I can get the sanken's from companies like Profusion (and be guaranteed they aren't fakes), but when they are nearly £5 each this is out of my price range. Especially when you consider I can order many OnSemi devices and pay only £10 for shipping.
Cheers for the recommendation, I'll look into it.
Had planned on around 50mA bias per pair. Can't see any crossover distortion on LTSpice even at 30mA. We shall see.
Try not to laugh at me, but can you elaborate on what you mean by "open loop linearity" ? I'm still learning about these things...
Does low THD figure explain why I have seen many an amplifier schematic that has very high feedback of 40dB or more? ~30dB seemed to be the usual figure for feedback in most of the circuits I looked at.
As long as it sounds good 🙂 I have been quite surprised by the schematics of some commercial amplifiers I thought sounded excellent, most were just regular EF stages with very simple designs. Most now seem to use these Sanken SAP devices.
Most of it has been designed from reading about the Leach low TIM amplifier as his web page is very thorough and gives step by step explanation into each section of the amplifier. I would probably build the Leach myself but I really don't need the 250W power it produces... 100W or thereabouts into 8 ohms is enough for me.
There is a sort of tradition/standard in the world of pro (as in touring) amps that a 1Vp (or is it RMS?) at the input will result in maximum (or at least rated) power at the output. This is translated in to a "sensativity" spec. This is, I surmise, because in that context one usually buys more power for the express objective producing louder sound. Consequently, if this relationship is to be maintained, gain (Vo/Vi) has to increase along with rated power.
Obviously (to me anyway) there is no reason why this practice needs or should be followed for home amplification.
EF gets a bad rap in some circles. Your design moots one significant part of the objections to EF simply by the presense of R25 and R26. These serve to force the drivers of one half "off" just as their complements are starting to conduct thus reducing that part of crossover distortion caused when both top and bottom conduct simultaneously. The other major objection to EF is that as the output transistors heat or cool with the level of the program the bias does't track track exactly plus there is a (sensible) tendency to under bias a little bit anyway out of fear of thermal run-away. Many designers of EF ytages have concluded that the real world isn't as bad as feared. EF stages seem to tolerate quite a bit of under/over biasing before anything audible (or easily measurable) appears. Placing the Vbe multiplier as close an output device as possible helps further. Some other designs may have lower crossover distortion when set optimally, but the performce degrades more quickly when bias drifts away from optimal.
There are many counter arguments to the above. This is one of those "discussions" that show little sign of reaching a definive resolution.
Currently I've got 1VRMS for 100W out into 8 ohms. I'm thinking of decreasing the gain a little to give some headroom (much more than 1VRMS causes clipping), and increasing the power supply rails to 55v.
This is exactly what I saw when I simulated the design, as originally I tried all three. The standard EF had crossover distortion, the CFP was murder to stabilise especially at high frequency. The Triple was the only one that behaved well, so I chose it. This seems to be the conclusion Leach came to as well.
One thing I did think of trying was the driver arrangement that Knut Nygaard uses in his "Classical Class A" amplifier:
http://home.no.net/andiha/articles/audio/claclassa.htm
This complicates things though as it means more current sources.
When built it will be an MJE340 (or some other TO-126 packaged device) bolted between the outputs directly on the heatsink.
I did think about trying the OnSemi ThermalTrak devices...
That's better left as a variation to an existing, functioning design that one already has some experience with.. One thing you need to know ahead of time is the approriate bias voltage. The Thermal Tracks have a limited finite set of possible bias voltages available set by the drop per diode and the number of devices. This makes for some design constraints: on the one hand you may have to slip in an extra diode of some sort or you may have to jumper one or more of the integral diodes OR a combination of the two. I think it might work to "shim" the bias with either series or parrallel pot. All this gives up some of the simplicity and in the case of the pot you have to think about what happens if it fails either open or short.
Hi Jaycee,
I am not sure you understand the difference between amplifier gain and global negative feedback.
In post 22 you say "Does low THD figure explain why I have seen many an amplifier schematic that has very high feedback of 40dB or more? ~30dB seemed to be the usual figure for feedback in most of the circuits I looked at."
Very few designers quote the open loop gain and the closed loop gain to allow you to check the global feedback. Leach is the exception and his is a low feedback design.
Your figures of 30db to 40db quoted above are almost certainly amplifier gain figures. Domestic HiFi & music system power amps often have a gain in the range +26db to +35db. PA equipment tend to have gain about 10db lower than domestic.
Regarding your power requirement of 100w into 8ohms keep to a 35Vac transformer giving +-50Vdc on the PSU rails. Leach will work at this voltage.
The standard Leach is about 120W into 8ohm but due to 2 pairs of robust output devices is capable of driving a moderate 4ohm speaker. Be carefull with the Hi Ft (2sa1943, 2SA1302, 2SA1386) devices they usually need 50% more devices to match the slower devices (MJ15003/4 & MJ21193/4) for reliability particularly when you go to plastic cased (To264) which have a Tjmax of 150degC rather than Tjmax of 200degC.
I am not sure you understand the difference between amplifier gain and global negative feedback.
In post 22 you say "Does low THD figure explain why I have seen many an amplifier schematic that has very high feedback of 40dB or more? ~30dB seemed to be the usual figure for feedback in most of the circuits I looked at."
Very few designers quote the open loop gain and the closed loop gain to allow you to check the global feedback. Leach is the exception and his is a low feedback design.
Your figures of 30db to 40db quoted above are almost certainly amplifier gain figures. Domestic HiFi & music system power amps often have a gain in the range +26db to +35db. PA equipment tend to have gain about 10db lower than domestic.
Regarding your power requirement of 100w into 8ohms keep to a 35Vac transformer giving +-50Vdc on the PSU rails. Leach will work at this voltage.
The standard Leach is about 120W into 8ohm but due to 2 pairs of robust output devices is capable of driving a moderate 4ohm speaker. Be carefull with the Hi Ft (2sa1943, 2SA1302, 2SA1386) devices they usually need 50% more devices to match the slower devices (MJ15003/4 & MJ21193/4) for reliability particularly when you go to plastic cased (To264) which have a Tjmax of 150degC rather than Tjmax of 200degC.
Probably not, I don't have a lot of literature to go on ! If you would care to elaborate on which is which I'd be grateful 🙂
I have been calculating those figure from the feedback network values.
That would be the plan if my own design(s) turn out to be complete garbage and/or unworkable.
If I was to build the leach it would be with the 4281/4302 as I have them already. I had reasoned 2 pairs of these would be enough for my 100W@8 ohms, and should be safe for say 130W@4 ohms. I have Eltax Liberty 5 speakers which are nominal 6 ohms so this should be fine.
I reasoned this as Elliot's P3A runs about 70W@8 ohms with 42V rails on one pair of these devices. Therefore two should be enough for 100W@8 ohms with 50V rails
jaycee said:
Hi Jaycee,
I posted yesterday but it seems either I am totally wrong or nobody understood my ideas. Since I am a novice I would go for the first option.
Anyway, the link below points to a very interesting Application Note from Intersil explaining calculations around gain and feedback theory. It is a nice reading, specially the first pages. I found it very good because it is not too complicated and I could confirm some ideas with my circuit (both in simulation and practice). I think it explains the differences between feedback factor, gain and etc.
This application note was referenced by someone in a thread here I can't remember.
http://www.intersil.com/data/an/an9415.pdf
Also, very interesting amplifier circuit you have - interesting topology. Please report to us your results as soon as you have them. 🙂
Best regards,
João Pedro
Nice reference João! Thanks.
jpnascim said:
Hi Jaycee,
I posted yesterday but it seems either I am totally wrong or nobody understood my ideas. Since I am a novice I would go for the first option.
Anyway, the link below points to a very interesting Application Note from Intersil explaining calculations around gain and feedback theory. It is a nice reading, specially the first pages. I found it very good because it is not too complicated and I could confirm some ideas with my circuit (both in simulation and practice). I think it explains the differences between feedback factor, gain and etc.
This application note was referenced by someone in a thread here I can't remember.
http://www.intersil.com/data/an/an9415.pdf
Also, very interesting amplifier circuit you have - interesting topology. Please report to us your results as soon as you have them. 🙂
Best regards,
João Pedro
Wonderfull schematic!, if this your first one...my God!..the last one will have no
The last one will have no words to describe.
Those circuits, despite a hell complicated produce a very good treble...i personally feel afraid of triple darlington, because of oscilations.
Maybe some caps in input can be helpfull in the real world assembled schematic.
Olá João Nascim...tudo bem?......Hello JP Nascim, good to see you here, constructing many airplanes?
Congratulations...my God!...first schematic!..... your intelligence quociente may be 180 or up!
Carlos
The last one will have no words to describe.
Those circuits, despite a hell complicated produce a very good treble...i personally feel afraid of triple darlington, because of oscilations.
Maybe some caps in input can be helpfull in the real world assembled schematic.
Olá João Nascim...tudo bem?......Hello JP Nascim, good to see you here, constructing many airplanes?
Congratulations...my God!...first schematic!..... your intelligence quociente may be 180 or up!
Carlos
Hi Jaycee,
just ran an SOAR for you based on the following:-
2 pairs MJ4281, Tc max = 45degC, Vrail = +-50V, Transformer regulation 5% & 350VA, +-45mF smoothing.
It predicts 120w into 8ohms at 60 degrees phase angle
220w into 4ohms at 45 degree phase angle.
Your 6 ohms speaker will not overload your output stage.
Your proposal will drive any severe 8ohms speaker and a moderate 4 ohms speaker.
just ran an SOAR for you based on the following:-
2 pairs MJ4281, Tc max = 45degC, Vrail = +-50V, Transformer regulation 5% & 350VA, +-45mF smoothing.
It predicts 120w into 8ohms at 60 degrees phase angle
220w into 4ohms at 45 degree phase angle.
Your 6 ohms speaker will not overload your output stage.
Your proposal will drive any severe 8ohms speaker and a moderate 4 ohms speaker.
Re: Wonderfull schematic!, if this your first one...my God!..the last one will have no
😀
I will have 15,000uF capacitance per rail, this is for both channels though. I have some BHC capacitors that are 4700uF that I had though of adding to the boards (right next to the output transistors) also. I have a 500VA transformer, again this will run both channels.
One thing I am wondering about is a heatsink. Planning to use a Fischer SK47 in 75MM length (about 0.8K/W according to the datasheet). Is that enough for 2 channels?
😀
I will have 15,000uF capacitance per rail, this is for both channels though. I have some BHC capacitors that are 4700uF that I had though of adding to the boards (right next to the output transistors) also. I have a 500VA transformer, again this will run both channels.
One thing I am wondering about is a heatsink. Planning to use a Fischer SK47 in 75MM length (about 0.8K/W according to the datasheet). Is that enough for 2 channels?
Hi Jaycee,
as might be expected the smaller transformer and less capacitance drops the power prediction, it now comes out at 190w into 4 ohms at 50 degree phase angle. This is for one channel driven which is your worst case for the output stage.
However it depends totally on Tc NEVER exceeding 45degC.
Can you check your heatsink? 0.8K/W seems low for 75mm length of sink unless it is very wide and thick with deep finning.
as might be expected the smaller transformer and less capacitance drops the power prediction, it now comes out at 190w into 4 ohms at 50 degree phase angle. This is for one channel driven which is your worst case for the output stage.
However it depends totally on Tc NEVER exceeding 45degC.
Can you check your heatsink? 0.8K/W seems low for 75mm length of sink unless it is very wide and thick with deep finning.
AndrewT said:
The heatsink is 200mm wide and 40mm deep.
How do you do these tests ? I have added a loudspeaker simulation circuit to my LTSpice schematic (copied from Rod Elliot's page) and have been doing some simple tests with this. It shows a resistance at DC as 6 ohms, so about right for the speakers i would be running.
Hi Jaycee,
look up your heatsink in the manufacturer's literature.
I still think 0.8K/W is low for a 200 by 40 by 75 long heatsink.
Did you follow my note about Tc < 45degC?
Your speaker has a normal DC resistance for a nominal 8ohm unit.
I hope it also has a normal 8ohm treble then your total amp loading will be 8ohms not 6ohms, which the manufacturer averages out over the whole audio band. Your 2 pairs will easily cope with this if you keep them cool (cold heatsink)
look up your heatsink in the manufacturer's literature.
I still think 0.8K/W is low for a 200 by 40 by 75 long heatsink.
Did you follow my note about Tc < 45degC?
Your speaker has a normal DC resistance for a nominal 8ohm unit.
I hope it also has a normal 8ohm treble then your total amp loading will be 8ohms not 6ohms, which the manufacturer averages out over the whole audio band. Your 2 pairs will easily cope with this if you keep them cool (cold heatsink)
That's where I got the figure from 🙂
http://www.fischerelektronik.de/fischer/uploadfischerfcool/Fischer/A.1.5.pdf
Scroll through until you find SK47, theres a graph there. OK it's a little higher than 0.8K/W, nearer 0.9K/W i'd say. The base plate is 10mm thick.
If I were to go for another heatsink, I would have to scrap the case I was planning to use as it is only 80mm high. I've thought about this already as 100mm high heatsinks seem to be easier to get.
http://www.fischerelektronik.de/fischer/uploadfischerfcool/Fischer/A.1.5.pdf
Scroll through until you find SK47, theres a graph there. OK it's a little higher than 0.8K/W, nearer 0.9K/W i'd say. The base plate is 10mm thick.
If I were to go for another heatsink, I would have to scrap the case I was planning to use as it is only 80mm high. I've thought about this already as 100mm high heatsinks seem to be easier to get.
Hi Jaycee,
the 0.8 to 0.9K/W is OK for most music playing situations.
It is exactly the same as I used for a 2 by 110W into 8 ohm stereo amp for years. It never got warm but was driving a pair of 93db sensitivity speakers.
Run yours at low to moderate volume levels checking heatsink temp. Try running at loud volume (party levels) and check regularly again. If it gets warm (above blood temp) in either situation then you need to be carefull. Try measuring Tc.
If you like your music at the warm level then consider increasing the heatsink WIDTH.
the 0.8 to 0.9K/W is OK for most music playing situations.
It is exactly the same as I used for a 2 by 110W into 8 ohm stereo amp for years. It never got warm but was driving a pair of 93db sensitivity speakers.
Run yours at low to moderate volume levels checking heatsink temp. Try running at loud volume (party levels) and check regularly again. If it gets warm (above blood temp) in either situation then you need to be carefull. Try measuring Tc.
If you like your music at the warm level then consider increasing the heatsink WIDTH.
One feature I was planning to add would be a thermal monitor, so if the heatsink got too hot, the power would be removed. I think the widest I could go would be 250mm but that's really pushing it, and i'd still need to change case.
Time to see if I can lay my hands on a 3U 19" case I guess!
Time to see if I can lay my hands on a 3U 19" case I guess!
- Status
- Not open for further replies.