Dvv, Bob Cordell has shown it's not only the number of output devices, but also biasing them sufficiently. He recommended 150 ma per output pair for K1530/J201 Fet devices as a sweet spot. The trouble is the trade off between sufficient bias current per pair of devices, heatsink and power supply size, and SOA of the device.
I recall going from a Phase Linear 400 A to a Harmon Kardon 19, less power but instantly noticeable as less grainy and fatiguing.
I found the Carver Cube particularly hard to listen to, 3 pairs of mid sized output devices biased to the minimum level at a low voltage on a non existent heatsink with the signal tracking power supply. The T ( tube ) version had a tube transfer function and lower dampening factor was even less entertaining, unless you were Julian Hirsch. Carver used Holco resistors in the Lightstar preamp, it didn't save it.
I recall going from a Phase Linear 400 A to a Harmon Kardon 19, less power but instantly noticeable as less grainy and fatiguing.
I found the Carver Cube particularly hard to listen to, 3 pairs of mid sized output devices biased to the minimum level at a low voltage on a non existent heatsink with the signal tracking power supply. The T ( tube ) version had a tube transfer function and lower dampening factor was even less entertaining, unless you were Julian Hirsch. Carver used Holco resistors in the Lightstar preamp, it didn't save it.
What is the difference between GNFB and error correction (global or local)?
In GNFB, we apply a percentage of the output signal (signal +amp's distortion) in inverse to the input.
In error correction, we first remove the original signal from the output signal before to apply this "error signal" to the input. The goal is to add feedback without impacting the gain of the amp.
In a way, it is like moving an amplifier stage from the amp to the feedback loop.
But, both the comparator and error amplifier stage add poles and distortions to the feedback signal. This distortion, you will have-it at the final output of the system. The limited bandwidth of the comparator will create phase errors at the HF, making the removal of the original signal not accurate here. With gain and stability impacts. And it needs fine tuning if you want a lot of feedback.
I had played extensively with this, some decades ago, and finally, gave-it up: Not so much benefits, much more complicated and stability issues in production. (Just an opinion)
It is like multiplying stages (poles) in the amplifier open loop. Hard Rock'n'roll at HF !!!
My religion is, now, to try to design the best open loop amp as i can, then apply as much feedback as i can.
When i say the best open loop, it means low distortion in itself, and flat phase open loop up to 20KHz.
Because we want feedback to be applied with no phase error for all frequencies some can hear.
That no one was really able to achieve in any amplifier i know, with full satisfaction, because the limited speed of the actual active devices we are usually using.
Well, i was not so bad with my previous video link: Some minutes of silence at this funeral and the JC bashing party is back again ;-)
Last edited:
You don't have to resort to Voo Doo (voodoo) for a laugh.
People in Cameroon shed a tear at funerals (majority has the christian faith), then make music and laugh, with lots of food and beverages.
Added a Cameroon memento picture belonging to my g/f, she worked there for Médecins sans Frontières.
(Glen Kleinschmidt's A+ concept with HEC, at a more practical continuous power level)
Attachments
With all respect, Ticknpop, but you have obviously started reading this thread not so long ago, or you would know that many a word was written on the topic, including a bit by me.
I do not use MOSFETs in the output stages, only BJTs. It is my experience that devices like Toshiba 2SC5200/2SA1948, and Motorola/ON Semi's MJL 3281/1302 A do what they can do with incrased bias by about 120-130 mA of bias current (per device). More brings you only problems, and the only step above that is going full class A, and accepting the penalties of that.
Just my opinion.
Last edited:
Can-you simulate and publish the bandwidth/phase curves at the input of the VAS in closed loop ?
The best FT that I'm able with very fast CFAs is around 10KHz, by habit.
Last edited:
Of course Ron Quan found some popular low feedback amps had horrible FM/PIM and some of those awful op-amps were below the detection threshold. Did you even read his second paper?
Setting bias current usually involves more than just the device you use. Most often it is limited by the heatsink's capability to carry off the heat. For passive heatsinks, that are normal for hi fi gear (no fan), 100-200 ma is about the best you can do, depending on how many devices that you put in parallel.
A given heatsink can only handle so much power without getting too hot to touch. This is usually the limit for a practical design.
Actually, 500 ma/device would be better, but with the voltages used today, it is impossible to run this hard without overheating the heatsink, and even the devices. My first quality amp, used +/- 20V or so and 500 ma. It really worked best at the beta peak (500ma) rather than less or even more.
However in recent decades, my power supplies go from +/- 60V-90V, with a minimum of 2 output pairs to 9 output pairs/amp channel. Now, .5A/device would overwhelm the heatsink, and even stress the safe area of the output devices. We use more devices in parallel to get more peak current, a better 'safe area' of operation, and a better spread of the generated heat on the heatsink selected. Even a relatively large heatsink will overheat with 9 output devices per side at idle currents higher than 200ma or so.
A given heatsink can only handle so much power without getting too hot to touch. This is usually the limit for a practical design.
Actually, 500 ma/device would be better, but with the voltages used today, it is impossible to run this hard without overheating the heatsink, and even the devices. My first quality amp, used +/- 20V or so and 500 ma. It really worked best at the beta peak (500ma) rather than less or even more.
However in recent decades, my power supplies go from +/- 60V-90V, with a minimum of 2 output pairs to 9 output pairs/amp channel. Now, .5A/device would overwhelm the heatsink, and even stress the safe area of the output devices. We use more devices in parallel to get more peak current, a better 'safe area' of operation, and a better spread of the generated heat on the heatsink selected. Even a relatively large heatsink will overheat with 9 output devices per side at idle currents higher than 200ma or so.
Let me make a couple of up-to-date comments about IC op amp selection and use.
There are some pretty darn good IC op amps. Scott's AD797 is darn good, and often used in precision measurement equipment, because it measures so well, but it IS finite in slew-rate (20V/us) and could be problematic for RFI etc. However, it is RELATIVELY expensive and is NOT the USUAL op amp selected for either mid-fi or pro audio use. Therefore, what most of us can afford, does not use the AD797 or its rough equivalent.
No, they use devices like the 4580, because they are many times cheaper to buy, and seem to measure 'OK' with traditional measurements. I would not use the cheaper/dirtier IC's, but almost everybody else does.
Now, IF you are careful, can you make an 'acceptable' preamp or phono stage with just IC's? Yes, IF you are careful not to load the outputs excessively. Guess what? Almost everybody excessively loads the outputs with both the feedback resistors and the external load with lower than optimum impedance. They do this to make the design work as a system, AND it is difficult to measure any problem without an advanced look at FM distortion. However, Ron Quan has done it again. He has measured FM distortion and its relatives in a number of TYPICAL IC op amps, that are and have been used by the multitudes for many decades. Beyond hi end, they are the only competitive IC's used.
Guess what? They don't measure that well, typically.
There are some pretty darn good IC op amps. Scott's AD797 is darn good, and often used in precision measurement equipment, because it measures so well, but it IS finite in slew-rate (20V/us) and could be problematic for RFI etc. However, it is RELATIVELY expensive and is NOT the USUAL op amp selected for either mid-fi or pro audio use. Therefore, what most of us can afford, does not use the AD797 or its rough equivalent.
No, they use devices like the 4580, because they are many times cheaper to buy, and seem to measure 'OK' with traditional measurements. I would not use the cheaper/dirtier IC's, but almost everybody else does.
Now, IF you are careful, can you make an 'acceptable' preamp or phono stage with just IC's? Yes, IF you are careful not to load the outputs excessively. Guess what? Almost everybody excessively loads the outputs with both the feedback resistors and the external load with lower than optimum impedance. They do this to make the design work as a system, AND it is difficult to measure any problem without an advanced look at FM distortion. However, Ron Quan has done it again. He has measured FM distortion and its relatives in a number of TYPICAL IC op amps, that are and have been used by the multitudes for many decades. Beyond hi end, they are the only competitive IC's used.
Guess what? They don't measure that well, typically.
Dvv, actually I have been around since the first hundred pages, have built a Blowtorch clone with 389/109 IDSS 17ma inputs, quad 2SJ72/k147 input phono stage, all the rest of my system is all discrete direct coupled jfet/fet amplification path and Class A Borbely power amp running 500 ma per Hitachi device. 2.5 A standing current. Yes not a Krell Audio Standard for device count, but 5 pairs are good for 100 watt Class A per channel.
Ticknpop, using a basketful of output devices is not the point, and was in fact used as a cheap trick to impress the customer. What use so many power devices if fed from teeny weeny 6,800 uF caps for more than 200WRMS per side?
Not to repeat what has already been said, see John's note a few messages back and you will see a high level of correlation with what I said. For good peak current, I will use a maximum of 4 pairs of devices, and if biased at say 130 mA each, that gives me a constant power dissipation high enough to be just under the capacities of a medium size heat sink per channel.
130 mA is an arbitrary value, quite simply that's what I found most amps will do well with, but the actual value will have to go from case to case. Over that value, and some designs will sound a bit sweter, but may also lose some focus and detail, others will take more, and so forth.
The other problem are the voltages used. If one wants a lot of power, one will have to use higher voltages, so with the same bias current power dissipation will rise. One must also take that into account, given the cost of decent heat sinks and worse, their quickly swelling areas required.
BTW, no insult intended, it's just that I haven't seen your handle before, but then I got in it seems way after you.
Not to repeat what has already been said, see John's note a few messages back and you will see a high level of correlation with what I said. For good peak current, I will use a maximum of 4 pairs of devices, and if biased at say 130 mA each, that gives me a constant power dissipation high enough to be just under the capacities of a medium size heat sink per channel.
130 mA is an arbitrary value, quite simply that's what I found most amps will do well with, but the actual value will have to go from case to case. Over that value, and some designs will sound a bit sweter, but may also lose some focus and detail, others will take more, and so forth.
The other problem are the voltages used. If one wants a lot of power, one will have to use higher voltages, so with the same bias current power dissipation will rise. One must also take that into account, given the cost of decent heat sinks and worse, their quickly swelling areas required.
BTW, no insult intended, it's just that I haven't seen your handle before, but then I got in it seems way after you.
Or you can add a pair of complementary BJTs as current boosters. It's usually not done because people seem to believe that to be an unacceptable sacrifice in maximum output voltage. My choices usually are MPSA 05/56 or BC 639/640. Reasonable gain, good linearity, easily available, can handle current peaks of over 1A, and still cheap.
Others, such as say reVox, do it routinely
Personally, I find that simple trick to take the ball game to a new level. Haven't heard an op amp yet which did not benefit from it sonically.
But you're right, the industry-at-large still regards op amps in general as cost savers rather than bona fide parts which need to be properly catered for if you want their best.
Dvv, you are very specific. I find that more devices make it easier to create a more Class A design, because you can use the bipolar output devices to spread the source of the heat. This makes the heat sink much more efficient. Secondly, you are not so compromised by the non-conductive spacer that is usually necessary. If you have (for example) 9 spacers instead of 2 spacers, you have 4.5 times the surface area on the spacers and this lowers potential temperature difference across the spacers. It is always a balance between how many output pairs and the optimum Idle current. More current REQUIRES very small emitter resistors. I have successfully used 0.05 ohm wirewound resistors for this, but you have to have a really good thermal tracking system to get away with such a small Re.
Be careful with discrete output devices. Studer Revox used them with a CLASS C bias (or dead zone) and they measured as such. Best is a true high speed IC buffer within the feedback loop, OR a truly class AB biased complementary jfet or bipolar pair.
I showed the measured results of the Studer line amp, back in 1975 to Willi Studer himself. Was he 'pssed' off! Lots of higher order distortion added to the relatively low annoyance distortion of the analog magnetic tape.
I showed the measured results of the Studer line amp, back in 1975 to Willi Studer himself. Was he 'pssed' off! Lots of higher order distortion added to the relatively low annoyance distortion of the analog magnetic tape.
I'll produce evidence when someone provides the test gear, the environment, a wage to do the job - that's called "motivation" ...
As has been expressed by many, ad nauseum, the "evidence" is very difficult to gather. We are dealing with here, firstly, how the ear/brain reacts to the level and type of distortion it hears; and secondly, that the signs of the crucial distortion mechanisms are barely suggested at by conventional measuring methods. The farce of the current audio industry is because current measurements are so feeble as predictors of subjective audio quality.
The evidence, for me, is the strength of the cause and effect linkage - I hear a problem, I apply a fix from experience that's appropriate and the sound improves, subjectively. Terribly bad science
, I know, but it works ...
- Status
- Not open for further replies.