I'm glad you do agree on this, Chris.
After following the discussions over lots of capacitor's brands and models, in which I get usually lost, I have been trying to get to some common ground to make real and valid comparisons between the caps.
My recent question on what was the bias voltage that was being mentioned was a first attempt at that. I still do not see how that bias can be set (or if can be set) to the ideal, lower distortion point.
Then comes the position of that cap on the power supply, particularly when using some regulator. This can be before the regulator, inside the regulator or at the output.
At the same time, we can not forget what the capacitor is doing at each position in the power supply. This dynamic part the capacitor has on each stage should be of utmost consideration.
As very few here have measuring instruments, we have the manufacturer's data and the subjective tests most can do and are already doing.
Carlos (not Carl)
Carlos,
dia bom - I worked in Belo Hoizonte for a while.
The LT 3042 is very low noise, but surface mount.
http://www.analog.com/media/en/tech...le/LTJournal-V25N1-00-df-LT3042-AmitPatel.pdf
The Jung super regulator can be built with through hole components which makes it good for DIY , i use it for many projects.
dia bom - I worked in Belo Hoizonte for a while.
The LT 3042 is very low noise, but surface mount.
http://www.analog.com/media/en/tech...le/LTJournal-V25N1-00-df-LT3042-AmitPatel.pdf
The Jung super regulator can be built with through hole components which makes it good for DIY , i use it for many projects.
Hi Carlos,
When looking at capacitors without the aid of test equipment, you really are at the mercy of the manufacturer. From talking to a few members from your country I understand that used test equipment isn't really available to you either, not to mention the prices of new equipment. You aren't in the position to challenge the manufacturer's data either. I was last in your position as a young boy and I do remember how frustrating it was.
So the best you can do is very careful and systematic testing of capacitors. This probably has to be done with a piece of audio equipment (a receiver or integrated amplifier) that is already very good (expensive) or you will not be able to control all the variables at play. Even that requirement cuts many people out of the running. I would suggest something like the Symasym. I found that various components really showed up in that amplifier because it is good enough to resolve differences in some parts. To get this amplifier up to snuff you need to match the differential pair carefully. I designed a jig for this that uses a DVM reading in mV. The DVM is only a null device, so you might be able to use a null meter (centre tune meter from a receiver) with less absolute accuracy, but you are really looking for a null (zero deflection). Once you have that built, you can begin looking at capacitors seriously.
Listening tests will need good notes and repeat sessions. This is something you already know. For me, I just look for parts with near zero dissipation factor, or a low dissipation factor if you are looking at electrolytic capacitors. Most simple test setups require an oscilloscope and oscillator. If you can obtain those items, you can replicate some tests and much more easily sort out the good from the bad. If you know someone with this equipment, you can team up with them. I hope this would be possible for you.
Okay, now that you have capacitors separated by dissipation factor, or waveform distortion if using an oscillator and 'scope, you can listen to the different types of capacitors and get some general ideas of capacitor behavior vs test results. Then you can quantify your results in a way that makes sense to you.
I have it easy these days Carlos. I just connect a capacitor to my LCR meter and read the dissipation. I can also read series impedance or even get phase angles. If you see an HP or Agilent (newer HP) 4263A or B LCR meter, grab it! It will test parts at different frequencies. I have found inductive metal film resistors by doing that!
I hope some of this helps you define and quantify capacitors more easily with some test indications that you are on the right path.
So, what did you want to discuss with regard to capacitors?
Best, Chris
When looking at capacitors without the aid of test equipment, you really are at the mercy of the manufacturer. From talking to a few members from your country I understand that used test equipment isn't really available to you either, not to mention the prices of new equipment. You aren't in the position to challenge the manufacturer's data either. I was last in your position as a young boy and I do remember how frustrating it was.
So the best you can do is very careful and systematic testing of capacitors. This probably has to be done with a piece of audio equipment (a receiver or integrated amplifier) that is already very good (expensive) or you will not be able to control all the variables at play. Even that requirement cuts many people out of the running. I would suggest something like the Symasym. I found that various components really showed up in that amplifier because it is good enough to resolve differences in some parts. To get this amplifier up to snuff you need to match the differential pair carefully. I designed a jig for this that uses a DVM reading in mV. The DVM is only a null device, so you might be able to use a null meter (centre tune meter from a receiver) with less absolute accuracy, but you are really looking for a null (zero deflection). Once you have that built, you can begin looking at capacitors seriously.
Listening tests will need good notes and repeat sessions. This is something you already know. For me, I just look for parts with near zero dissipation factor, or a low dissipation factor if you are looking at electrolytic capacitors. Most simple test setups require an oscilloscope and oscillator. If you can obtain those items, you can replicate some tests and much more easily sort out the good from the bad. If you know someone with this equipment, you can team up with them. I hope this would be possible for you.
Okay, now that you have capacitors separated by dissipation factor, or waveform distortion if using an oscillator and 'scope, you can listen to the different types of capacitors and get some general ideas of capacitor behavior vs test results. Then you can quantify your results in a way that makes sense to you.
I have it easy these days Carlos. I just connect a capacitor to my LCR meter and read the dissipation. I can also read series impedance or even get phase angles. If you see an HP or Agilent (newer HP) 4263A or B LCR meter, grab it! It will test parts at different frequencies. I have found inductive metal film resistors by doing that!
I hope some of this helps you define and quantify capacitors more easily with some test indications that you are on the right path.
So, what did you want to discuss with regard to capacitors?
Best, Chris
Hi,
Belo Horizonte is a very nice city. My wife has relatives there. Found some regulator kits on eBay using the LT3042.
Low Noise LT3042 Linear Regulator Power Supply Board DC Converter Overvoltage | eBay
Yes, the Jung Super regulator can be done using TH components alright. The pcb is critical though, preferring dual sided boards, which is not too DIY friendly to realize.
IMHO people who are using capacitors here on regulated power supplies should specify which supply that is. It could make a major difference to know that.
Hi Chris,
I'm in the process of building a Luxman "clone" amplifier, which is discussed on as specific thread in DIYAudio. This is an amplifier I am very familiar with, and I only involve in building amps I have listened to or were listened by people I rely on.
The architecture is quite similar to the Symasim, but much more sophisticated. There's dual fet at the LTP input, which already takes care of the matching problem. A dual FET or dual transistor is essential on an LTP input. Matching is not easy.
The original Luxman amp I talk about was considered one of the best at the time it was released ('80s) and still holds its own very well with more modern amps. A friend of mine who builds high quality speakers in Argentina uses the Luxman as his reference amplification to test his speakers. That's why I know the amp so well.
An LCR meter looks like a very good way to measuring caps nowadays, and they are quite affordable. And DF is an excellent reference to start evaluating a cap. You can read its series impedance with it too? What can you do with phase angles in practical terms?
There are no electrolytic caps on the Luxman, except the bipolar on the NFB end. The original Luxman was pure DC, no cap or servo, but that is not feasible on a clone. So I will try it with a bipolar on the NFB and with a servo.
There's a third way you can try on a FET input amp, which is what Luxman called duo beta arrangement, that they used on the NFB on some models. It allowed using a 1uF or 2uF film cap, which is a very good thing.
But going back to the question of this thread, best electrolytic capacitors, the two really critical areas for caps in a preamplifier or power amplifier is where the capacitor is in the signal path.
That would be at the input and output, linking to other stages, and in the NFB capacitors.
On some cases, like on RIAA preamps, this NFB cap can be really large, which demands an electrolytic type.
This application can be extremely critical and affecting audio quality a lot more than power supply bypassing. Particularly if the preamp already has a high PSRR.
Carlos
Belo Horizonte is a very nice city. My wife has relatives there. Found some regulator kits on eBay using the LT3042.
Low Noise LT3042 Linear Regulator Power Supply Board DC Converter Overvoltage | eBay
Yes, the Jung Super regulator can be done using TH components alright. The pcb is critical though, preferring dual sided boards, which is not too DIY friendly to realize.
IMHO people who are using capacitors here on regulated power supplies should specify which supply that is. It could make a major difference to know that.
Hi Chris,
I'm in the process of building a Luxman "clone" amplifier, which is discussed on as specific thread in DIYAudio. This is an amplifier I am very familiar with, and I only involve in building amps I have listened to or were listened by people I rely on.
The architecture is quite similar to the Symasim, but much more sophisticated. There's dual fet at the LTP input, which already takes care of the matching problem. A dual FET or dual transistor is essential on an LTP input. Matching is not easy.
The original Luxman amp I talk about was considered one of the best at the time it was released ('80s) and still holds its own very well with more modern amps. A friend of mine who builds high quality speakers in Argentina uses the Luxman as his reference amplification to test his speakers. That's why I know the amp so well.
An LCR meter looks like a very good way to measuring caps nowadays, and they are quite affordable. And DF is an excellent reference to start evaluating a cap. You can read its series impedance with it too? What can you do with phase angles in practical terms?
There are no electrolytic caps on the Luxman, except the bipolar on the NFB end. The original Luxman was pure DC, no cap or servo, but that is not feasible on a clone. So I will try it with a bipolar on the NFB and with a servo.
There's a third way you can try on a FET input amp, which is what Luxman called duo beta arrangement, that they used on the NFB on some models. It allowed using a 1uF or 2uF film cap, which is a very good thing.
But going back to the question of this thread, best electrolytic capacitors, the two really critical areas for caps in a preamplifier or power amplifier is where the capacitor is in the signal path.
That would be at the input and output, linking to other stages, and in the NFB capacitors.
On some cases, like on RIAA preamps, this NFB cap can be really large, which demands an electrolytic type.
This application can be extremely critical and affecting audio quality a lot more than power supply bypassing. Particularly if the preamp already has a high PSRR.
Carlos
Hi Carlos,
I was doing warranty service on Luxman through that time. Which one are you building?
A servo can really affect an amplifier in a bad way if it is applied to the diff pair. A servo unbalances the diff pair to achieve low DC offset. You don't need to control the DC offset that tightly. You can apply DC offset compensation later in the signal chain after the voltage amplifier stage.
A really good LCR meter will tell you a lot about a component, accurately. The low cost models will only give you a rough idea on any parameter. Mine is ancient as far as test equipment goes. I'd love a new one, but it is accurate enough, repeatable and reliable.
I recently bought and restored a C-02, M-02 and T-03 combination. I'm just waiting for speakers now.
-Chris
I was doing warranty service on Luxman through that time. Which one are you building?
A servo can really affect an amplifier in a bad way if it is applied to the diff pair. A servo unbalances the diff pair to achieve low DC offset. You don't need to control the DC offset that tightly. You can apply DC offset compensation later in the signal chain after the voltage amplifier stage.
Mine, yes.
The phase angle just gives you a feel for how reactive the component is at that frequency. You could also measure the DC resistance, then the impedance at some frequency and figure it out that way. It's a way of expressing how much the reactive portion (imaginary) of the impedance compares to the real impedance (resistive).
A really good LCR meter will tell you a lot about a component, accurately. The low cost models will only give you a rough idea on any parameter. Mine is ancient as far as test equipment goes. I'd love a new one, but it is accurate enough, repeatable and reliable.
Yup, familiar with that.
I recently bought and restored a C-02, M-02 and T-03 combination. I'm just waiting for speakers now.
-Chris
Luxman 5M21.
The servo would be connected to the large feedback resistor, not to the differential pair.
Hi Carlos,
Correct me if I'm mistaken, but you are still feeding the DC correction signal into the diff pair aren't you? That would, if it be the case, unbalance your diff pair and you can't really allow that to happen. Unbalancing the diff pair directly affects it's ability to cancel audio signals and that will increase distortion.
What is the uncorrected DC offset?
-Chris
Correct me if I'm mistaken, but you are still feeding the DC correction signal into the diff pair aren't you? That would, if it be the case, unbalance your diff pair and you can't really allow that to happen. Unbalancing the diff pair directly affects it's ability to cancel audio signals and that will increase distortion.
What is the uncorrected DC offset?
-Chris
Hi Chris,
Isn't that unbalance you mention happening any way due to the NFB? AFAIK the servo acts on the NFB, acting on the DC part of it, the lower the better.
In any case, yes, the simulation does show a slight increase in THD when you use the servo. Whether it's positive or not, there are commercial products from Luxman or other brands, or audio designers like Erno Borbely who use DC servos instead of NFB caps. So in the end it might be a question of compromise, either objective or subjective, from the designer, isn't it? In the case of the 5M21, a DC offset trim is part of the design. On the simulation the DC offset is 156uV, which is very low.
The problem is not just the offset, but the distortion the NFB cap will add. That is what should be compared to the DC servo added distortion. Shouldn't it? If you can, have a look at the final simulations I did on what would be the prototype 5M21 clone. When a Luxman clone is not a clone
Carlos
Isn't that unbalance you mention happening any way due to the NFB? AFAIK the servo acts on the NFB, acting on the DC part of it, the lower the better.
In any case, yes, the simulation does show a slight increase in THD when you use the servo. Whether it's positive or not, there are commercial products from Luxman or other brands, or audio designers like Erno Borbely who use DC servos instead of NFB caps. So in the end it might be a question of compromise, either objective or subjective, from the designer, isn't it? In the case of the 5M21, a DC offset trim is part of the design. On the simulation the DC offset is 156uV, which is very low.
The problem is not just the offset, but the distortion the NFB cap will add. That is what should be compared to the DC servo added distortion. Shouldn't it? If you can, have a look at the final simulations I did on what would be the prototype 5M21 clone. When a Luxman clone is not a clone
Carlos
Hi Carlos,
To some extent - yes. The trick is to obtain the DC offset control outside of the main diff pair. So to do that you apply the correction elsewhere. You really, really want the diff pair to concentrate on the in band signals and not the out of band stuff, like DC offset.
This is a question of degree of course. But I have always found that a closely matched pair with minimal DC unbalance sounds the best by a long shot. Experiments in matching currents in the driver stage and output stage are effective ways of doing this. Of course that means at least attempting to match the complementary pairs of drivers. Match the outputs, and if they are close to the complementary outputs, so much the better. This also evens out the load on the VAS. Basically you would like to have the system naturally balanced. It's a tall order, but the closer you get, the better the sound tends to be. That is assuming a good design to begin with of course! All the matching and balancing in the world will not help a poor design.
-Chris
To some extent - yes. The trick is to obtain the DC offset control outside of the main diff pair. So to do that you apply the correction elsewhere. You really, really want the diff pair to concentrate on the in band signals and not the out of band stuff, like DC offset.
This is a question of degree of course. But I have always found that a closely matched pair with minimal DC unbalance sounds the best by a long shot. Experiments in matching currents in the driver stage and output stage are effective ways of doing this. Of course that means at least attempting to match the complementary pairs of drivers. Match the outputs, and if they are close to the complementary outputs, so much the better. This also evens out the load on the VAS. Basically you would like to have the system naturally balanced. It's a tall order, but the closer you get, the better the sound tends to be. That is assuming a good design to begin with of course! All the matching and balancing in the world will not help a poor design.
-Chris
Hi Chris,
I am using a dual fet pair at the input, so the matching issue is being taken care of.
Matching the output transistors is probably quite difficult, so it's not something I can count with. Same thing with the drivers.
On the drivers I might a few more and try to match them.
The design I already showed you, and I think is a very good one.
Carlos
I am using a dual fet pair at the input, so the matching issue is being taken care of.
Matching the output transistors is probably quite difficult, so it's not something I can count with. Same thing with the drivers.
On the drivers I might a few more and try to match them.
The design I already showed you, and I think is a very good one.
Carlos
In a multi-stage amplifier using loop feedback, the first stage's performance controls the performance of the entire amplifier. For example, any subsequent differential second stage offset would be reduced by a factor equal to the gain of the first stage. If there's a DC offset, and only one differential stage, it thus _must_ be the fault of the first stage, and so correcting it with a servo should actually re-balance the input pair from its imperfect state that causes the offset. Many power amplifiers only have one differential stage anyway, so I'm not clear how offset can be the fault of any stage other than the first stage.
For example, in a classic op amp, the second stage is a single ended transistor that has no inherent DC set point other than what the first stage tells it to do through feedback and the output stage, so it's not clear that correcting DC offset can be done anywhere else. Maybe there are specific circuits that I'm not familiar with where DC can be corrected elsewhere, but in a DC coupled amp using overall feedback, the first stage errors are what cause DC offset, and also where they can be corrected.
I'm not sure why correcting offset causes a change in distortion, but maybe it's that correcting DC offset subtly changes the standing current in the amplifier's devices, and it's just a change, depending on the polarity of the offset, and not always a degradation or improvement. This sort of behavior doesn't seem to exist with IC op amp amplifiers in my experience, and maybe is more of an issue with discrete, low feedback, higher distortion circuits that are more sensitive to individual device variations in the first place.
When I started talking about power supplies and amplifiers, my intention was to make this thread more concrete and focused on actual applications of caps (electrolytics or whatever) in specific circuits.
Different power supplies and applications, of course, and with subjective listenings on what changed with each different cap.
Different power supplies and applications, of course, and with subjective listenings on what changed with each different cap.
Hi Monte,
For proof, look at the BGW-750B/C schematic and read about the setup procedure. This works in real life. Citing a successful commercial design should settle any misgivings you might have about what is not the way most people approach the problem. It is one of the right ways to do this.
-Chris
I'm sorry, but that is not correct at all. If the driver and output stage tends to one polarity, the diff pair would normally be the stage that corrects for this. But in doing so it must unbalance itself. You can see this in practice. By correcting the tendency in the pre-driver stage (say), you can balance the rest of the circuit and allow the diff pair to run in a balanced condition. Of course, there will be minor corrections needed from the differential pair, but the effect of unbalancing the diff pair will be greatly reduced.
For proof, look at the BGW-750B/C schematic and read about the setup procedure. This works in real life. Citing a successful commercial design should settle any misgivings you might have about what is not the way most people approach the problem. It is one of the right ways to do this.
-Chris