Hi Key,
chee whiz, you make life difficult.
chee whiz, you make life difficult.
yes, but take account of that ratio.
ask Jens. Or read Self for an alternative viewpoint.
In my opinion, very few amps follow any sensible guidelines for the overall performance and how this could be affected by individual component values. Leach is one of the few to explain his reasoning (for much, but not quite all) and that gives the reader the opportunity to question/agree/disagree with his conclusions. Adopt or change to suit your own ideas, but at least they are informed decisions, or should be.
how does the cost of that 6u8 compare to the total cost of all the other amp components?
surely you've seen other posts from both myself and others quoting the ratio between the input filter, the NFB filter and the PSU filter to ensure the low frequency stability of an amplifier! i.e. avoid motorboating.
AndrewT said:
I don't have a specific reference, but I think he chose the breakpoint around the power bandwidth of the amplifier. I think he wants the main feedback to operate alone for the useful extent of the audio needs. The high frequency feedback it taken off before the output stage to reduce the interaction with the amplifier load.
AndrewT said:
I have seen your posts on this before, but I have never heard of a modern SS amp motorboating. It used to be a problem with tube amps, I think because of the output transformer phase change at low frequencies causing stability problems, if I recall. But I don't see any need to put a cap in series with the input if the amp doesn't motorboat without it.
keypunch said:
I've seen pros and cons too, including the idea that the thermal resistance through the top of the plastic package is higher causing poorer coupling. My motivation was at least partly laziness - I didn't want to unmount the entire output stage to drill and tap a hole to mount the Vgs multiplier when I made the change.
But then again OnSemi has their (fairly) new devices that include a compensation diode on the die.
AndrewT said:
I worked, but was slightly temperature over-compensated.
I tried 3 diodes and found that to be better when the diodes are placed on the output devises.
I desided to go for 4 diodes mounted on a small PCB on the heatsink (as you know)
\Jens
thanks Jens,
I wonder if Bob's obsevation that output device sensing being slower is related to the plastic insulation of the package delaying the temperature getting to the sensor?
Placing the three separate diodes, one to a package, would not help.
Using integrated diodes (as ONsemi) may be the solution to accurate and fast sensing.
Any feedback form successful or unsuccessful implementations?
I wonder if Bob's obsevation that output device sensing being slower is related to the plastic insulation of the package delaying the temperature getting to the sensor?
Placing the three separate diodes, one to a package, would not help.
Using integrated diodes (as ONsemi) may be the solution to accurate and fast sensing.
Any feedback form successful or unsuccessful implementations?
AndrewT said:
Andrew,
Sorry if my breaking up my prior questions into seperate quotes caused you more effort. It was not my intent. I thought the extra work I put in to do so would be more helpful than not. I guess I was wrong.
The UltraCap 6u8 PP against the total by way of my guessing some hardware (case/heatsink) is about $80.00-$100.00/amp (skiping other add on circuits), using the 6u8 in parallel with the NFB lower leg would be about $3.00 of that cost. The module cost exclusive of the output power devices is about $10.00 - $13.00, so the 6u8 would add $3.00 to the module cost. Even though this woudl be as high as a 30% added cost I think it would be worth it. I would still use a quality electrolytic as well, but for the tweeter based amps I may only the 6u8. and no other capacitor in parallel.
Actually with all my diyAudio reading and other internet audio reading I had not seen the ratio at all. I know it is hard to believe, but it is the case. The odd time something still arises I even stumble into on my own that I am surprised I had not encountered by osmosis before with all the reading and research I have done to learn. That is why I asked about the ratios ording. Now that you suggested I should have run into this ratio before now I will do some research on this later in week. At least I know about it now and can read up by doing some searching.
As far as the NFB lower leg capacitance I think I will be inclined for a general or woffer amp to choose a -3db frequency of about 6-7Hz. SL seems to suggest 10u0 for tweeter amps, which is nets out to be about 160Hz -3db point. and he seems to favour about around a 14Hz NFB Lower leg -3db point for general/woofer use.
Regards,
John L. Males
Willowdale, Ontario
Canada
29 November 2006 13:41
pooge said:
pooge,
Are you suggesting the capacitor in series with the signal input is nto really necessary? My understanding for using the signal input capacitor in series was to more to block DC than anything else. That said, then of course if one uses the input singal capacitor to block DC one needs to determine the dimension of the capacitor to minimize the sonic inpact of the input signal.
I know everyone agrees it is better not to use the input signal coupling capacitor in series with the signal input. There seems to be a general agreement it is better to use this input coupling capacitor just in case there is, or develops, unwanted DC offset element or DC to the input signal that can otherwise be amplified by the amplifier and cause serious problems.
Am I correct in my understanding above?
Regards,
John L. Males
Willowdale, Ontario
Canada
29 November 2006 13:55
BobEllis said:
Hi Bob,
Yes, some have also pointed out the top of the case of an output device (plastic, never seen done with a TO-3) has thermal resistance considerations. The basis of that is the die is closer to bottom of the plastic devices and/or the metal back to provide the thermal coupling.
Is your reference to not wanting to unmount the output stage to drill and tap hole for Vgs because the Vgs was moved or was added to the circuit after fact?
Regards,
John L. Males
Willowdale, Ontario
Canada
29 November 2006 14:01
Jens,
I have a few questions when you have a moment. If any of these questions have been answered in this or some other thread, just say so and look or go to the reference you cite. These are not about questioning you decisions, but to understand the decisions for my an possibly others education.
1) Why did you choose to use such a large value for the input coupling capacitor?
2) Why did you choose such a large capacitance for the NFB lower leg?
3) it was suggested you reduced the NFB HP capacitor from 47pF to 39pF to improve the better phase margin due to peaking nature of first order filters with same -3db corner frequency. Correct? If so, you determined this by simulationws, experience, math, listening and/or testing? Details for the answer are not required, I am just curious how you determined.
4) I noticed you even used 1u0 PE or PP capacitors for bypassing where most use 0u1. I have rarely see this on a selective basis, let alone for entire module. I am curious why you do so - as in advantage vs cost reasons?
I know you have have very good reasons why you make your decisions, so again I am not questing, I am just trying to learn.
Regards,
John L. Males
Willowdale, Ontario
Canada
29 November 2006 14:34
I have a few questions when you have a moment. If any of these questions have been answered in this or some other thread, just say so and look or go to the reference you cite. These are not about questioning you decisions, but to understand the decisions for my an possibly others education.
1) Why did you choose to use such a large value for the input coupling capacitor?
2) Why did you choose such a large capacitance for the NFB lower leg?
3) it was suggested you reduced the NFB HP capacitor from 47pF to 39pF to improve the better phase margin due to peaking nature of first order filters with same -3db corner frequency. Correct? If so, you determined this by simulationws, experience, math, listening and/or testing? Details for the answer are not required, I am just curious how you determined.
4) I noticed you even used 1u0 PE or PP capacitors for bypassing where most use 0u1. I have rarely see this on a selective basis, let alone for entire module. I am curious why you do so - as in advantage vs cost reasons?
I know you have have very good reasons why you make your decisions, so again I am not questing, I am just trying to learn.
Regards,
John L. Males
Willowdale, Ontario
Canada
29 November 2006 14:34
keypunch said:
As the name implies, the A75 is a class A mosfet amp. As designed the Vgs multiplier is mounted on the circuit board and the bias is heavy enough to keep the amp reasonably stable thermally. The 250W per channel waste heat was unbearable in the summer, since I biamp with four channels running at increased rail voltage.
I moved the Vgs multiplier to compensate the amps at lower bias. I was able to do this without totally disassembling the amp. There is still ~400W heating the room, but it's better than it was.
Hi Bob,
I wish we had kept the dual bias switch in the Krell Klones after the first PCB got superceded.
I am sure many users could have benefitted in the hot weather rather than bringing out the low bias alternatives until things cooled a bit.
Here in Scottish Borders, I may have to use low bias once in a blue moon (this might not translate for our non English speakers) or rarely.
I wish we had kept the dual bias switch in the Krell Klones after the first PCB got superceded.
I am sure many users could have benefitted in the hot weather rather than bringing out the low bias alternatives until things cooled a bit.
Here in Scottish Borders, I may have to use low bias once in a blue moon (this might not translate for our non English speakers) or rarely.
BobEllis said:
Bob,
I am familar with, but have never built, the A75.
I can appreciate the bi-amping as well as 4 channels heat generation, not to mention stiff electric bills you would have as a result.
So are you saying you reduced 250W * 8 (bi-amp 4 channels) to 400W / 8? If so you have done a great job doing so. I am sure that at least means not so many $$ to the hydro company. I can see you have to have at least one dedicated 15A circuit previously.
I choose not to consider Class A even for a Stereo, as in pair of channels due to not just he heat, but the electrical draw even at 30W rating. Even a Class A stereo amp would almost need its own dedicated circuit which I cannot have or arrange in my building. My target is to tri-amp and the number of channels involved is at least 7 due to HT 7.1, but using a DSP-3000 I have had for some time and the DSP-1 prior. That means at least 21 amps, likely 28. I cannot imagine even bi-amping 7 channels as first stage towards tri-amping, let alone tri-amping in Class A in terms of heat, skipping fact I could nto afford the hyrdo even if I coudl run at least 10 dedicated 15 amp circuits to supply the amps. I have thought about Class A for just the tweeters, but I still do not like the heat and wasted hydro.
I am very happy you have reduced the power and heat of your A75's. I think, all considered, placing the Vgs ontop of one of the output devices is well worth the extra time the amps take to stabilize with the Vgs.
Regards,
John L. Males
Willowdale, Ontario
Canada
29 November 2006 15:19
keypunch said:
I did 1 to optimize BOM after selecting 2
3 was selected after simulating and measuring on the board....with the BOM specified.
4 I had a lot of 1µF caps in stock - simple really 😉
Later I used 220nF and 100nF
\Jens
AndrewT said:
Andrew,
I have been thinking of using a switch to allow the O/P coil to be switched out or be able to change from standard to tweeter based amplifier for the larger value capacitor of the NFB leg. Was the Krell switch you refer to implemented as a relay or as a longish lead to a rear switch?
Regards,
John L. Males
Willowdale, Ontario
Canada
29 November 2006 21:40
JensRasmussen said:
Hi Jens,
Thanks for your reply.
Interesting corner frequency as a result of 1, like really really really low!
Thanks for the insight on 3.
LO LOL LOL for you EE's with a small store at home re 4 😉 If I understand the second part of your reply to 4 correctly in the actual build you used 220n0 or 100n0 rather than the 1u0's on the BOM.
Regards,
John L. Males
Willowdale, Ontario
Canada
29 November 2006 21:48
KP - it was only 2 biamped channels - 4 A75 channels dissipating 1KW total reduced to 400W. I decreased the bias until I heard a difference in sound then bumped it up 20%. Coincidentally, this matched the 100 mA per MOSFET pair generally regarded as the minimum bias.
I'm not as bad as some, it didn't take long to realize that certain items are used over and over again. If I buy 10 of anything more than once I'll buy 100 the next time to have some handy. I have at least 100 1 uf/50V MKTs on hand, and buy 100nf by the 1000. It's amazing how quickly that bag gets emptied out. I've got a pile of boards, heat sinks and transformers waiting in the wings for me. "Hey look at this great surplus deal at ApexJr - I know I'll use 2 dozen 15,000 uF/100V caps someday" is my downfall. 😉
It sounds like Jens built his first few Leach amps with 1 uf then used other in stock values. The original design's 100 nF works just fine for me, but my next ones will use 1 uF since I have them.
AndrewT - the dual bias switch would be great had it been included on the KSA100 boards. With the A75s I couldn't hear a difference at current bias so it didn't make sense to mess with doing it.
I'm not as bad as some, it didn't take long to realize that certain items are used over and over again. If I buy 10 of anything more than once I'll buy 100 the next time to have some handy. I have at least 100 1 uf/50V MKTs on hand, and buy 100nf by the 1000. It's amazing how quickly that bag gets emptied out. I've got a pile of boards, heat sinks and transformers waiting in the wings for me. "Hey look at this great surplus deal at ApexJr - I know I'll use 2 dozen 15,000 uF/100V caps someday" is my downfall. 😉
It sounds like Jens built his first few Leach amps with 1 uf then used other in stock values. The original design's 100 nF works just fine for me, but my next ones will use 1 uF since I have them.
AndrewT - the dual bias switch would be great had it been included on the KSA100 boards. With the A75s I couldn't hear a difference at current bias so it didn't make sense to mess with doing it.
"Quite why Leach arranges his input components this way is a mystery but he must have had a reason/s.
The Leach arrangement , which many others also adopt, deliberately unbalances the input conditions on the two sides of the LTP."
In a complementary diff input the bias current flowing through this resistor is due to the difference in beta between the NPN and PNP transistors. If the pairs were perfectly matched, no current would flow to ground, and there would be no input off-set voltage.
The Leach arrangement , which many others also adopt, deliberately unbalances the input conditions on the two sides of the LTP."
In a complementary diff input the bias current flowing through this resistor is due to the difference in beta between the NPN and PNP transistors. If the pairs were perfectly matched, no current would flow to ground, and there would be no input off-set voltage.
Hi Djk,
the cancellation between the two halves works quite well for reducing the absolute value of the input offset current. What it's not good at is creating a stable or near stable input offset current.
Look at the datasheets for some compensated opamp where trimming is used to nearly zero the input offset current after some other method of reduction has been applied. I think you will find that the input offset current and the input bias current are nearly the same, normal opamps have a ratio of about 10:1
now compare the typical and maximum for these two parameters. Enlightening!
The same happens in these dual complementary LTP stages if the gain for all four transistors are very similar.
The problem comes about because the LTP has a variable input offset current that depends on the conditions applied to the two junctions. Changes in current, voltage or temperature will cause changes in the gain and this gives rise to changes in the base currents (or is some of that the other way round?).
The two LTP thus have a variable input offset current. Now subtract these two variables from each other and the result is smaller but VERY variable. This will be exaggerated if the junctions are far apart, as in discrete assembly.
Some will deliberately choose the pair match in each LTP to be very close. Then choose the gain match between the two LTPs to be quite different. The low gain side is often the PNP side (NPN are often higher gain than PNP). This produces a slightly bigger input offset current but the cancellation leaves a bias in ONE direction and the value is more consistent with changes in environmental and electrical conditions.
Now back to Leach. He has shown the resistors attached to the inverting and non-inverting inputs in a different order and this means one can never balance the input conditions. The only reason I can see for doing it this way is to avoid the input attenuation that occurrs if the series resistor is in line before the grounding resistor and capacitor that form the two input filters. There may be another reason that over-rides the benefit of balancing. I do not believe the attenuation is that big a problem and I would always prefer the resistors to be placed to ensure balance. Go have a look at the steps Nelson P goes to, to ensure balance on the inputs of his amps.
the cancellation between the two halves works quite well for reducing the absolute value of the input offset current. What it's not good at is creating a stable or near stable input offset current.
Look at the datasheets for some compensated opamp where trimming is used to nearly zero the input offset current after some other method of reduction has been applied. I think you will find that the input offset current and the input bias current are nearly the same, normal opamps have a ratio of about 10:1
now compare the typical and maximum for these two parameters. Enlightening!
The same happens in these dual complementary LTP stages if the gain for all four transistors are very similar.
The problem comes about because the LTP has a variable input offset current that depends on the conditions applied to the two junctions. Changes in current, voltage or temperature will cause changes in the gain and this gives rise to changes in the base currents (or is some of that the other way round?).
The two LTP thus have a variable input offset current. Now subtract these two variables from each other and the result is smaller but VERY variable. This will be exaggerated if the junctions are far apart, as in discrete assembly.
Some will deliberately choose the pair match in each LTP to be very close. Then choose the gain match between the two LTPs to be quite different. The low gain side is often the PNP side (NPN are often higher gain than PNP). This produces a slightly bigger input offset current but the cancellation leaves a bias in ONE direction and the value is more consistent with changes in environmental and electrical conditions.
Now back to Leach. He has shown the resistors attached to the inverting and non-inverting inputs in a different order and this means one can never balance the input conditions. The only reason I can see for doing it this way is to avoid the input attenuation that occurrs if the series resistor is in line before the grounding resistor and capacitor that form the two input filters. There may be another reason that over-rides the benefit of balancing. I do not believe the attenuation is that big a problem and I would always prefer the resistors to be placed to ensure balance. Go have a look at the steps Nelson P goes to, to ensure balance on the inputs of his amps.
Hi Bob,
Sory I did nto do my math correct on number of A75's you had in your Bi-Amp configuration. Still made a great effort to keep the heat, hydro use and maintain the sound quality of your A75's. Hats off for you.
I looks like you and some otehrs here do alot of amp and other audio related building. I do not have such time. I did realize I needed lots of certain parts for some of the select projects I will build and some I may just see, like a small 10-15W Class A amp, to see if will suit what purpose I like to try them for. Where I may have my downfall is when friends of mine hear what I have built and ask me to build one for them as they do nto have the skills, knowledge (I have enough to be dangerous 😉 ) nor patience to build.
Like you when I see something I likely can use I pick it up as it is a deal. I know I will use lots of 0u1's and piicked up over 400 at excellent price. I have ctually looked over many scehmatics for amps, pre-amps and active crossovers to get a sense of the comon parts and when I see them at good price I pick them up. It just seems the ones doing the designing have a second career as a parts store where they are the only customer. lol lol lol lol
I happened to come across a 10uF PP 5% gold plated leads cap last night at a steal of price. This was by accident as I was not shoping for parts, just roaming the surplus store as I was in area. I gather you do same or via intenet cruising 😉 Sadly they only had the one of these 10u0 PP. I would have liked at least 8 for some tweeter specific amplifiers. The one at least allows me to test and see if I like 10u0 for the lower leg of the NFB, but at the price buying 10 would not have broken the bank. I only wish I could find some filter capacitors here in Toronto with voltage of 100V as you have run across at times.
Regards,
John L. Males
Willowdale, Ontario
Canada
30 November 2006 09:38
Sory I did nto do my math correct on number of A75's you had in your Bi-Amp configuration. Still made a great effort to keep the heat, hydro use and maintain the sound quality of your A75's. Hats off for you.
I looks like you and some otehrs here do alot of amp and other audio related building. I do not have such time. I did realize I needed lots of certain parts for some of the select projects I will build and some I may just see, like a small 10-15W Class A amp, to see if will suit what purpose I like to try them for. Where I may have my downfall is when friends of mine hear what I have built and ask me to build one for them as they do nto have the skills, knowledge (I have enough to be dangerous 😉 ) nor patience to build.
Like you when I see something I likely can use I pick it up as it is a deal. I know I will use lots of 0u1's and piicked up over 400 at excellent price. I have ctually looked over many scehmatics for amps, pre-amps and active crossovers to get a sense of the comon parts and when I see them at good price I pick them up. It just seems the ones doing the designing have a second career as a parts store where they are the only customer. lol lol lol lol
I happened to come across a 10uF PP 5% gold plated leads cap last night at a steal of price. This was by accident as I was not shoping for parts, just roaming the surplus store as I was in area. I gather you do same or via intenet cruising 😉 Sadly they only had the one of these 10u0 PP. I would have liked at least 8 for some tweeter specific amplifiers. The one at least allows me to test and see if I like 10u0 for the lower leg of the NFB, but at the price buying 10 would not have broken the bank. I only wish I could find some filter capacitors here in Toronto with voltage of 100V as you have run across at times.
Regards,
John L. Males
Willowdale, Ontario
Canada
30 November 2006 09:38
- Status
- Not open for further replies.