2Q CCS's for alan's higher current OPS app.
With a input stage diamond that is either decoupled or regulated ,
LED CCS's work well.
Reference from rail to rail , any ground connection makes PSRR
quite inferior.
Whatever CCS he uses (in the OPS) , it's unique Tc will have to be factored
into the OPS's total thermal behavior. It was this complexity
that kept my design a simple EF3 with just 6 PTC Vbe's to worry
about.
PS - lower powered OPS powered by regulators - I used LED CCS's (below).
OS
Attachments
Last edited:
All I can say is that lower Re sound better, you should shift the bias current value when changing them but I can clearly hear that lower Re sounds better and I don`t have extra bad *** speakers. Also different Re value wants different bias current so you could get the best sound at crossover region, there is no magic number for bias current and you should go for the lower base stoppers also.
All that I noted I can hear at home without any measuring instrument but I listen in mono with 2 Ohms load (two paralleled 4 Ohm speakers)
Interesting.... I too heard greater resolving/clarity after lowering the Re.... but it tended to be easily blown up when connected/disconnecting cables etc especially if the supplies had not completely drained to zero, first. It was more robust with higher Re.
It has been my understanding the distortion was always lower with lower Re (single stage). Which is why I tried lowering the Re to .1 Ohm. And, it did sound better.
BUT, later on further reflection, I didnt change the protection circuit sensitivity at the same time..... leading to occasional damage but also the cleaner sound could be caused by less intrusion by the over-current OPS protection circuitry (SOA). That would need to be tested by removing the OPS protection and listening with .1 and .2 Re. Something I have not done. I'm just leaving it at .1 (rebiased to stock Id setting....never had a thermal run-away issue or bias stability issue) and being more careful.
Question..... any other ideas ... why does the changing of Re make audible change when this is a moderately high gnfb... VFA.... amplifier topology?
THx-RNMarsh
Last edited:
There is always poison in OPS, thermal stability and current sharing being two of the most significant. Current sharing is improved by having larger Re. Larger Re also lowers the optimum idle current. I see both being very beneficial for making durable and reliable amplifiers. I use several pairs with 0,47ohm and 60mA idle
I seek my improvements in OPS performance by making the output transistors close to non switching with very gradual turn on/off. Only when pushed very close to the rails will the symmetric pairs stop conducting.
I seek my improvements in OPS performance by making the output transistors close to non switching with very gradual turn on/off. Only when pushed very close to the rails will the symmetric pairs stop conducting.
You are missing my whole point. You have to compare apple with apple. In order to compare, you have to:
1) Keep the total idle current equal. that is if you run 100mA per stage in 5 stages, you have total or 500mA. If you run 3 stages, you have to run 167mA per stage to be equal. You are thinking about only running 300mA with 3 stages. This is not equal by any stretch.
2) Keep the total output power equal whatever that might be. Not 60% of the power.
I stress 1) because that will give you the same Class A region in the most critical small signal where you listen most of the time.
If you truly make it equal, that the Class A region and total output power remains the same, there is only disadvantage using 3 stages. You push the SOA of the transistor more in 3 stages.
If you reduce everything by 40%, of cause 3 stages generates less heat. There is no disagreement there.
You design an amp, you better know what rail voltage you plan to use and choose the right value of R3 in your schematic. I choose to run either 40V or 25V, you can see in my picture I have a 10K parallel onto the original 10K 1/2W resistor on the far right to make 5K for 25V as I am using 25V for testing. Later I can remove that to use for 40V rail.
I don't like two diode as you need more parts. I don't like to use extra parts unless I have to, more parts, more space on pcb and might compromise layout. I definitely don't like 2 diode because it's only 1.4V. I only get 0.7V drop to control the current. 3 diodes gives 2.1V, eating more headroom and one more component. The only way to get 1V across the current setting resistor is 2 diode and 1 schottky diode!!!
It's my personal preference. I never seen using LED for biasing in the industry, I am not comfortable with LED. I don't know the tempco of LED and what is the voltage variation. More important, they as specified at 1 or 2 mA, I want to run at least 4 to 5 mA.
I actually gave this some thoughts and I decided to go this route. Layout is important for me, it is a lot more components doing the diamond and so complicate to get the critical transistors to screwed together to get the best thermal tracking, this is the last thing of my concern.
Last edited:
1) Why would anyone really care what the total idle current of all the devices is? We are biasing each transistor to eliminate crossover distortion.
2) Unless heatsinks are horribly undersized, total output power will only be creating heat into a dummy load on the test bench. Idle current of the devices creates the heat in the real world unless the amp is pro-sound or subwoofer duty. Most people don't have their home system blasting full power for long periods of time. If they do, chances are they would never hear the difference in idle current anyway.
Here's a friendly suggestion that may help you with your build. There are many cheap amp boards available all over this website. It would likely help you a lot to put a couple together and experiment a bit. Figure out some of this stuff hands on. It would likely save a lot of arguing obvious points. It's easy to miss the basic important stuff by overthinking things that are really irrelevant.
I am not saying one way or the other as I am new. I have no idea which sound better. But please look at my DC calculation in post #5645. If you go through the calculation, using one pair of output devices with 0.47ohm Re will turn off Q1 when driving large signal going negative. Of cause if you use a few pairs in parallel, the effect will be divided by the number of pairs and works better.
I don't know my assumption is correct as nobody comment on my calculation.
That's where we disagree, I care a lot about the class A region. when I increase the bias of my Acurus, I can hear quite an improvement in transparency even though I did not even decrease the Re and violates the Oliver's condition.
that's the reason I design with 5 pairs from the start to get the bias current up and still can attempt to conform to Oliver's condition.
Last edited:
Miib is right , "choose your poison".
Higher Re , more reliable/cooler - Lower Re , maybe better sound.
How about better Re , or non-inductive Re ?
I just ran my 4 year old triple variant , Alan's variant , and
a classic Locanthi.
A Plain old 3EF can ...
- run the lowest driver/predriver currents.
- have the most (predictable) thermal compensation.
- and have the lowest THD.
The only thing the diamond and CCS's do is replace the driver "suckout" cap
and burn off more current. Underbiasing the CCS's in Alan's or removing
the cap in the Locanthi do the same thing to distortion.
So , the tradeoff is a simple Vbe , no cap , more heat versus having
the cap and a compound bias spreader (and less heat).
I had alan's design and the Locanthi nearly equal in simulation. Alan's
required 20-30ma current to have the same performance as my
8ma setup.
Glad I did this , and glad I waited 4 years.
OS
Higher Re , more reliable/cooler - Lower Re , maybe better sound.
How about better Re , or non-inductive Re ?
I just ran my 4 year old triple variant , Alan's variant , and
a classic Locanthi.
A Plain old 3EF can ...
- run the lowest driver/predriver currents.
- have the most (predictable) thermal compensation.
- and have the lowest THD.
The only thing the diamond and CCS's do is replace the driver "suckout" cap
and burn off more current. Underbiasing the CCS's in Alan's or removing
the cap in the Locanthi do the same thing to distortion.
So , the tradeoff is a simple Vbe , no cap , more heat versus having
the cap and a compound bias spreader (and less heat).
I had alan's design and the Locanthi nearly equal in simulation. Alan's
required 20-30ma current to have the same performance as my
8ma setup.
Glad I did this , and glad I waited 4 years.
OS
Can you explain a little more, your comments are valuable. I don't know Locanthi.
I go for diamond mainly for the thermal tracking capability and may be little improvement in large signal linearity. I was thinking about in the diamond, the current through the pre-driver EF decrease while the current in the driver increases, or vise verse. The Vbe variation are always opposite between the two and have a first order canceling effect and lower distortion.
Is I suggested earlier, Why not experiment a bit with your Acurus amp and see why it sounded better? Maybe your class A idea may have some merit, maybe not, but making an assumption proves nothing. You will never know by arguing theory. You may know more by actually testing real world parts. Personally I couldn't care less if it's class A, AB, B, C, D.... The proof is in the finished product. It sounds good, or it sounds great. Some of the most unlikely designs I've assembled to date fit in the great category. I've been lucky so far in having none sound awful.
I come from an automotive background. When you are building a race car, the first thing a sensible new designer would do is go to the race track and see what everyone else is doing. Look closer at the car that is winning. The worst thing a designer could do is blindly build from a possibly wrong theory with no real world experience behind them. This is why I suggest to build a couple predesigned amps and tinker with them a bit. Ostripper's got some great designs. So does Vzaichenko.
Ha ha, I can't argue with real experience and I can't argue with people that actually listen to the design. I am talking a lot here lately because you guys convince me to use 0.22ohm resistors, I am waiting for them to arrive. Until then, I can only talk!!!
I am trying to build a few, remember, I am just starting. This is my first one. I don't have most of the right parts, every time I change something, usually I have to order it. They usually have to come from China. try looking for a 0.22ohm 3W in Digikey and Newark, they are all either super expensive(AND non stock) or wire wound.
So I am stuck in this mode for another week minimum. You have to look at my room with all the almost stuffed pcb laying around waiting for parts.
I always order extra amount on the components, hopefully one day, when I need a component value, I can just go to the draw and get it. Not now!!!
Not to invalidate your point, I spent at least 15 years on the bench instead of staying in my office doing bench work. In the last years of my career, I changed and pull back from the hands on approach. I find my mind is clearer when I am not in the mist of everything. But that's just me.
Last edited:
Besides the thermal modeling , I saw no difference in any FFT - cancelling.
NFB and the IPS take care of that.
What I did was simulate with a 5ppm blameless and A/B compared
The 2 OPS's. The EF3 also was "faster" , could do >350v/us (CFA).
Simple is better for speed .....
Locanthi is just your standard 3EF (class A /class A / AB -outputs).
PS - It's just better to leave the pre-drivers right out of the thermal
"picture" - run them at <3ma.
Since you want to pursue this diamond variant , you should continue
that EF3 thread I posted -
OS
Yes, very expensive at 20 cents apiece at your corner store, with next day delivery: MOSX3CT631RR22J KOA Speer | Mouser
Besides the thermal modeling , I saw no difference in any FFT - cancelling.
NFB and the IPS take care of that.
What I did was simulate with a 5ppm blameless and A/B compared
The 2 OPS's. The EF3 also was "faster" , could do >350v/us (CFA).
Simple is better for speed .....
Locanthi is just your standard 3EF (class A /class A / AB -outputs).
PS - It's just better to leave the pre-drivers right out of the thermal
"picture" - run them at <3ma.
Since you want to pursue this diamond variant , you should continue
that EF3 thread I posted -
OS
Thanks
1) Why is regular 3EF faster than diamond? If I run the pre-driver hotter so I never run out of steam, I should get the same speed. Yes, I might have to have more current, but even at 20mA is hardly in the big picture.
2) I have spent a little time reading how the CFA works. Most of the articles are from Ti, LT etc. that explain about their opamp. I yet to find a book that explain about the pros and cons of different variations like how Mr. Cordell did in his book on the LTP IPS. Do you have any suggestion in a book or articles that explain how to optimize CFA for low distortion? I have every intention to do a CFA in the next go around. Just not now, I want to do a conventional one as the first project.
3) Why leaving the pre-driver out of the thermal picture? I thought it's a big advantage of diamond being able to use the pre-driver to compensate the driver.
I would expect my OPS can keep up with your Slew Master stuff. I started out with 3EF(diamond), I use a very big driver transistor with 60MHz fT to drive the big output transistors. I use high current to increase the speed and keep it in Class A. I never done the simulation on the OPS alone, but I would expect mine can keep up with your fastest circuits.
If you don't think so, can you explain?
Thanks
Alan
Most CFA are ugly beasts, I personally found some issues that I don't really like about them primarily that the feed back signal is voltage, that is then transformed into current over a smallish value feedback resistor network, the smaller the values are in the feedback network, the better this V to I transform is, price is heat.. So one have to settle at a compromise with not ideal network values. The advantage of CFA is that you omit a V to I device in the feedback path
All the CFA's I have built over the last years seem to lack bite and definition in the lowest registrers, I contribute this to dynamic effects with a mish mash of Voltage and Current modulation in the feedback path.
All the CFA's I have built over the last years seem to lack bite and definition in the lowest registrers, I contribute this to dynamic effects with a mish mash of Voltage and Current modulation in the feedback path.
Hi Alan,
I am leaning closer to giving you your own thread. Just saying.
You appear to be "book smart", but without the actual bench smarts that should accompany the way you demand other folks do the proof for you. Asking the odd question is fine, but you are taking an entire course in this thread, and that is mighty unfair to the other readers who may not have a chance to ask Bob questions or make helpful comments.
Buying parts from China and complaining about the travel time isn't a sensible thing to do. You can buy non-inductive, wire wound resistors from Digikey. I have used them. Shipping time to me in Canada is one single day at a cost of $8 to go through customs and ship. Keep in mind that "non-inductive" will never be exactly true, more like greatly reduced inductance. Another resistor for low inductance is the plate variety used in many far east amplifier builds. I have found some to have a strong positive temperature co-efficient. Frankly, that might be more of a problem than some slight inductance.
Never seen red LEDs used as the voltage reference in commercial designs? I have, many in fact. The tempco is pretty easy to figure out if you build the thing and measure it. As a hint for you, a 3mm led will fit inside the mounting hole on a TO-126 package. You may have to file some LEDs down slightly, but they are very, very close to a snug fit. Encase this in foam and you can approximate a zero tempco depending on relative current densities. Theory should help you there. Use the older, non-efficient 3mm red LEDs.
Diamond buffer circuits can have odd tempcos of their own. I have built some over many years. Experiment on the bench and forget the simulator for these simple circuit blocks.
Are you using 1% metal film resistors from China? Measure the the tempco of those, then buy a Dale 1% part and try the same thing. That should be a very quick lesson on what to expect from various parts.
I'll just say in closing that you are talking down to a bunch of people who have a great deal of experience. Don't do that.
-Chris
I am leaning closer to giving you your own thread. Just saying.
You appear to be "book smart", but without the actual bench smarts that should accompany the way you demand other folks do the proof for you. Asking the odd question is fine, but you are taking an entire course in this thread, and that is mighty unfair to the other readers who may not have a chance to ask Bob questions or make helpful comments.
Buying parts from China and complaining about the travel time isn't a sensible thing to do. You can buy non-inductive, wire wound resistors from Digikey. I have used them. Shipping time to me in Canada is one single day at a cost of $8 to go through customs and ship. Keep in mind that "non-inductive" will never be exactly true, more like greatly reduced inductance. Another resistor for low inductance is the plate variety used in many far east amplifier builds. I have found some to have a strong positive temperature co-efficient. Frankly, that might be more of a problem than some slight inductance.
Never seen red LEDs used as the voltage reference in commercial designs? I have, many in fact. The tempco is pretty easy to figure out if you build the thing and measure it. As a hint for you, a 3mm led will fit inside the mounting hole on a TO-126 package. You may have to file some LEDs down slightly, but they are very, very close to a snug fit. Encase this in foam and you can approximate a zero tempco depending on relative current densities. Theory should help you there. Use the older, non-efficient 3mm red LEDs.
Diamond buffer circuits can have odd tempcos of their own. I have built some over many years. Experiment on the bench and forget the simulator for these simple circuit blocks.
Are you using 1% metal film resistors from China? Measure the the tempco of those, then buy a Dale 1% part and try the same thing. That should be a very quick lesson on what to expect from various parts.
I'll just say in closing that you are talking down to a bunch of people who have a great deal of experience. Don't do that.
-Chris
Last edited:
I’m under the impression that “Locanthi” additionally implies that the emitters of the drivers are connected via resistance that is not split with the midpoint connected to the output.
Isn’t it nicer to have something that will work on a range of rail values without alteration?
Apart from that, with your design at 40 V, if the voltage goes up 5%, the current goes up 6.4%. With the 2 diode alternative, the current goes up 0.7%. Presumably you are using an unregulated linear supply to give your 40 V rail? So as that rail is bouncing around, you’re going to get a nice 100 Hz component in the output of your current sources. I’d just like to hear from more people whether this is a problem, because as you say, your arrangement is simpler!
I don’t see the problem here. A smaller drop seems better to me as it gives you a wider compliance range of the current source.
I apologize for not being able to respond sooner and more frequent. Have a day job and it is a busy one....
I would think that is how one makes things worse unless I got you wrong. Cutting a slit between transistors would impede thermal transfer across the heat sink, can't see how it could even out the temperature. Maybe I miss-read you.
Alan, the OPS are to idle at slightly under 60mA per device, and the total heat sink dissipation at idle, including that of drivers and pre-drivers, is about 46W. The heat sink in picture measures 300W x 150L x 50H mm^3, has 28 fins. The 50mm height is made up of a 10mm base height and 40mm fin height, possibly came out of the same factory somewhere in China as did the heat sink in one of your recent pictures. I expected it to be ample for a 200-watt amp for domestic use, and was not concerned about the thermal problems as much as the VAS standing current problems, even with the OPS transistors lumped together.
I would think that is how one makes things worse unless I got you wrong. Cutting a slit between transistors would impede thermal transfer across the heat sink, can't see how it could even out the temperature. Maybe I miss-read you.
Alan, the OPS are to idle at slightly under 60mA per device, and the total heat sink dissipation at idle, including that of drivers and pre-drivers, is about 46W. The heat sink in picture measures 300W x 150L x 50H mm^3, has 28 fins. The 50mm height is made up of a 10mm base height and 40mm fin height, possibly came out of the same factory somewhere in China as did the heat sink in one of your recent pictures. I expected it to be ample for a 200-watt amp for domestic use, and was not concerned about the thermal problems as much as the VAS standing current problems, even with the OPS transistors lumped together.
I'm trying to catch up on this thread after being out of the loop for a couple of days. Sorry if some of my comments are redundant or obsolete.
With respect to driver current and whether it turns off or not, and for the Locanthi-like topology where a single emitter resistor establishes bias current between the P and N drivers, the driver generally will remain in class A and not turn off. I consider this generally desirable, so that only the output transistors are not in class A. I also prefer fairly high idle current in the drivers so as to readily sweep out minority carriers on the turn-off slope.
In particular, at higher frequencies it becomes more difficult to switch off the output transistors quickly enough. This is especially the case when driving high current into lower-impedance loads because there are then more minority carriers that must be swept out by reverse base current to turn the power transistor off. This problem is worse for output transistors with low ft.
You never want the driver transistor to lose control over the output transistor.
This phenomenon is referred to as "secondary" crossover distortion or "dynamic" crossover distortion. I believe it was originally explained by either Bart Locanthi or Jim Bongiorno, not sure which one at this time.
Some authors and designers advocate the use of a "speedup" capacitor across the driver emitter resistor in order to allow the driver transistor on the other side to help pull current out of the power transistor base. It effectively attempts to make the driver push-pull. This approach helps, but is less effective than just running the drivers at higher bias current (20-60mA). I touch on this in my book.
When you simulate your power amplifier, look at the waveform of the current in the driver transistor when the amplifier is driving full power at 20kHz into a 4-ohm or even 2-ohm load.
Cheers,
Bob