to see an amplifier have at least 50 V/us of slew rate, but realize that this is pretty arbitrary.
As you pointed out, the edge rate of a 20 kHz sinusoid is about 0.125 V/us per volt peak. A 100W amplifier will produce 40V peak, implying a slew rate of 5.0 v/us. In this case, the 50 V/us recommendation corresponds to a slew rate headroom factor of 10:1 against the maximum 20 kHz sinusoidal edge rate. This does not seem unreasonable.
Cheers,
Bob
Hi Bob,
Thanks, very much, for the advice.
The 50 V/us is a little bit surprising, to me. Maybe THAT'S why some people prefer discrete amplifier designs versus chipamps.
The LM1875 datasheet, for example, shows a typical minimum slew rate of 7 V/us, while the LM3875 shows 11 V/us and the LM3886 shows 19 V/us. The acclaimed LM4562 audio opamp's datasheet shows 20 V/us. They appear to not actually say what the "maximum" slew rates might be.
When I was playing around with opamp-and-chipamp-based error-correction amplifier designs, a few months ago, about the only decent power chipamp spice model I could find was the OPA541E from ti.com, which was limited to 7.5 V/us. I basically couldn't do anything about that. But in the feedback loops, I found it was usually best to use the fastest opamps for which I had spice models, e.g. the LT1363 @ 1000 V/us.
Maybe I should revisit those EC amp circuits and substitute a classic "opamp with discrete power booster" circuit in place of the OPA541E, so I can see what's possible when much higher output slew rates are available.
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
-
Thanks, very much, for the advice.
The 50 V/us is a little bit surprising, to me. Maybe THAT'S why some people prefer discrete amplifier designs versus chipamps.
The LM1875 datasheet, for example, shows a typical minimum slew rate of 7 V/us, while the LM3875 shows 11 V/us and the LM3886 shows 19 V/us. The acclaimed LM4562 audio opamp's datasheet shows 20 V/us. They appear to not actually say what the "maximum" slew rates might be.
When I was playing around with opamp-and-chipamp-based error-correction amplifier designs, a few months ago, about the only decent power chipamp spice model I could find was the OPA541E from ti.com, which was limited to 7.5 V/us. I basically couldn't do anything about that. But in the feedback loops, I found it was usually best to use the fastest opamps for which I had spice models, e.g. the LT1363 @ 1000 V/us.
Maybe I should revisit those EC amp circuits and substitute a classic "opamp with discrete power booster" circuit in place of the OPA541E, so I can see what's possible when much higher output slew rates are available.
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
-
And then, coincidentally, I find this!
The 50V/us slew rate is what caught my eye, because of Bob's comments. But it looks like a pretty nice chip.
Interestingly, it has pins for an external Vbe multiplier. There's also an output pin for a clipping indicator, such as an LED.
Looks like its initial Web introduction was quite recent: 5/24/07.
http://www.national.com/pf/LM/LME49810.html
"LME49810 - 200V Audio Power Amplifier Driver with Baker Clamp"
■ Wide operating voltage range ±20V to ±100V
■ Slew Rate 50V/μs (typ)
■ Output Drive Current 60mA (typ)
■ PSRR (f = DC) 110dB (typ)
■ THD+N (f = 1kHz) 0.0007 (typ)
"The LME49810 is a high fidelity audio power amplifier driver designed for demanding consumer and pro-audio applications. Amplifier output power may be scaled by changing the supply voltage and number of power transistors. The LME49810’s minimum output current is 50mA. When using a discrete output stage the LME49810 is capable of delivering in excess of 300 watts into a single-ended 8Ω load.
Unique to the LME49810 is an internal Baker Clamp. This clamp insures that the amplifier output does not saturate when over driven. The resultant “soft clipping” of high level audio signals suppresses undesirable audio artifacts generated when conventional solid state amplifiers are driven hard into clipping.
The LME49810 includes thermal shutdown circuitry that activates when the die temperature exceeds 150°C. The LME49810’s mute function, when activated, mutes the input drive signal and forces the amplifier output to a quiescent state."
My samples are on their way.
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
-
The 50V/us slew rate is what caught my eye, because of Bob's comments. But it looks like a pretty nice chip.
Interestingly, it has pins for an external Vbe multiplier. There's also an output pin for a clipping indicator, such as an LED.
Looks like its initial Web introduction was quite recent: 5/24/07.
http://www.national.com/pf/LM/LME49810.html
"LME49810 - 200V Audio Power Amplifier Driver with Baker Clamp"
■ Wide operating voltage range ±20V to ±100V
■ Slew Rate 50V/μs (typ)
■ Output Drive Current 60mA (typ)
■ PSRR (f = DC) 110dB (typ)
■ THD+N (f = 1kHz) 0.0007 (typ)
"The LME49810 is a high fidelity audio power amplifier driver designed for demanding consumer and pro-audio applications. Amplifier output power may be scaled by changing the supply voltage and number of power transistors. The LME49810’s minimum output current is 50mA. When using a discrete output stage the LME49810 is capable of delivering in excess of 300 watts into a single-ended 8Ω load.
Unique to the LME49810 is an internal Baker Clamp. This clamp insures that the amplifier output does not saturate when over driven. The resultant “soft clipping” of high level audio signals suppresses undesirable audio artifacts generated when conventional solid state amplifiers are driven hard into clipping.
The LME49810 includes thermal shutdown circuitry that activates when the die temperature exceeds 150°C. The LME49810’s mute function, when activated, mutes the input drive signal and forces the amplifier output to a quiescent state."
My samples are on their way.
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
-
gootee said:
Hi Tom,
This looks like a nice part. I'll take a look at the data sheet. I noticed there is no pricing on the net yet.
I like Baker clamps. However, I would not go so far as to say that they provide soft clipping, as there is still NFB around it that sharpens it up. However, the Baker clamp keeps transistors out of saturation, and tends to prevent "sticking", which is a further exacerbation of sharp clipping.
Cheers,
Bob
Bob Cordell said:
While I was posting that, I remembered following a discussion, here, about your KleverKlipper (IIRC the name) circuit, probably six to nine months ago. Sometime soon, I'll have to check your website, again, to see if there is more-detailed information available, about that, and also track down those messages.
At this point I have almost no recollection of how that was done, or even exactly where in a discrete amplifier circuit it might best be implemented. But now I'm somewhat curious about whether or not a chip like this one would tend to prevent attempting implementation of better handling of clipping, by making the relevant circuits inaccessible.
- Tom
gootee said:
Hi Tom,
The Klever Klipper is a passive-adaptive soft clip circuit that goes in front of the power amplifier-proper, outside the feedback loop. Because of this, I see no reason that such a circuit could not be used with a chip like this one. Indeed, I use a Klever Klipper circuit in front of my Super Gain Clone amplifier based on the LM3886.
In its simplest implementation, it just uses a pair of diodes that are back-biased by an amount proportional to the main rail voltage available to the output stage. In this way, its clipping level adapts to the available power supply headroom, preserving dynamic headroom. It never lets the power amplifier itself clip. The Klever Klipper is, of course, arranged to be defeatable.
The theory is that, if you must clip, do it in a controlled, civilized, soft, memoryless, passive way.
Some variations of the Klever Klipper give an idication when they are clipping the signal. I'll try to get some more information on it up on the site soon.
I believe that many others have implemented similar approaches out there.
Cheers,
Bob
gootee said:
Yes, Tom.
I have written a couple of notes on this, in several topics:
chips vs. dsicrete op-amps
....................................
- opamps, safe and good if trimmed alright for task
- not much chances ( read: pins) of anyone manipulating and upset from outside
- transistor discrete circuit, unlimited possibilities, if you know how
Of course this above
is elementary and obvious to anyone
that has given these things some thought.
OPA637 AD825 AD797 and LM6031 may be good
for some quick & dirty, safe and instant setup.
... but oh, what boring, if you want to
explore & test some amplifier/transistors/trimming matters
Audio 3rd harmonics dist
regards
Lineup Audio Lab Inc.
...............................................................................................
we hope you all have a Nice Day, you are Rich & truly Happy
and that you do not never: Lack one major important thing.
... and finally, you never, please! call Lineup 'good':
There is but ONE that is good =
ALLAH, the great .... Spirit In the Sky
Follow up to single point error correction
The attached schematic illustrates an error corrected output stage utilizing a single point of error correction that is applied to both the positive and negative halves of the output stage. By use of floating + and - 15V supplies, it is possible to utilize a high speed opamp as the correction circuit and generate only the error signal. In this case the + and - 15V supplies float on the output of the amplifier, and the error voltage is generated by comparing the outputs of the driver and output stages. With a little arithmetic one can show that the voltage applied to the output stage's input is equal to the sum of the driver stage voltage plus a correction voltage exactly equal to the error introduced by the output stage.
Note the tolerance on the four 100 ohm resistors. They need not achieve an absolute accuracy that great, but they must be matched to this tolerance. Simulations showed that 1% resistors caused a distortion increase of approx 15 dB. This is one of the few places where I have found this degree of matching is required in audio circuitry.
Some minor modifications are required for the driver and output stages. It is necessary to configure the driver stage such that it generates a single output signal, rather than a pair of signals offset by a few volts as is usually the case. This is easily achieved by adding a pair of source followers as would be the case for a real output stage. The only difference is the amount of current this stage must furnish is modest (~25 ma). The source follower is also necessary to present a low impedance driving source to R1.
The output stage needs to re-establish a DC offset between the drive to the upper and lower halves, and this is easily done with a pair of VBE multipliers. Constant current sources connected to the pos and neg rails guarantee the bias circuits active independent of the voltage swing. (clipping will turn them off). As I posted in an earlier thread, this circuit achieves (in simulations at least) distortion suppression in the range of 30 - 50 dB, depending on the frequency.
I'll post some waveforms in the next few days.
The attached schematic illustrates an error corrected output stage utilizing a single point of error correction that is applied to both the positive and negative halves of the output stage. By use of floating + and - 15V supplies, it is possible to utilize a high speed opamp as the correction circuit and generate only the error signal. In this case the + and - 15V supplies float on the output of the amplifier, and the error voltage is generated by comparing the outputs of the driver and output stages. With a little arithmetic one can show that the voltage applied to the output stage's input is equal to the sum of the driver stage voltage plus a correction voltage exactly equal to the error introduced by the output stage.
Note the tolerance on the four 100 ohm resistors. They need not achieve an absolute accuracy that great, but they must be matched to this tolerance. Simulations showed that 1% resistors caused a distortion increase of approx 15 dB. This is one of the few places where I have found this degree of matching is required in audio circuitry.
Some minor modifications are required for the driver and output stages. It is necessary to configure the driver stage such that it generates a single output signal, rather than a pair of signals offset by a few volts as is usually the case. This is easily achieved by adding a pair of source followers as would be the case for a real output stage. The only difference is the amount of current this stage must furnish is modest (~25 ma). The source follower is also necessary to present a low impedance driving source to R1.
The output stage needs to re-establish a DC offset between the drive to the upper and lower halves, and this is easily done with a pair of VBE multipliers. Constant current sources connected to the pos and neg rails guarantee the bias circuits active independent of the voltage swing. (clipping will turn them off). As I posted in an earlier thread, this circuit achieves (in simulations at least) distortion suppression in the range of 30 - 50 dB, depending on the frequency.
I'll post some waveforms in the next few days.
Attachments
Re: Follow up to single point error correction
This looks like a nice circuit. The only question that I have is the pros and cons of doing something like this with an op amp as opposed to the discrete EC implementation. Doing it with an op amp dosn't really seem to reduce complexity by the time you put in things like the floating power supplies. On the other hand, it would certainly make sense if it improved performance compared to the discrete version.
The other key thing to bear in mind in comparisons is that the amount of high-frequency error correction is what is most important, so speed is king in the EC circuit.
Any thoughts on these pros and cons?
Thanks,
Bob
analog_guy said:
This looks like a nice circuit. The only question that I have is the pros and cons of doing something like this with an op amp as opposed to the discrete EC implementation. Doing it with an op amp dosn't really seem to reduce complexity by the time you put in things like the floating power supplies. On the other hand, it would certainly make sense if it improved performance compared to the discrete version.
The other key thing to bear in mind in comparisons is that the amount of high-frequency error correction is what is most important, so speed is king in the EC circuit.
Any thoughts on these pros and cons?
Thanks,
Bob
Re: Re: Follow up to single point error correction
Bob,
The choice of op amp is indeed critical. It needs to be fast and capable of driving ~ 100 ma. The EL2045 meets those requirements and is one of the few opamps supporting a +/- 15V rail voltage that will. Most high speed opamps are limited to +/- 5V or something in that range. Configured for 2x non-inveting gain, the EL2045's power bandwidth exceeds 100 MHz. In fact, it was necessary to roll off its response with an inductor as part of the feedback return path. Without the inductor the EC circuit briefly oscillates if presented with a sharp step response.
The floating power supplies are not a big deal, since small (1-2 watt) isolated DC-DC converters are available at reasonable cost. Also, if I remember correctly, your 1984 design example required separate supplies for the driver and output stages, so the cost should be comparable.
BTW, when designing the circuit, I purposely put performance ahead of simplicity. I have not tried to simulate your original circuit to see how it compares to an opamp-based approach. Have you ever done so? It would be interesting to compare the performance of the two different approaches. If you want, I can provide you with the signal levels and frequencies and load characteristics that I used, so we'll have an apples-to-apples comparison.
Regards,
Jeff
Bob Cordell said:
Bob,
The choice of op amp is indeed critical. It needs to be fast and capable of driving ~ 100 ma. The EL2045 meets those requirements and is one of the few opamps supporting a +/- 15V rail voltage that will. Most high speed opamps are limited to +/- 5V or something in that range. Configured for 2x non-inveting gain, the EL2045's power bandwidth exceeds 100 MHz. In fact, it was necessary to roll off its response with an inductor as part of the feedback return path. Without the inductor the EC circuit briefly oscillates if presented with a sharp step response.
The floating power supplies are not a big deal, since small (1-2 watt) isolated DC-DC converters are available at reasonable cost. Also, if I remember correctly, your 1984 design example required separate supplies for the driver and output stages, so the cost should be comparable.
BTW, when designing the circuit, I purposely put performance ahead of simplicity. I have not tried to simulate your original circuit to see how it compares to an opamp-based approach. Have you ever done so? It would be interesting to compare the performance of the two different approaches. If you want, I can provide you with the signal levels and frequencies and load characteristics that I used, so we'll have an apples-to-apples comparison.
Regards,
Jeff
Re: Re: Follow up to single point error correction
Hi, Bob.
Yes, "speed is king" in the EC circuit(s). I found that out when I tried to implement a "delay", to try to get a better comparison (time-alignment-wise) for generation of the error signal. I actually was able to see a pretty-significant improvement by delaying the input to be better time-aligned with the output. But it was a bit "painful", circuit-wise. One of my better EC amps of that genre used a "six-pole Sallen-Key 0 dB low-pass" filter with "linear-phase .05-degree equiripple error" response-type, where I used a higher-frequency and a higher-order filter (relative to ?... OK...) to be able to get more more delay with much less LP filtering effect. And the opamps I used for that had to be very, very fast, indeed; the faster the better!
But... that made me think... If most or all of that could be done (gasp!) "digitally", somehow, it might be much easier and better (maybe very nearly perfect). However, the eventual output might need to be significantly delayed, time-wise, versus the actual souce. BUT, no one (else) would know, if it was only a millisecond or two behind.
There might be some basic violation of the laws of physics involved. But, if so, I'm guessing that that could be surmounted with, say, two sets of enough memory and way more (digital) speed than would normally be needed, for "practical" situations, at least.
I'm sorry if that's too uncomfortably-far off-topic. I'm an analog guy, myself, but usually have to (or try to) keep digital in mind, too.
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
-
Bob Cordell said:
Hi, Bob.
Yes, "speed is king" in the EC circuit(s). I found that out when I tried to implement a "delay", to try to get a better comparison (time-alignment-wise) for generation of the error signal. I actually was able to see a pretty-significant improvement by delaying the input to be better time-aligned with the output. But it was a bit "painful", circuit-wise. One of my better EC amps of that genre used a "six-pole Sallen-Key 0 dB low-pass" filter with "linear-phase .05-degree equiripple error" response-type, where I used a higher-frequency and a higher-order filter (relative to ?... OK...) to be able to get more more delay with much less LP filtering effect. And the opamps I used for that had to be very, very fast, indeed; the faster the better!
But... that made me think... If most or all of that could be done (gasp!) "digitally", somehow, it might be much easier and better (maybe very nearly perfect). However, the eventual output might need to be significantly delayed, time-wise, versus the actual souce. BUT, no one (else) would know, if it was only a millisecond or two behind.
There might be some basic violation of the laws of physics involved. But, if so, I'm guessing that that could be surmounted with, say, two sets of enough memory and way more (digital) speed than would normally be needed, for "practical" situations, at least.
I'm sorry if that's too uncomfortably-far off-topic. I'm an analog guy, myself, but usually have to (or try to) keep digital in mind, too.
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
-
Re: Re: Re: Follow up to single point error correction
I agree with your points. That op amp looks like it should perform well in this application. I guess your use of the floating supplies eliminates the need for boosted supplies for the front-end and the VAS? If so, I guess you're right about the complexity being similar.
I have never simulated my amplifier for purposes of distortion analysis, so we'd have to do any comparison based on measured results.
Cheers,
Bob
analog_guy said:
I agree with your points. That op amp looks like it should perform well in this application. I guess your use of the floating supplies eliminates the need for boosted supplies for the front-end and the VAS? If so, I guess you're right about the complexity being similar.
I have never simulated my amplifier for purposes of distortion analysis, so we'd have to do any comparison based on measured results.
Cheers,
Bob
EC Amp Boards are Back
The first boards for my EC amp design have come back. I have soldered up the first one, and with one minor adjustment to the compensation network, they appear to work fine. Since this design does not use GFB around the output stage, the voltage gain stage can be tested in isolation.
DC idle current and voltage values, bandwidth, and rise/fall time are all within 20% of simulated values. Right now, I do not have access to a distortion analyzer, so that measurement will be deferred until the output stages are completed. Then I'll either buy or rent an Audio Precision analyzer.
One interesting finding: The design uses LM317/LM337 regulators in a floating mode. Since the supply voltage is +/- 70V, it is important to ensure that the 317/337 max input/output differential never exceeds 40V. Otherwise they will fail. The recommended protection scheme with diodes does not protect against this failure mode. It is necessary to replace the diodes with transzorb zeners with a ~30V knee voltage. During power-up the filter cap on the 317/337 output is at 0V. So if the supply voltage dv/dt exceeds the response time of the regulator, the in/out voltage differential can be exceeded. The zener's breakover voltage guarantees this cannot happen. The circuitry serviced by the 317/337 is capaple of tolerating + or - 70V, so there is no problem there.
I'll post a photo when I get the opportunity.
The first boards for my EC amp design have come back. I have soldered up the first one, and with one minor adjustment to the compensation network, they appear to work fine. Since this design does not use GFB around the output stage, the voltage gain stage can be tested in isolation.
DC idle current and voltage values, bandwidth, and rise/fall time are all within 20% of simulated values. Right now, I do not have access to a distortion analyzer, so that measurement will be deferred until the output stages are completed. Then I'll either buy or rent an Audio Precision analyzer.
One interesting finding: The design uses LM317/LM337 regulators in a floating mode. Since the supply voltage is +/- 70V, it is important to ensure that the 317/337 max input/output differential never exceeds 40V. Otherwise they will fail. The recommended protection scheme with diodes does not protect against this failure mode. It is necessary to replace the diodes with transzorb zeners with a ~30V knee voltage. During power-up the filter cap on the 317/337 output is at 0V. So if the supply voltage dv/dt exceeds the response time of the regulator, the in/out voltage differential can be exceeded. The zener's breakover voltage guarantees this cannot happen. The circuitry serviced by the 317/337 is capaple of tolerating + or - 70V, so there is no problem there.
I'll post a photo when I get the opportunity.
Re: EC Amp Boards are Back
Sounds great! Keep us posted on progress and details.
That's a good point about the LM317/337 use in a floating arrangement for the high supplies. I like the idea of the transorbs. I had a similar situation on an amp I did many years ago, using those same regulators to regulate the boosted rails. If I recall correctly, I had a small transformer with isolated secondaries to supply the low-current boost, its rectifier sitting on top of the main rails. I believe I used a diode going from the main rails to the boosted rails in that case. That diode would charge up the boost filter capacitor fast enough that the LM317/337 would never see a drop greater than the boost delta. I don't remember exactly what I did with the voltage-setting part of the regulator.
Cheers,
Bob
analog_guy said:
Sounds great! Keep us posted on progress and details.
That's a good point about the LM317/337 use in a floating arrangement for the high supplies. I like the idea of the transorbs. I had a similar situation on an amp I did many years ago, using those same regulators to regulate the boosted rails. If I recall correctly, I had a small transformer with isolated secondaries to supply the low-current boost, its rectifier sitting on top of the main rails. I believe I used a diode going from the main rails to the boosted rails in that case. That diode would charge up the boost filter capacitor fast enough that the LM317/337 would never see a drop greater than the boost delta. I don't remember exactly what I did with the voltage-setting part of the regulator.
Cheers,
Bob
What is this?
Is this an EC? The whole thing works as a follower (no voltage gain). Input from VAS comes to node (b), entering point (3) of the opamp. This opamp feeding the Vbe multiplier (bias scheme) which is loaded by R526. The VAS working lightly, because it only feeds the base of the non-inverting input of the opamp.
The opamp itself has feedback from the output node via R508, entering point (2) of the opamp (inverting input).
The supply of the opamp is bootstrapped to the input node, always follows the input (thus the output, because the whole thing is a follower), D506 and D508 are referenced to point (b) of the input.
Is this an EC? The whole thing works as a follower (no voltage gain). Input from VAS comes to node (b), entering point (3) of the opamp. This opamp feeding the Vbe multiplier (bias scheme) which is loaded by R526. The VAS working lightly, because it only feeds the base of the non-inverting input of the opamp.
The opamp itself has feedback from the output node via R508, entering point (2) of the opamp (inverting input).
The supply of the opamp is bootstrapped to the input node, always follows the input (thus the output, because the whole thing is a follower), D506 and D508 are referenced to point (b) of the input.
