Golmund clone
From the measurements I made with heat sink 320mm x 100mm x 40mm
I counted 52 C degrees temperature with bias current 290 ma over to one rail +or - .....95-100 ma per mosfet bias current which remains stable.
Ιf the amplifier begins to play on a high level the temperature rise to 60-65 C degrees temperature and continues to rise.
Τhe pictures I've seen from the amplifier Μimesis 9.2 ι make the conclusion that the heat sink is about 320mmx 200mm x 40mm
What is the maximum safe temperature on the heatsink where it can reach the amplifier?
Would it be preferable to be fastened fans and small heatsinks on tunnel model ?
From the measurements I made with heat sink 320mm x 100mm x 40mm
I counted 52 C degrees temperature with bias current 290 ma over to one rail +or - .....95-100 ma per mosfet bias current which remains stable.
Ιf the amplifier begins to play on a high level the temperature rise to 60-65 C degrees temperature and continues to rise.
Τhe pictures I've seen from the amplifier Μimesis 9.2 ι make the conclusion that the heat sink is about 320mmx 200mm x 40mm
What is the maximum safe temperature on the heatsink where it can reach the amplifier?
Would it be preferable to be fastened fans and small heatsinks on tunnel model ?
Attachments
Last edited:
Most (cost) efficient is spreading the temperature rise evenly between the output devices and the heatsink.
Rjs for a 100W Hitachi MOSFET with regular mica insulator is about 1.6 C/W.
For 6 output devices on 1 heatsink, the thermal resistance number of the sink would be 1.6C/W divided by 6, makes ~0.27 C/W.
(the 200mm length Fischer SK56 of the 9.2 does ~0.30C/W)
Thermal resistance of 100mm SK56 is about 0.40C/W.
65C temperature of the heatsink at 25C ambient makes a temperature rise of 40 degrees.
40C temperature rise for 0.40C/W thermal resistance means 100W average dissipation.
100W divided by 6 output devices is ~17W dissipation per device.
For an Rjs of 1.6C/W, temperature rise between the die junction and the heatsink is about 27C.
At 65C heatsink temperature, die temperature would be 92C.
No big deal for the output devices, provided the temperature does not go up another 10 points.
+65C heatsink temperature is a bit unpractical, investing too little in heatsinks is both bad engineering and bad economics.
(a 200mm SK56 does about €52 or $75 each, makes $12.5 per output device. The higher the purchasing cost of a TO3 Hitachi, the more sensible it becomes to buy even larger heatsinks)
Last edited:
Maximum operating junction temperature is 150°C for the devices however according to profusion application note: A good rule of thumb when using these MOSFETs is to set a maximum operational case temperature of 110°C.
The heatsink in the picture will pass this criteria, C/W = 0.3
I have mention before that the heatsink will be very hot on the original amp.
The heatsink in the picture will pass this criteria, C/W = 0.3
I have mention before that the heatsink will be very hot on the original amp.
Thank you Jacco Vermeulen
Εach time we need clarification on any part of the amplifier you're here.
This small heat sink is part of the amplifier Crescendo Millennium I had to construct prior years and test the existing heatsink how cools well before buying final heatsink.
Ι will put greater heatsink 320mm x 200 x 40 mm .
Τhe question is how well it fits the mosfet Exicon ECF10N20 ECF10P20 with mosfet Hitachi .
From Exicon we have so poor datasheet about ECF10 Ν-P-20
Τhe transistors fit as I could before placing.
But if not raise the resistance of the gate mosfet N channel about 680 Ω would have oscillation .
Βut the rise time is the same now Ν and p channel ?
how can we know this ?
Regards
Εach time we need clarification on any part of the amplifier you're here.
This small heat sink is part of the amplifier Crescendo Millennium I had to construct prior years and test the existing heatsink how cools well before buying final heatsink.
Ι will put greater heatsink 320mm x 200 x 40 mm .
Τhe question is how well it fits the mosfet Exicon ECF10N20 ECF10P20 with mosfet Hitachi .
From Exicon we have so poor datasheet about ECF10 Ν-P-20
Τhe transistors fit as I could before placing.
But if not raise the resistance of the gate mosfet N channel about 680 Ω would have oscillation .
Βut the rise time is the same now Ν and p channel ?
how can we know this ?
Regards
Do you know this?
Hi,
LOL,I like that color one ,I bet it work's just like my old b&W one ,lol
Golmund clone- DETERMINING GATE RESISTOR VALUES
Reading this article I decided to experiment with the generator and oscilloscope determining gate resistor values.
The result is the following :
100 Ω resistance for the P channel mosfet
and 220 Ω resistance for the N channel mosfet
The sound is completely changed, incredible sound, the sound was colorless before.
1KHZ CLIP SINE
1KHZ-10KHZ-40KHZ-100KHZ
1KHZ-10KHZ-20KHZ ----100nf
Reading this article I decided to experiment with the generator and oscilloscope determining gate resistor values.
The result is the following :
100 Ω resistance for the P channel mosfet
and 220 Ω resistance for the N channel mosfet
The sound is completely changed, incredible sound, the sound was colorless before.
1KHZ CLIP SINE
1KHZ-10KHZ-40KHZ-100KHZ
1KHZ-10KHZ-20KHZ ----100nf
Attachments
Last edited:
Nikosokey
Your recent oservations and experiments with gate resistors are interesting.
Have I understood you correctly when you now have "landed" on the gate resistor values of 100 and 220 ohms, and that you now are experiencing these values offering you a more "musically" amplifier (not "colorless" as before)?
At such low values, you do not have any oscillation problems, if I have understood you right (In a previous post,1785, you mention that you used a gate value of 470/680 ohms to avoid oscillation)?
I use gate resistors of 470 and 680 ohms. My FETS are Exicon 10N16 and 10P16. I do not have an easy access to gauges, but I feel tempted to follow your example.
But will the results for me, with slightly different FETs, be the same as Nikos tell us.
Does anyone have views on that question?
Eivind Stillingen
Your recent oservations and experiments with gate resistors are interesting.
Have I understood you correctly when you now have "landed" on the gate resistor values of 100 and 220 ohms, and that you now are experiencing these values offering you a more "musically" amplifier (not "colorless" as before)?
At such low values, you do not have any oscillation problems, if I have understood you right (In a previous post,1785, you mention that you used a gate value of 470/680 ohms to avoid oscillation)?
I use gate resistors of 470 and 680 ohms. My FETS are Exicon 10N16 and 10P16. I do not have an easy access to gauges, but I feel tempted to follow your example.
But will the results for me, with slightly different FETs, be the same as Nikos tell us.
Does anyone have views on that question?
Eivind Stillingen
Hi,
There is an important hidden variable not seen.
As the output from the VAS is around 10K, the output impedance from the BSS71/74 followers is somewhere around 80 Ohm. So, this is more or less equivalent to preceding each mosfet with another resistor of around 500 Ohm...
Later GM's with individual drivers per mosfet pair use additional resistors in the Emitters to get similar results, once all is calculated through.
So the commonly recommended 680/470 Ohm gate resistors are likely not ideal for the GM Amp.
Also, looking at the frequency compensation of the Amp when using BSS71/74 (which have a smashingly low Cob of only 3.5pF each) seems almost entirely applied to the Mosfet Gate, so I am unsurprised the sound changed so much.
Ciao T
There is an important hidden variable not seen.
As the output from the VAS is around 10K, the output impedance from the BSS71/74 followers is somewhere around 80 Ohm. So, this is more or less equivalent to preceding each mosfet with another resistor of around 500 Ohm...
Later GM's with individual drivers per mosfet pair use additional resistors in the Emitters to get similar results, once all is calculated through.
So the commonly recommended 680/470 Ohm gate resistors are likely not ideal for the GM Amp.
Also, looking at the frequency compensation of the Amp when using BSS71/74 (which have a smashingly low Cob of only 3.5pF each) seems almost entirely applied to the Mosfet Gate, so I am unsurprised the sound changed so much.
Ciao T
The last two contributions from ThorstenL and Nico Ras about the values on the gate resistors are interesting and for me new knowledge about FETs.
In connection with the building of my two Goldmund clones, I have visited quite a few threads here on diyAudio (and on others sites) to find out what to have in mind to make a FET amplifier stable.
There seems to be broad agreement about this: The different speed between the N FET an P FET must be compensated by using higher value of gate resistance on the N FET versus P FET.
How many FETs connected in parallel will also affect the size: The "right" value for three pairs must be higher than if two pairs are used.
ThortsenL wrote:
As the output from the VAS is around 10K, the output impedance from the BSS71/74 followers is somewhere around 80 Ohm. So, this is more or less equivalent to preceding each mosfet with another resistor of around 500 Ohm.
And:
Also, looking at the frequency compensation of the Amp when using BSS71/74 (which have a smashingly low Cob of only 3.5pF each) seems almost entirely applied to the Mosfet Gate, so I am unsurprised the sound changed so much.
Nico Ras wrote:
Gate resistors should be matched to the gate capacitance.
I asume that the drivers' (BSS71/74) capasistance influences at the capasistanse on the FETs.
I have no background to calculate the total capasistance we get when a pair of BSS71/74 "looks into" a pair of Exicon 10N16/10P16, and from this calculations determine the value on the gate resistors.
Have Nikos,using 100 ohm and 220 ohms, found an optimal value??. He says that he subjectively experience this as a more musical solution than to use the 470/680 ohm (witch I use). It should mention that he uses a "larger" version of Exicon FETs.
I have studied one aplication note from Exicon, but unfortunately they provided no values for the gate resistors.
As mention before I have built two Goldmund clones, one with BSS71/74 as drivers and one with MJE 340/350 as drivers.
I am looking forward to hear from some of our experts out there, to give us some good advice on the choice of values on the gate resitors with the drivers and FETs described here.
Eivind Stillingen
In connection with the building of my two Goldmund clones, I have visited quite a few threads here on diyAudio (and on others sites) to find out what to have in mind to make a FET amplifier stable.
There seems to be broad agreement about this: The different speed between the N FET an P FET must be compensated by using higher value of gate resistance on the N FET versus P FET.
How many FETs connected in parallel will also affect the size: The "right" value for three pairs must be higher than if two pairs are used.
ThortsenL wrote:
As the output from the VAS is around 10K, the output impedance from the BSS71/74 followers is somewhere around 80 Ohm. So, this is more or less equivalent to preceding each mosfet with another resistor of around 500 Ohm.
And:
Also, looking at the frequency compensation of the Amp when using BSS71/74 (which have a smashingly low Cob of only 3.5pF each) seems almost entirely applied to the Mosfet Gate, so I am unsurprised the sound changed so much.
Nico Ras wrote:
Gate resistors should be matched to the gate capacitance.
I asume that the drivers' (BSS71/74) capasistance influences at the capasistanse on the FETs.
I have no background to calculate the total capasistance we get when a pair of BSS71/74 "looks into" a pair of Exicon 10N16/10P16, and from this calculations determine the value on the gate resistors.
Have Nikos,using 100 ohm and 220 ohms, found an optimal value??. He says that he subjectively experience this as a more musical solution than to use the 470/680 ohm (witch I use). It should mention that he uses a "larger" version of Exicon FETs.
I have studied one aplication note from Exicon, but unfortunately they provided no values for the gate resistors.
As mention before I have built two Goldmund clones, one with BSS71/74 as drivers and one with MJE 340/350 as drivers.
I am looking forward to hear from some of our experts out there, to give us some good advice on the choice of values on the gate resitors with the drivers and FETs described here.
Eivind Stillingen
Last edited:
Hi,
The speed as such does not come into it.
However the N-Channel Fet's have 1/3rd of the Cgd (Crss) and 2/3rd's of the Cgs (Ciss) of the P-Channel types. Now Cgs (Ciss) will be bootstrapped by the fact that an emitter follower is used.
For small signals the Z-Out of a bank of three Fet Pairs (K135/J50) is around 0.4R while for high frequencies the load Z is mostly dominated by the Boucherot cell, in this case around 10R.
So the effective gain will be Cgs will be 0.96 and Cgs will be bootstrapped to around 0.04 Times the actual value. For the N-Channel Fet this leaves 24pF per fet and for the P-Channel Fet this leaves 36pF per Fet.
The Cgd is not subject to any reduction though and is around 10pF for the N-Channel Fet and 30pF for the P-Channel Fet. So the effective Cin on this particular Amp (three pairs of Fets) would be 34pF per N-Channel Fet and 66pF per P-Channel Fet.
So using twice the gate resistor value for the N-Channel Fet makes sense IN THIS case.
I am unsure the logic you present is "sound" (pun intended).
The main purpose of the gate resistor is to "de-Q" resonant circuits formed from various parasitics so that the individual Fet does not turn into a Colpits Oscillator.
The resistor value for this is not neccesarily very high, but will depend on layout etc (e.g. the old Mosfet Amp's with the Fets on far away heatsinks connected with long wires are a much worse case than good PCB layouts).
Another part is that the source impedance to the Fet's must be accounted for.
Not quite. An Emitter follower is a bit like a gearbox, or perhaps better servo assisted steering. The effective capacitive load from the Mosfet's onto the VAS is reduced by the BJT's Beta (BSS71/74 have around 120 at 10mA). The price we pay is that we add the BJT's Cbc (Cob) to the capacitive load. The BSS71/74 are fantastically low at only 3.5pF each and equal for Npn and Pnp, not common.
Equally, the output impedance of the follower is basically the source impedance (around 10K) divided by the beta. So, with a beta of 120 and 10K source the output impedance will be around 83 Ohm (that is for both transistors combined).
The addition of the 330 Ohm resistor between the Emitters (without bypass) complicates things a little, but not too much.
This additional impedance, which is not in the schematic (but is normally simulated) forms a common "gate resistor" for the bank of Fet's it is connected to.
The larger fets may have more capacitance, but the relative relations remain. I would suggest that if doing such an Amp DIY and wishing to optimise it, one needs a fast (>50MHz) 'scope.
Then start initially with very "safe" values for the gate resistors and tune the amplifiers overall compensation. Then lower the gate resistors as much as possible, until the amplifier becomes unstable, try if changing the overal compensation restores stability, otherwise back of a little in lowering the gate resistors.
Ciao T
The speed as such does not come into it.
However the N-Channel Fet's have 1/3rd of the Cgd (Crss) and 2/3rd's of the Cgs (Ciss) of the P-Channel types. Now Cgs (Ciss) will be bootstrapped by the fact that an emitter follower is used.
For small signals the Z-Out of a bank of three Fet Pairs (K135/J50) is around 0.4R while for high frequencies the load Z is mostly dominated by the Boucherot cell, in this case around 10R.
So the effective gain will be Cgs will be 0.96 and Cgs will be bootstrapped to around 0.04 Times the actual value. For the N-Channel Fet this leaves 24pF per fet and for the P-Channel Fet this leaves 36pF per Fet.
The Cgd is not subject to any reduction though and is around 10pF for the N-Channel Fet and 30pF for the P-Channel Fet. So the effective Cin on this particular Amp (three pairs of Fets) would be 34pF per N-Channel Fet and 66pF per P-Channel Fet.
So using twice the gate resistor value for the N-Channel Fet makes sense IN THIS case.
I am unsure the logic you present is "sound" (pun intended).
The main purpose of the gate resistor is to "de-Q" resonant circuits formed from various parasitics so that the individual Fet does not turn into a Colpits Oscillator.
The resistor value for this is not neccesarily very high, but will depend on layout etc (e.g. the old Mosfet Amp's with the Fets on far away heatsinks connected with long wires are a much worse case than good PCB layouts).
Another part is that the source impedance to the Fet's must be accounted for.
Not quite. An Emitter follower is a bit like a gearbox, or perhaps better servo assisted steering. The effective capacitive load from the Mosfet's onto the VAS is reduced by the BJT's Beta (BSS71/74 have around 120 at 10mA). The price we pay is that we add the BJT's Cbc (Cob) to the capacitive load. The BSS71/74 are fantastically low at only 3.5pF each and equal for Npn and Pnp, not common.
Equally, the output impedance of the follower is basically the source impedance (around 10K) divided by the beta. So, with a beta of 120 and 10K source the output impedance will be around 83 Ohm (that is for both transistors combined).
The addition of the 330 Ohm resistor between the Emitters (without bypass) complicates things a little, but not too much.
This additional impedance, which is not in the schematic (but is normally simulated) forms a common "gate resistor" for the bank of Fet's it is connected to.
The larger fets may have more capacitance, but the relative relations remain. I would suggest that if doing such an Amp DIY and wishing to optimise it, one needs a fast (>50MHz) 'scope.
Then start initially with very "safe" values for the gate resistors and tune the amplifiers overall compensation. Then lower the gate resistors as much as possible, until the amplifier becomes unstable, try if changing the overal compensation restores stability, otherwise back of a little in lowering the gate resistors.
Ciao T
ThorstenL
Thanks for a very precise and satisfactory answers to my many questions.
When it comes to the question of what values are optimal gate resistor, then it is best determined by trying different values, while the results is observed on an oscilloscope, if I have understood you correctly.
Nico your answer confirms probably my conclusion on what ThorstenL recommended.
Nikos uses 100/220 ohms in his Goldmund. Let me add to (hope that's ok, Nikos) that in addition he has made the following changes from the original in order to stabilize:
C5 = 10 pF(mica ?, from 4,7 pF)
C6 = out
C7 = out
T7 between B-C: 100 pF(mica?)
Eivind Stillingen
Thanks for a very precise and satisfactory answers to my many questions.
When it comes to the question of what values are optimal gate resistor, then it is best determined by trying different values, while the results is observed on an oscilloscope, if I have understood you correctly.
Nico your answer confirms probably my conclusion on what ThorstenL recommended.
Nikos uses 100/220 ohms in his Goldmund. Let me add to (hope that's ok, Nikos) that in addition he has made the following changes from the original in order to stabilize:
C5 = 10 pF(mica ?, from 4,7 pF)
C6 = out
C7 = out
T7 between B-C: 100 pF(mica?)
Eivind Stillingen
ThorstenL
If you keep the driver voltage below +/- 75 V wouldn´t the pair 2sc3423/2sa1360 be a good substitute for BSS71/74?
In fact maybe for the MPSA types as well? The 2sc3423 has even lower Cob (1,5 pf) and 2sa 1360 has 2,5 pf.
They are in TO126 house , so it is far easier to place all drivers/predrivers on a common heatsink.
If you keep the driver voltage below +/- 75 V wouldn´t the pair 2sc3423/2sa1360 be a good substitute for BSS71/74?
In fact maybe for the MPSA types as well? The 2sc3423 has even lower Cob (1,5 pf) and 2sa 1360 has 2,5 pf.
They are in TO126 house , so it is far easier to place all drivers/predrivers on a common heatsink.
