Yes, this makes sense. How do I know the base current? Do I get this by dividing the VAS CCS current by the hfe of the Vbias transistor? (I guess so.)AndrewT said:
Also, your suggestion gives me an upper limit for the resistors. Is there anything I should keep in mind about the lower limit?
This was totally an embarassing mistake of mine.... Rod says clearly that one must set max resistance to begin with. I had confused it. Sorry.
Okay, got it, thanks.
And I think I need to re-word an earlier question. What's the relationship between the time constant of the PSU and the time constant of the NFB's RC?
What do you think of the PCB? Or don't you think much of it at all?
Hi Tc,
I have seen and used a recommendation to keep at least half an octave between each pair.
I use input=80 to 100mS, NFB=120 to 140mS, PSU=160 to 200mS (ClassAB) and gives about -1db @ 3 to 4Hz. These give fairly large capacitors compared to most published designs, particularly so if you aim for low impedances.
Assume the multiplier produces 0.7Vbe for the minimum output and similarly 0.6Vbe for the maximum output, now warm up your calculator or spreadsheet.
input RC<NFB RC<PSU RC.
I have seen and used a recommendation to keep at least half an octave between each pair.
I use input=80 to 100mS, NFB=120 to 140mS, PSU=160 to 200mS (ClassAB) and gives about -1db @ 3 to 4Hz. These give fairly large capacitors compared to most published designs, particularly so if you aim for low impedances.
find the minimum and maximum voltage across the output stage bases. An EF Darlington will be min =4*0.65V and max = 4*0.65V + Vre. A triplet needs 6times. CFP is different.
Assume the multiplier produces 0.7Vbe for the minimum output and similarly 0.6Vbe for the maximum output, now warm up your calculator or spreadsheet.
The PCB: v3
Made some changes to the schematic and PCB:
I've added a first-order RC low-pass filter at the signal input, and reduced the power ground points by one. Now there are just two power ground points, one on either side of the board, plus the signal ground.
I've also added a small ceramic disk capacitor on each supply rail (C14 and C15) in addition to the old electrolytic ones, near the input section of the circuit. I was told that this additional cap can help keep RF out of the circuit and help stability of MOSFET amps.
Made more space for the L1 inductor. I have an inch of length now, and can probably do more than 25 turns of 14SWG copper.
The Q3-Q4 and Q1-Q2 pairs of transistors are now close together and amenable to being tied together to keep close thermal coupling. Also, the two transistors which make up the VAS Darlington pair (Q5 and Q6) are now thermally coupled with a common heatsink. Apparently, as per some passing mention in Randy Slone's book, this is a good idea.
Set different values for the gate resistors for the P-channel and N-channel OPS devices. Earlier, they were all equal.
Made some changes to the schematic and PCB:
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
I've added a first-order RC low-pass filter at the signal input, and reduced the power ground points by one. Now there are just two power ground points, one on either side of the board, plus the signal ground.
I've also added a small ceramic disk capacitor on each supply rail (C14 and C15) in addition to the old electrolytic ones, near the input section of the circuit. I was told that this additional cap can help keep RF out of the circuit and help stability of MOSFET amps.
Made more space for the L1 inductor. I have an inch of length now, and can probably do more than 25 turns of 14SWG copper.
The Q3-Q4 and Q1-Q2 pairs of transistors are now close together and amenable to being tied together to keep close thermal coupling. Also, the two transistors which make up the VAS Darlington pair (Q5 and Q6) are now thermally coupled with a common heatsink. Apparently, as per some passing mention in Randy Slone's book, this is a good idea.
Set different values for the gate resistors for the P-channel and N-channel OPS devices. Earlier, they were all equal.
Hi,
669 is a wonderful (and relatively cheap) transistor.
I am using them to drive both the 140W Leach and 120W Krell.
BUT not for q5. Use 2sc3423 or similar
Any comment on C8 being electrolytic or plastic film? I have a view!
R34 could be raised to match R12 or even higher.
C7*R12<<1.4*C1*R4.
But if the second DC blocking cap in the source unit is also 10uF then the combined RC is about right. This will result in an effective 5uF giving -1db @ about 7Hz.
669 is a wonderful (and relatively cheap) transistor.
I am using them to drive both the 140W Leach and 120W Krell.
BUT not for q5. Use 2sc3423 or similar
Any comment on C8 being electrolytic or plastic film? I have a view!
R34 could be raised to match R12 or even higher.
C7*R12<<1.4*C1*R4.
But if the second DC blocking cap in the source unit is also 10uF then the combined RC is about right. This will result in an effective 5uF giving -1db @ about 7Hz.
I don't know where you get those transistors.... I looked up findchips.com and they turned up a complete blank in all the resellers listed with them.AndrewT said:
Should I aim for something like a polyester? My PCB currently doesn't have space for something like that.
Okay. Will note it.
Which is the second DC blocking cap? I can't see any other cap of 10uF. Do you mean the (presumed) output cap at the output of the CD player or preamp or whatever which is upstream of this power amp?
How do I calculate the HF F3 for my C1+C2+R34+R4 combo?
Hi,
yes, DC blocking cap in the immediately preceeding buffered source component (pre-amp or CD player etc.).
Ignoring the small value of R34
then F-3db=1/2/Pi/R/C
R=R4
C=C1 & C(pre-amp), C=C1*Cp/[C1+Cp]
10uF & 10uF =5uF
10uF & 4u7F=3u2F
10uF & 2u2F=1u8F
Notice how the bass cut-off gets worse as the preamp DC blocker moves to a lower value.
R4=50k, C1=10uF, Cpre=2u2F gives a good compromise.
or swap C1 & Cpre, same result. Buy a good 10uF pp and fit it in the pre, then all your poweramps need just 2u2F pp.
HF F-3db uses R34 & C2
150r & 100p gives -1db @ 5MHz (a bit high don't you think?).
330r & 1nF gives -1db @ 240kHz. (even this is a bit high, but if Rs>1k0 then the story is completely different).
2sc3423 is the complement to 2sa1360. or try complements to 1209, 1210, 1370, 1371, 1380, 1403, 1405, 1406, 1407, 1476, 1477, 1478, 1538, 1539, 1540, 1697, 1707, 1777, 1875, most of these are more expensive. But you should be able to find a number of low Cob alternatives that are suitably voltage rated.
yes, DC blocking cap in the immediately preceeding buffered source component (pre-amp or CD player etc.).
Ignoring the small value of R34
then F-3db=1/2/Pi/R/C
R=R4
C=C1 & C(pre-amp), C=C1*Cp/[C1+Cp]
10uF & 10uF =5uF
10uF & 4u7F=3u2F
10uF & 2u2F=1u8F
Notice how the bass cut-off gets worse as the preamp DC blocker moves to a lower value.
R4=50k, C1=10uF, Cpre=2u2F gives a good compromise.
or swap C1 & Cpre, same result. Buy a good 10uF pp and fit it in the pre, then all your poweramps need just 2u2F pp.
HF F-3db uses R34 & C2
150r & 100p gives -1db @ 5MHz (a bit high don't you think?).
330r & 1nF gives -1db @ 240kHz. (even this is a bit high, but if Rs>1k0 then the story is completely different).
2sc3423 is the complement to 2sa1360. or try complements to 1209, 1210, 1370, 1371, 1380, 1403, 1405, 1406, 1407, 1476, 1477, 1478, 1538, 1539, 1540, 1697, 1707, 1777, 1875, most of these are more expensive. But you should be able to find a number of low Cob alternatives that are suitably voltage rated.
Understood your calculations, thanks again.AndrewT said:
So, basically if I need to be preamp-independent in my LF cut-off, I'll need to spend richly on large C1.
I think I'll switch R34 and C2 to approximately the values you've mentioned, between 300 and 350 for R34 and about 1nF for C2. I agree that 5MHz is very high.... this high cutoff was considered adequate by Randy Slone in his book to protect MOSFET OPS against oscillations, but I thought that there's no audible harm in playing safe and bringing the cutoff down by a decade or so.
Will a ceramic disc do for C2, or should I opt for more expensive caps?
Okay, will search. Where do you usually find these transistors being sold? I used to look up Digikey and one or two others, now I'll look elsewhere.
Also, sorry to admit this, but I still don't know what Cob is. I didn't find any Cob mentioned in the one or two transistor specs I studied.
Found 2SA1360 on MCM (www.mcminone.com). It's very inexpensive, at about forty cents each. 2SC3423 is more expensive, at a shade over a dollar each. Both are Toshiba devices. 2SA1370 and 1371 are also available, at prices broadly in the same range as 1360. I've stopped searching for the remaining parts for now... it seems a sufficient portion of your list is available from here.AndrewT said:
Strange how these are so hard to find through the findchips.com resellers.
Can you please explain what will be the audible or stability effect if I use these transistors instead of the 649/669 devices? And why is the swap needed only for Q5?
Hi,
Cob is the capacitance seen at the base going to the collector of the transistor. It varies with Vcb quite markedly. Lower at high volts and much higher at low Vcb.
The LTP has a resistor or active load in parallel with the load that the VAS or it's EF provides.
The LTP works best when the load on it's collector is constant and at a value that balances the LTP.
Any extra current that escapes into the VAS robs the load and unbalaances the LTP.
The VAS reflects the downstream impedance back to it's base and the LTP sees this reflected load in // to the collector load.
However, in addition there is a third parallel load on the LTP. The Cob is zero load at DC, but since it is capacitance it's impedance(reactance) decreases with increasing frequency. At high audio frequencies the capacitance is sufficient to cause LTP unbalance. But it gets worse. At the frequencies say one decade above audio (20kHz to 200kHz) the load falls by a further factor of ten and this causes severe loading that may be heard as intermodulation (and I guess other effects). It also causes slew rate decrease since the capacitance charges at the maximum current the LTP can provide and also causes slew asymetry.
But there is more. The capacitance varies as the voltage across the VAS changes.
This capacitance change with voltage is eliminated by cascoding the VAS. But it's advantage is then almost thrown away by adding the Miller comp cap.
Adding an EF before the VAS reduces the reflected load and reduces the variable capacitance, but this is only an advantage if the Cob is deliberately selected to be low, if it were high we have that reducing impedance with increasing frequency. 669 is about 27pF and 3423 is about 2.5pF. anything below 5pF is probably OK.
Cob is the capacitance seen at the base going to the collector of the transistor. It varies with Vcb quite markedly. Lower at high volts and much higher at low Vcb.
The LTP has a resistor or active load in parallel with the load that the VAS or it's EF provides.
The LTP works best when the load on it's collector is constant and at a value that balances the LTP.
Any extra current that escapes into the VAS robs the load and unbalaances the LTP.
The VAS reflects the downstream impedance back to it's base and the LTP sees this reflected load in // to the collector load.
However, in addition there is a third parallel load on the LTP. The Cob is zero load at DC, but since it is capacitance it's impedance(reactance) decreases with increasing frequency. At high audio frequencies the capacitance is sufficient to cause LTP unbalance. But it gets worse. At the frequencies say one decade above audio (20kHz to 200kHz) the load falls by a further factor of ten and this causes severe loading that may be heard as intermodulation (and I guess other effects). It also causes slew rate decrease since the capacitance charges at the maximum current the LTP can provide and also causes slew asymetry.
But there is more. The capacitance varies as the voltage across the VAS changes.
This capacitance change with voltage is eliminated by cascoding the VAS. But it's advantage is then almost thrown away by adding the Miller comp cap.
Adding an EF before the VAS reduces the reflected load and reduces the variable capacitance, but this is only an advantage if the Cob is deliberately selected to be low, if it were high we have that reducing impedance with increasing frequency. 669 is about 27pF and 3423 is about 2.5pF. anything below 5pF is probably OK.
Sorry if I'm responding with half-unclear concepts, but your description reminded me of the effect of the compensation capacitor on the input stage loading. Doesn't the presence of the CC have a similar effect on the input stage, with its impedance reducing with frequency, etc? Will the Cob of the transistor dominate the load that the input stage sees, or will it be dominated by the CC?
Am I at all on the same page as you? If not, please say so, because I have lots of grey areas about my understanding of amp internals. I ask questions with the aim of clearing up those greys, not with the intention to find fault in anyone else's explanations.
Thanks for the note.
Am I at all on the same page as you? If not, please say so, because I have lots of grey areas about my understanding of amp internals. I ask questions with the aim of clearing up those greys, not with the intention to find fault in anyone else's explanations.
Thanks for the note.
You mean I cut the negative rail between the bottom of R22 and the bottom of R26, and insert this resistor?darkfenriz said:
On this subject, I was wondering about the entire idea of doing additional decoupling of both rails between the OPS and the VAS. Is this a good idea? So few amps seem to do it.
If I imagine that the input power supply flows into the circuit near the OPS and then flows from right to left towards the VAS and then towards the input, then the maximum surges of current are in the OPS, and the VAS and input stages get the side-effects of this. Is it a good idea to cut each supply rail between the OPS and VAS, and put in an RC there?
Taking the idea one step forward, since the OPS will occasionally sink huge amounts of current, maybe there will be momentary current flow depletion for the VAS and input stages. So, to prevent this, maybe we could fit diodes (1N4001?) before the RC filters. That way, the supply current would flow in and charge the cap, but would never flow out.
What do you think? And if yes, then what values of RC make sense? We can estimate the current draw of the VAS and input stages by looking at the CCS resistors I guess.
Hi,
a diode between r22 & r26 and similarly between r23 & r14 keep the volt amp stage running from the decoupling caps when the mains rails sag. You may need to increase the caps size slightly.
You can also fit bypass caps at the output devices.
Q.
When output device bypasses are added, do they return to the ground side of the load?
What if the ground side of the load does not return to the PCB?
a diode between r22 & r26 and similarly between r23 & r14 keep the volt amp stage running from the decoupling caps when the mains rails sag. You may need to increase the caps size slightly.
You can also fit bypass caps at the output devices.
Q.
When output device bypasses are added, do they return to the ground side of the load?
What if the ground side of the load does not return to the PCB?
Great. I'll see how I can fit them into the PCB. Will 1N4001-type things do?AndrewT said:
Didn't understand this. I already have bypass caps C11 and C12 near the entry points for the supply lines, on the PCB. What are bypass caps specifically near the output devices? Do you mean I should add one more pair of caps, as close to the OPS devices as possible? If I do, then what kind of cap values will be needed? Won't they need to be large and super-low-ESR to make any difference at all at those high currents?
What does this question mean? I'm confused. And in case it helps, my ground side of the load does not return to the PCB... it returns straight to the star-ground point.
Managed to add OPS drain resistors
This is the new schematic. No surprises here, other than the addition of four power resistors:
I've also removed the power resistor which was in parallel with the inductor in the output rail. This does not mean I'll not put the resistor... it just means I've decided not to show it on the PCB. When I build the amp, I'll insert a 5W 10 Ohm resistor inside the inductor and solder both together in parallel.
This is the new PCB:
You can see the four power resistors. I've added one vertically (R34). This one will almost certainly have to be fitted on the underside of the PCB. (This of course implies that I'll need that much empty space below the PCB in my chassis.) The other three may either be fitted on the top surface, raised by a quarter inch, or on the bottom, as is found convenient. The inductor (marked "L1+R") will be fitted on top, raised by half an inch or even more above the PCB.
What do you think?
This is the new schematic. No surprises here, other than the addition of four power resistors:
An externally hosted image should be here but it was not working when we last tested it.
I've also removed the power resistor which was in parallel with the inductor in the output rail. This does not mean I'll not put the resistor... it just means I've decided not to show it on the PCB. When I build the amp, I'll insert a 5W 10 Ohm resistor inside the inductor and solder both together in parallel.
This is the new PCB:
An externally hosted image should be here but it was not working when we last tested it.
You can see the four power resistors. I've added one vertically (R34). This one will almost certainly have to be fitted on the underside of the PCB. (This of course implies that I'll need that much empty space below the PCB in my chassis.) The other three may either be fitted on the top surface, raised by a quarter inch, or on the bottom, as is found convenient. The inductor (marked "L1+R") will be fitted on top, raised by half an inch or even more above the PCB.
What do you think?
Hi,
move the whole Thiel network to the back of the speaker terminals. i.e. R//L & C+R, saving the space of all four components.
That Q is really for me to anybody that knows the solution. Ignore it till some expert tells us how to do it.
Bypass caps provide the very short term current into or out of a device that changes it's current. It helps attenuate the spike or notch that otherwise gets onto the supply rails as devices modulate their consumption. A local bypass works best when it is mounted right beside the device that causes the current change i.e. the source pin of the FETs and the ground return of the speaker. The closest we can get to the speaker return is the power ground on the PCB.
C11 & C12 are far too far from the FETs. Use very small, very low esr, very low inductance, very short lead=ceramic.
1N400X is OK.
Add 6V to 8V zeners to protect the FETs. // to R23 & R26.
Add 1N400x from output rail to both supply rails.
Are you making the PCB yourself? then move the mounting holes outside the PCB size limits. Similarly the Fuse holders could move to outside or put them with the smoothing caps.
move the whole Thiel network to the back of the speaker terminals. i.e. R//L & C+R, saving the space of all four components.
That Q is really for me to anybody that knows the solution. Ignore it till some expert tells us how to do it.
Bypass caps provide the very short term current into or out of a device that changes it's current. It helps attenuate the spike or notch that otherwise gets onto the supply rails as devices modulate their consumption. A local bypass works best when it is mounted right beside the device that causes the current change i.e. the source pin of the FETs and the ground return of the speaker. The closest we can get to the speaker return is the power ground on the PCB.
C11 & C12 are far too far from the FETs. Use very small, very low esr, very low inductance, very short lead=ceramic.
1N400X is OK.
Add 6V to 8V zeners to protect the FETs. // to R23 & R26.
Add 1N400x from output rail to both supply rails.
Are you making the PCB yourself? then move the mounting holes outside the PCB size limits. Similarly the Fuse holders could move to outside or put them with the smoothing caps.
Feeling lazy to do this change now.AndrewT said:
If I run out of space, I'll probably do it. In fact, I can make a small general-purpose PCB for this and use it for all future power amps.How do I do this? Layout-wise, I mean... it seems so tight now. Let me see what I can do. I guess in order to do this, I'll have to move the C11 and C12 "after" the fuses, not before, right?
How does this work? I am somewhat confused. Does this limit the voltage drop across R23 and R26? How will this protect the OPS FETs?
Is this called "clamping diodes"? I'll try to find space for them.
Yes, I'll get the PCBs made myself, but I won't make them with my own hands. Therefore, I'll have to submit the Gerber files to someone "as is". I don't know whether they can do such editing for me. I'll check.
And about the fuse holders, I was initially thinking of putting them with the smoothing caps, but then I read that L-MOSFETs can actually be protected by fuses, i.e. the fuse will blow before the L-MOSFET will be damaged. Therefore, for just the L-MOSFET designs, I thought it might be a good idea to have one fuse per rail. For BJT designs (I'm also working on the Slone "Fig 11.4" design), I can move the fuses to the smoothing caps, I guess.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.