Have you made sims with this amp ?
Have you checked distortionlevels for higher freqs like 10 khz ?
Connecting mosfetgates directly to a lowcurrentvas can be problematic...
With these 2 gates you have ~4nF capacity-load to the vas.
Or is it intended for a subwoofer ?
At these voltagelevels the to220 case from the mosfets can
be too small. You might consider a IRFP-type. (irfp240/9240
these have even lower inputcapacitance)
Your vas looks like ~6ma, that makes 250mw dissipation for
Q4/Q8, that gives ~80° heat. This works, but i wouldn't feel
good with that.
But, it's a "simple" amp, why don't you just try it ?
Everything else is naked theory...
Mike
Have you checked distortionlevels for higher freqs like 10 khz ?
Connecting mosfetgates directly to a lowcurrentvas can be problematic...
With these 2 gates you have ~4nF capacity-load to the vas.
Or is it intended for a subwoofer ?
At these voltagelevels the to220 case from the mosfets can
be too small. You might consider a IRFP-type. (irfp240/9240
these have even lower inputcapacitance)
Your vas looks like ~6ma, that makes 250mw dissipation for
Q4/Q8, that gives ~80° heat. This works, but i wouldn't feel
good with that.
But, it's a "simple" amp, why don't you just try it ?
Everything else is naked theory...
Mike
I would make R14,16 = R13,15 to get equal load on diff pair and speed up Q7
equalize dc bias point by adding a diode (or diode connected Q) in series w/R16 to improve Q8 current mirror operation
equailze Q4,7 collector V with a series R || C or a Zener in Q7 collector to get the same dc power dissapation and better balanced operation in Q4,7 (and diode+ R16)
more current (and pwr without going to power packages) in vas/mirror can easily be had by paralleling extra Qs and emitter Rs with Q4,7,8
C2,5 should be non-polar, C2 should be at least 10X higher value
tail current resistor can be bootstrapped with R from output to get nearly constant tail current; add ( R1 / R20 ) * R6 between output and top of R6
R4,5 drop a lot of V but reducing them costs loop gain; added gain can be had with a R between Q4,7 emitters - once you've equalized their current as in my 1st recommendation
equalize dc bias point by adding a diode (or diode connected Q) in series w/R16 to improve Q8 current mirror operation
equailze Q4,7 collector V with a series R || C or a Zener in Q7 collector to get the same dc power dissapation and better balanced operation in Q4,7 (and diode+ R16)
more current (and pwr without going to power packages) in vas/mirror can easily be had by paralleling extra Qs and emitter Rs with Q4,7,8
C2,5 should be non-polar, C2 should be at least 10X higher value
tail current resistor can be bootstrapped with R from output to get nearly constant tail current; add ( R1 / R20 ) * R6 between output and top of R6
R4,5 drop a lot of V but reducing them costs loop gain; added gain can be had with a R between Q4,7 emitters - once you've equalized their current as in my 1st recommendation
Banned
Joined 2002
Looks exactly like my amp's : O )
http://www.diyaudio.com/forums/showthread.php?s=&postid=594331#post594331
Except now i have a newer version of the board design and schematic..
http://www.diyaudio.com/forums/showthread.php?s=&postid=594331#post594331
Except now i have a newer version of the board design and schematic..
Dear Friends,
Thanks for all your comments. It will be useful at the time to build the final version.
I designed this amp on CircuitMaker 2000. I intend to use it to hear CDs, DVD-Audio and / or to plug my steel guitar on it. My intention is to get Hi Fi sound, but I can build an identical separate unit for subwoofer.
By the way, I have an idea to improve this circuitry that is to install a dynamic variable bias system on this amp, to work like the Super A amplifier, by modulating the base of Q6 with signal rectified output signal.
Now I am at the office, when I get home I will do these changes and submit a new design for you, including the above idea.
PS: jleaman, this circuit really was not based on you project.
Thanks for attention of all.
Thanks for all your comments. It will be useful at the time to build the final version.
I designed this amp on CircuitMaker 2000. I intend to use it to hear CDs, DVD-Audio and / or to plug my steel guitar on it. My intention is to get Hi Fi sound, but I can build an identical separate unit for subwoofer.
By the way, I have an idea to improve this circuitry that is to install a dynamic variable bias system on this amp, to work like the Super A amplifier, by modulating the base of Q6 with signal rectified output signal.
Now I am at the office, when I get home I will do these changes and submit a new design for you, including the above idea.
PS: jleaman, this circuit really was not based on you project.
Thanks for attention of all.
Banned
Joined 2002
Hi,
I don't entirely agree with all the previous comments, so you'll have to make your mind up which you adopt. Mine are:
1) C5 is the wrong way around (the base of Q3 will be slightly negative due to base current flow through R11)
2) If C6 is intended to filter out HF then it needs to be connected between Q3 base and ground and needs to be 300pF to 500pF.
3) Q6 is acting like a battery so it is prudent to place a wide-band capacitor between its collector and emitter, such as a 50uF tantalum.
4) The zobel network, R10/C4, is on the wrong side of L1. The zobel is there to provide a defined output load to the amp at HF. L1 does the opposite as its reactance increases with frequency.
It may be a little unstable in the flesh so check the performance carefully.
Have fun,
BAM
I don't entirely agree with all the previous comments, so you'll have to make your mind up which you adopt. Mine are:
1) C5 is the wrong way around (the base of Q3 will be slightly negative due to base current flow through R11)
2) If C6 is intended to filter out HF then it needs to be connected between Q3 base and ground and needs to be 300pF to 500pF.
3) Q6 is acting like a battery so it is prudent to place a wide-band capacitor between its collector and emitter, such as a 50uF tantalum.
4) The zobel network, R10/C4, is on the wrong side of L1. The zobel is there to provide a defined output load to the amp at HF. L1 does the opposite as its reactance increases with frequency.
It may be a little unstable in the flesh so check the performance carefully.
Have fun,
BAM
I would have to agree with some of the mods proposed. You are 1 resistor and 1 diode away from having a two stage differential amp with current mirror, so why not make one - it will work better.
Q4 and Q7 will form a differential VAS is you make R14 = R13 and connect a resistor between their emitters (this will have to be quite small, on the order of 10 ohms).
Making R16=R15, and connecting a diode in series with R16 (or better still a diode connected 2SA2240) will give you a proper current mirror.
The various minor things mantioned in the thred should be done too (C6 connects to the other side of R12, R10 to the other side of L1, C2 is the wrong way around).
I would strongly advise a constant current source in the tail of the first stage, as the VAS current will also depend on it, and hence the voltage across Q6, which is the bioas voltage. Otherwise, you will get bias variable with supply voltage. not a major issue but cheap to fix.
Three other things I would like to add:
1) This amp has two Cdom, the capacitance betwen C and B of Q4 and Q7 (normally one would find a small cap here, you will probably have to add them in the real amp). But beware: because C of Q7 sees nearly no voltage swing, a capacitor between C and B of Q7 will have little effect, unless you make C of Q7 swing about the same as C of Q4. You can do that by inserting a resistor between C of Q7 and where it now connects to Q8 + R16. The resistor will also take some of the power dissipation of of Q7. The actual value is largely a compromise and you will have to play with it in your simulator. You can use te AC analysis seto to a very high upper limit frequency to see how the gain curve of the amp looks like. Cdom compensation will give you a 6db/oct falling gain slope that should stay th same right down to the point gain = 0dB, this way the amp will be unconditionally stable.
2) A Vbe multiplier used for biasing MOSFETs (Q6 in your schematic) will invariably overcompensate bias voltage with temperature the way you are using it. The standard way of dealing with this is to insert a resistor in the emitter circuit or Q6. the non-standard way, which i have found to work far better, is to use the same element for temperature sensing, as the one making the heat - therefore, use a Vbe multiplier for BJT outputs, Vgs multiplier for MOS outputs. In other words, make Q6 a MOSFET. A small one like 2N7000 or BS170 will work, but I like something in TO220 best (IRF510, 610, 520) - even though it requires mounting hardware, it offers little thermal resistance to the heatsink and reacts fast and accurate. Vgs multipliers are considerably 'worse' than Vbe, because of lower MOSFET gain. Because of this you need a low ESR capacitor between it's D and S lines. I have to disagree with one of the previous posters - tantalums, even though evolved from what they used to be, REALLY should be avoided. They are the most unreliable capacitor. Use a regular or low ESR electrolytic with a non-inductive foil cap on the order of 10-47nF in parallel. unusually, as this kind of cap should not be used in audio, ceramic caps can work good here as well.
Another possibility is to bypass the top resistor (between D and G) with a 10-47n non-inductive ceramic cap.
A Vgs multiplier has the advantage that you can use relatively large resistors for R17, R18, VR1. Also, because you need to compensate for 2 x Vgs (output transistors) + Ibias x (R8 + R9), R17 and R18+VR1 will be nearly equal.
A Vgs multiplier also has the advantage that it tracks the thermal characteristics of the output transistors precisely, which means you can use smaller R8 and R9, i.e. the values you can use are less crytical so you can fine-tune them better.
3) Driving MOSFETs directly from a Vas may be asking for serious trouble, in the form of parasitic oscillations of the output MOSFETs. Your Vas is fairly low current so you may get away with it, but I would advise you to put series resistors in te gates of the output transistors (often even 10 ohms is enough). The added cost is negligible and it avoids a potential problem with little or no degradation of the amps characteristic
Finally, I really don't want to disclose all my secrets
but I may be able to point out the way. Consider this question carefully:
Given that unmatched output transistors result in a less linear output stage, and output stage nonlinearity is typically the largest contributor to distortion, why do people use IRF540/9540, IRFP240/9240 and other 'complementary' MOSFETs when the only three things that make them complementary are the way they look, the fact one is N and the other P channel, and that they both have the same maximum voltage?
Q4 and Q7 will form a differential VAS is you make R14 = R13 and connect a resistor between their emitters (this will have to be quite small, on the order of 10 ohms).
Making R16=R15, and connecting a diode in series with R16 (or better still a diode connected 2SA2240) will give you a proper current mirror.
The various minor things mantioned in the thred should be done too (C6 connects to the other side of R12, R10 to the other side of L1, C2 is the wrong way around).
I would strongly advise a constant current source in the tail of the first stage, as the VAS current will also depend on it, and hence the voltage across Q6, which is the bioas voltage. Otherwise, you will get bias variable with supply voltage. not a major issue but cheap to fix.
Three other things I would like to add:
1) This amp has two Cdom, the capacitance betwen C and B of Q4 and Q7 (normally one would find a small cap here, you will probably have to add them in the real amp). But beware: because C of Q7 sees nearly no voltage swing, a capacitor between C and B of Q7 will have little effect, unless you make C of Q7 swing about the same as C of Q4. You can do that by inserting a resistor between C of Q7 and where it now connects to Q8 + R16. The resistor will also take some of the power dissipation of of Q7. The actual value is largely a compromise and you will have to play with it in your simulator. You can use te AC analysis seto to a very high upper limit frequency to see how the gain curve of the amp looks like. Cdom compensation will give you a 6db/oct falling gain slope that should stay th same right down to the point gain = 0dB, this way the amp will be unconditionally stable.
2) A Vbe multiplier used for biasing MOSFETs (Q6 in your schematic) will invariably overcompensate bias voltage with temperature the way you are using it. The standard way of dealing with this is to insert a resistor in the emitter circuit or Q6. the non-standard way, which i have found to work far better, is to use the same element for temperature sensing, as the one making the heat - therefore, use a Vbe multiplier for BJT outputs, Vgs multiplier for MOS outputs. In other words, make Q6 a MOSFET. A small one like 2N7000 or BS170 will work, but I like something in TO220 best (IRF510, 610, 520) - even though it requires mounting hardware, it offers little thermal resistance to the heatsink and reacts fast and accurate. Vgs multipliers are considerably 'worse' than Vbe, because of lower MOSFET gain. Because of this you need a low ESR capacitor between it's D and S lines. I have to disagree with one of the previous posters - tantalums, even though evolved from what they used to be, REALLY should be avoided. They are the most unreliable capacitor. Use a regular or low ESR electrolytic with a non-inductive foil cap on the order of 10-47nF in parallel. unusually, as this kind of cap should not be used in audio, ceramic caps can work good here as well.
Another possibility is to bypass the top resistor (between D and G) with a 10-47n non-inductive ceramic cap.
A Vgs multiplier has the advantage that you can use relatively large resistors for R17, R18, VR1. Also, because you need to compensate for 2 x Vgs (output transistors) + Ibias x (R8 + R9), R17 and R18+VR1 will be nearly equal.
A Vgs multiplier also has the advantage that it tracks the thermal characteristics of the output transistors precisely, which means you can use smaller R8 and R9, i.e. the values you can use are less crytical so you can fine-tune them better.
3) Driving MOSFETs directly from a Vas may be asking for serious trouble, in the form of parasitic oscillations of the output MOSFETs. Your Vas is fairly low current so you may get away with it, but I would advise you to put series resistors in te gates of the output transistors (often even 10 ohms is enough). The added cost is negligible and it avoids a potential problem with little or no degradation of the amps characteristic
Finally, I really don't want to disclose all my secrets
but I may be able to point out the way. Consider this question carefully:Given that unmatched output transistors result in a less linear output stage, and output stage nonlinearity is typically the largest contributor to distortion, why do people use IRF540/9540, IRFP240/9240 and other 'complementary' MOSFETs when the only three things that make them complementary are the way they look, the fact one is N and the other P channel, and that they both have the same maximum voltage?
What is the reliability issue with tantalums? I have never had any problems with them when used within spec. Tantalums fail if they are reverse-biased alright. In terms of capacitance they have the benefit of being very small and having very good HF performance, sometimes better than some ceramics. I quote from National Semi's LM317 datasheet:
I've had to replace lots of solid tantalums in '70s vintage test gear, they fail short. They also can fail short with excessive dV/dt but this isn't an issue with audio frequency applications. I use them only when low leakage current is needed and only for DC bypassing. Usually Al electrolytic is OK for DC bypassing, if you need extended frequency response, parallel with a low-inductance film or ceramic. Certainly film is preferable to electrolytics for AC coupling but for low impedance at 20Hz the choices are limited.
traderbam said:
Nuvistor actually answered this for me, and keep in mind that caps in test equipment are usually of high quality.
Until the advent of reasonably priced higher capacitance multilayer ceramics, tantalums were the only way to go for bulk bypasses on high speed digital stuff, and I used tons of them by necessity. I have had some explode on first application of voltage even though they were brand new and of course properly polarized (35A power supplies do not tolerate shorted caps well
). Some tantalums are partly self-healing which can actually be the origin of untold anguish, with equipment that occasionally fails. More than any other component, with tantalums you never know when they are going to turn from a decent cap into a short circuit.Of course, the better the quality, the less chance of that (but I've had even the best fail), but the price, which is high to begin with, tends to end up on the order of several low ESR electrolytics bypassed with ceramic caps, which give better results anyway.
Tantalums have one characteristic which is rare in high capacitances, low leakage. Unfortunately, the leakage current that there is, tends to be very noisy.
Also, while they have negligible series inductance (ecxept the inevitable leads, if present), tantalums do tend to have highish series resistance, especially if they are very small.
darkfenriz said:
Lower on resistance will hardly help in a linear application - especially if your case has high thermal resistance and the temperature rise increases the said resistance (people who design switchmode PSUs know that MOSFETs can exhibit a kind of thermal runaway effect because of this). What counts is the ability to shed heat (IRFP240/9240 are far better at that), and good gain, the more linera the better. IRFP240/9240 are much better in both respects, but also more expensive. And, granted, they may be a bit of overkill for 35W.
nuvistor said:
I can only agree. Also, there are ceramics that are almost ideal, but unfortunately, the largest one you can get of that type tends to be about 2.2nF
The typical ceramic cap tends to have large changes in capacitance with respect to the voltage applied, which is why you never want to be caught using those in filters or coupling caps that will have any appreciable voltage swing across them - the distortion gets quite high.
All of the failed tantalums I have seen were dipped radial lead, these are the cheapest type but are used in a lot of DVMs, signal generators, etc. I have not seen any hermetically sealed tantalums fail, but these are not as common. There are mil-spec tantalums with established reliability levels. Personally I just prefer to avoid components that fail short, unfortunately all semiconductors fail this way (and tubes usually do not).
Banned
Joined 2002
binary said:
It is a very simple amp.
Maybe too simple to really work, without some adjustments.
Well, it is only 35 HIFI WATTS output ......
Have you come around to build it, by now?
I like that you are using two different supplies, 40 and 33 Volt for output.
This also will make better use of output power.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.