Let's examine the dissipation of the fb resistor.
Assume that the complete output voltage is applied to the emitter of the input transistor, Q1. In fact it's a bit less by the input voltage, but the reduction is only worked out on the gain, and is only 3% (gain of 30, for example), so assume, with 42V rails, that we can produce 37.5x2=75Vpp across the resistor. This is 53Vpp rms.
Dissipation is voltage dropped multiplied by corresponding current.
Here is an interesting fact; the DC current is pretty much constant, but the AC does vary because its AC load is in fact the shunt resistor, 47R in this case. The dissipation is Eexp2/R from Ohms Law.
So for calculating the full dissipation, assume 1.5k + 47R = 1547 ohms.
This is 1.82W, but it's only relevant to continuous sine waye, or square wave on 1:1 mark space ratio.
For music, it really is about 25% of this. So the continuous power dissipation in this 1.5K series fb resistor on MUSIC is around 460mW. And this is at full output, that is, lifting the roof, burning up the speakers, and attracting the neighbours. Even at very high levels, and with your ears bleeding, you would max out at around 300mW on this resistor.
A 1W resistor would be absolutely fine, and in truth I would use a 600mW Beyschlag through hole in this position, even in my bigger amps, which in fact are voltage fb amps, not quite such a high dissipation regards at the higher output in volts.
Cheers,
Hugh
Assume that the complete output voltage is applied to the emitter of the input transistor, Q1. In fact it's a bit less by the input voltage, but the reduction is only worked out on the gain, and is only 3% (gain of 30, for example), so assume, with 42V rails, that we can produce 37.5x2=75Vpp across the resistor. This is 53Vpp rms.
Dissipation is voltage dropped multiplied by corresponding current.
Here is an interesting fact; the DC current is pretty much constant, but the AC does vary because its AC load is in fact the shunt resistor, 47R in this case. The dissipation is Eexp2/R from Ohms Law.
So for calculating the full dissipation, assume 1.5k + 47R = 1547 ohms.
This is 1.82W, but it's only relevant to continuous sine waye, or square wave on 1:1 mark space ratio.
For music, it really is about 25% of this. So the continuous power dissipation in this 1.5K series fb resistor on MUSIC is around 460mW. And this is at full output, that is, lifting the roof, burning up the speakers, and attracting the neighbours. Even at very high levels, and with your ears bleeding, you would max out at around 300mW on this resistor.
A 1W resistor would be absolutely fine, and in truth I would use a 600mW Beyschlag through hole in this position, even in my bigger amps, which in fact are voltage fb amps, not quite such a high dissipation regards at the higher output in volts.
Cheers,
Hugh
Last edited:
Let's examine the dissipation of the fb resistor.
Assume that the complete output voltage is applied to the emitter of the input transistor, Q1. In fact it's a bit less by the input voltage, but the reduction is only worked out on the gain, and is only 3% (gain of 30, for example), so assume, with 42V rails, that we can produce 37.5x2=75Vpp across the resistor. This is 53Vpp rms.
Dissipation is voltage dropped multiplied by corresponding current.
Here is an interesting fact; the DC current is pretty much constant, but the AC does vary because its AC load is in fact the shunt resistor, 47R in this case. The dissipation is Eexp2/R from Ohms Law.
So for calculating the full dissipation, assume 1.5k + 47R = 1547 ohms.
This is 1.82W, but it's only relevant to continuous sine waye, or square wave on 1:1 mark space ratio.
For music, it really is about 25% of this. So the continuous power dissipation in this 1.5K series fb resistor on MUSIC is around 460mW. And this is at full output, that is, lifting the roof, burning up the speakers, and attracting the neighbours. Even at very high levels, and with your ears bleeding, you would max out at around 300mW on this resistor.
A 1W resistor would be absolutely fine, and in truth I would use a 600mW Beyschlag through hole in this position, even in my bigger amps, which in fact are voltage fb amps, not quite such a high dissipation regards at the higher output in volts.
Cheers,
Hugh
but completely avoided the nonlineraity of resistance and ignored the tempco of the resistance.Well explained.
Thanks.
Transients do heat up the resistances and the resistances do change in value.
The result is that the feedback ratios change and that changes the output, which reverts to normal AFTER the transient has passed and the resistor element begins cooling.
Hi Andrew
It is true that some of my boards feature paralleled low(ish) wattage, high precision and very low temp co. feedback resistors. You made a good point here or elsewhere that perhaps series connections confer some benefits.
Unfortunately I don't own an APx555 and even if I did I'm doubtful it would be possible to measure the improvement; certainly I can't hear any difference when paralleling feedback resistors. Series connections? Well... I'll try it on my next project (that's a promise).
This is a "very simple quasi hybrid" and I'm prepared to guarantee that constructors will not disappointed with the sound on account of the single 1W feedback resistor as specified.
Cheers
Christian
It is true that some of my boards feature paralleled low(ish) wattage, high precision and very low temp co. feedback resistors. You made a good point here or elsewhere that perhaps series connections confer some benefits.
Unfortunately I don't own an APx555 and even if I did I'm doubtful it would be possible to measure the improvement; certainly I can't hear any difference when paralleling feedback resistors. Series connections? Well... I'll try it on my next project (that's a promise).
This is a "very simple quasi hybrid" and I'm prepared to guarantee that constructors will not disappointed with the sound on account of the single 1W feedback resistor as specified.
Cheers
Christian
Thiago, that s simply fantastic, thanks againPCB here Updated
Regards
J.
Unfortunately this is a common failing with layout designers that don't understand the implications of thermal effects on transients. <snip>
This is a bigger issue with "CFA" style amplifiers, where we see ~1k feedback values instead of 30k in a comparable "VFA".
certainly I can't hear any difference when paralleling feedback resistors. Series connections?
Ironical, isn't it. We're not anymore talking about thermal effects, Johnson's noise, etc. but about audibility of things.
Smaller feedback values tend to sound better. That's the phenomenon (not the opposite). Imo, why it is so is more interesting/useful than why it is not so.
but completely avoided the nonlineraity of resistance and ignored the tempco of the resistance.
Andrew,
The nonlinearity of resistors is a big issue in instrumentation. This amp has 0.05% THD, and the linearity which drive harmonics beyond H2, H3 and H4 would be less than the linearity issues of the actives! If you use a metal film through hole (like a Beyschlag I use) the predominating issue would be the inductance rather than the change of resistance with heating, which would be very low as well. I would think the changes of beta/transconductance in the actives would be far more important with heating than the linearity of the resistors.
Transients do heat up the resistances and the resistances do change in value.
The result is that the feedback ratios change and that changes the output, which reverts to normal AFTER the transient has passed and the resistor element begins cooling.
Yes, they do heat up, no question, but the effects are pretty slow and therefore would only affect <10Hz issues. This is the reason I always use a 600mW metal film in this role because the thermal mass is far more than a smd.
I believe your resistor thermal linearity issues on this amp are pretty small and would actually have no effect on the sound quality. If you are measuring a scientific instrument, yes, but this is pretty small potatoes compared to the actives which should be very, very carefully watched for dissipation during design since their higher temperatures might risk blowing them.
Thanks for your suggestions,
Hugh
Prasi we need rev. 4.3 Board
Hallo All.
Now it is a lucky day. I could good laughter, because last posts here.
Very good math, done- thanks Hugh- i could understand. #601
Yesterday i have placed in parts on board what i actually have.
So i came up to 1381 and discover that the holes are a bit to tiny.
Do not take a Hammer to place it in!
So i have careful drilled it to 1,0mm.
I have all new information noticed- thanks for the Excel- xrk- you
can look there.
The thiagomogi little board with 2 x 1000µ/16V is smart done.
Now i must order on big blue C- Apotheke parts.
Thimios: look at my little Amigo- hopefully it would blooming!
Bangla.
Hallo All.
Now it is a lucky day. I could good laughter, because last posts here.
Very good math, done- thanks Hugh- i could understand. #601
Yesterday i have placed in parts on board what i actually have.
So i came up to 1381 and discover that the holes are a bit to tiny.
Do not take a Hammer to place it in!
So i have careful drilled it to 1,0mm.
I have all new information noticed- thanks for the Excel- xrk- you
can look there.
The thiagomogi little board with 2 x 1000µ/16V is smart done.
Now i must order on big blue C- Apotheke parts.
Thimios: look at my little Amigo- hopefully it would blooming!
Bangla.
Attachments
Hallo All.
Now it is a lucky day. I could good laughter, because last posts here.
Very good math, done- thanks Hugh- i could understand. #601
Yesterday i have placed in parts on board what i actually have.
So i came up to 1381 and discover that the holes are a bit to tiny.
Do not take a Hammer to place it in!
So i have careful drilled it to 1,0mm.
I have all new information noticed- thanks for the Excel- xrk- you
can look there.
The thiagomogi little board with 2 x 1000µ/16V is smart done.
Now i must order on big blue C- Apotheke parts.
Thimios: look at my little Amigo- hopefully it would blooming!
Bangla.
Hi Bangla,
Seems you have been doing lots of homework with all those notes. Yes rev 3 had 0.7mm holes(some downloaded library package created by someone ). I have now increaded to 0.9mm here.
just noticed, Did you use a bigger resistor for R17 for any particular reason?
reg
prasi
Last edited:
Hi Christian,Hi Prasi
I've been fine tuning and tweaking the past 24 hours to perfect this on 35-42V rails....
C2 = 220p (with the 1n we run the risk of the HPF intruding in the audible band with high Z sources)
R5 = 100k (thanks Terry)
R7 = 680R
R9 = 1k5 /1W
R10 = 68k (I changed my mind after discussion with Hugh)
R12 should be a 600mW part
R15 = 1k2
R19 = 120R (thanks thimios)
R25 should be a 1W part
R23 = 10R
It also be good if you are able to add pads for a ceramic bypass cap around the feedback resistor - R9. More testing is required but if we are going to organise a group buy it would be good to have the option.
I sat down today and changed the above values. I found that with changing R9 to 1k5 I had to change R5 to get the offset in range so I did as before and installed a 100k pot in place of R5. I centered VR1 and adjusted the pot until I had +/- 1mV. When I removed the pot and measured it it was approximately 50k so I installed a 51k resistor there and the offset can be easily adjusted with the 200R trimmer in VR1. Both channels are the same so it is likely that 51k For R5 is a good value.
This testing was was done with +/- 37V rails. I will check it with higher voltage and make sure it still works. Also, with the 37V rails Q3 seems OK without a heatsink.
Yes you are correct. You may need to show a range for that transistor. Another thing I have been meaning to ask. Have you or Thimios noticed if your amp has an issue with being started cold with a load attached? Mine seems to not like that. Not an issue I suppose if used with a protection circuit that has a short delay.
The only problem I have observed is that when the trimpot is adjusted to null the offset once in the warmed up state (as you should) there can be a small DC offset on cold start (up to 200mV) - which quickly reduces. I believe I made reference to this earlier in the thread. The "cure" is to replace the bias chain with a proper DC servo. I build a prototype with a servo consisting of a dual package JFET opamp and a handful of discretes, and it worked flawlessly, holding the offset to within 1mV under all conditions. But this approach increases the complexity, component count and arguably the sound quality, so I've stuck with the simple method despite its (minor) weaknesses.
As with many amplifiers, there is also a small turn on pop - the amplitude and duration of the transient is not severe on my prototype.
Are these the sort or issues you are encountering, Terry?
As with many amplifiers, there is also a small turn on pop - the amplitude and duration of the transient is not severe on my prototype.
Are these the sort or issues you are encountering, Terry?
re posts #612 to #614,
The turn on behaviour of amplifier of this type (ie non-balanced, bootstrap VAS etc) will depend on the charging times of the capacitors C1, C3 and C4 (the ones circled in red).
When the FetZilla amp was being designed, I ran about 100 versions of LTspice to find the optimum value for these capacitors so that the turn on thump was a nice low frequency thump with no sudden transitions (so you don't hear it). For this Quasi, C1 is basically fixed (either 1uF or 1.5uF), and C4 needs to be around 1000uF or bigger, so only C3 can really be altered.
Someone (and I'm not volunteering) needs to run a series of LTprice runs, or try with a series of C3 values with a real amplifier. Try and get the turn on `thump' to have no sudden transitions (as these will end up in the tweeter), a suitable low frequency (so only the woofer moves), and that it doesn't stick to one of the power rails.
Terry, are you able to record what the turn on thump is on you amplifier - positive or negative, frequency etc. A storage oscilloscope is handy for this.
Also I notice that the gain is quite high (47 plus 1500 gives a gain of 33, reduced a bit by R2 + R4//R5). A lot of commercial amplifier have a gain of 26dB (ie a voltage gain of 21) and I think that would be about right for a low power amplifier. I suggest you try a larger value for R8, such as R8=68 and R9=1500 or R8=56 and R9=1200 (the yellow resistors on the schematic), which give a gain of about 22.
I think I'd make R4 bigger - probably 100k the same as R5.
Regards,
Paul Bysouth
The turn on behaviour of amplifier of this type (ie non-balanced, bootstrap VAS etc) will depend on the charging times of the capacitors C1, C3 and C4 (the ones circled in red).
When the FetZilla amp was being designed, I ran about 100 versions of LTspice to find the optimum value for these capacitors so that the turn on thump was a nice low frequency thump with no sudden transitions (so you don't hear it). For this Quasi, C1 is basically fixed (either 1uF or 1.5uF), and C4 needs to be around 1000uF or bigger, so only C3 can really be altered.
Someone (and I'm not volunteering) needs to run a series of LTprice runs, or try with a series of C3 values with a real amplifier. Try and get the turn on `thump' to have no sudden transitions (as these will end up in the tweeter), a suitable low frequency (so only the woofer moves), and that it doesn't stick to one of the power rails.
Terry, are you able to record what the turn on thump is on you amplifier - positive or negative, frequency etc. A storage oscilloscope is handy for this.
Also I notice that the gain is quite high (47 plus 1500 gives a gain of 33, reduced a bit by R2 + R4//R5). A lot of commercial amplifier have a gain of 26dB (ie a voltage gain of 21) and I think that would be about right for a low power amplifier. I suggest you try a larger value for R8, such as R8=68 and R9=1500 or R8=56 and R9=1200 (the yellow resistors on the schematic), which give a gain of about 22.
I think I'd make R4 bigger - probably 100k the same as R5.
Regards,
Paul Bysouth
Attachments
Strange,i haven't noticed any damp at start up.Yes you are correct. You may need to show a range for that transistor. Another thing I have been meaning to ask. Have you or Thimios noticed if your amp has an issue with being started cold with a load attached? Mine seems to not like that. Not an issue I suppose if used with a protection circuit that has a short delay.
Yes,trimming offset when amplifier is cold isn't useful .
I prefer the worm adjust.
Last edited:
Very simple stability test
Here is the test.
Please keep in mind that amplifier under test is version without capacitor in the inverter not at feedback resistor.
A Fluke 179 attached as thermometer to zobel resistor when another connected as offset voltmeter.
Sorry i must resize the pictures.....
Comming soon.
Ok,here is the test with 0.1uf//6R at 1KHz and 20KHz plus with 1uf//6R at 1KHz and 20KHz
Here is the test.
Please keep in mind that amplifier under test is version without capacitor in the inverter not at feedback resistor.
A Fluke 179 attached as thermometer to zobel resistor when another connected as offset voltmeter.
Sorry i must resize the pictures.....
Comming soon.
Ok,here is the test with 0.1uf//6R at 1KHz and 20KHz plus with 1uf//6R at 1KHz and 20KHz
Attachments
Last edited:
Hi thimios
Thanks for taking the time to run these tests.
All the sine waves appear extremely good. In terms of the square waves, well, they are horrible, lots of ringing and overshoot, but that is to be expected when driving squares into a reactive load.
Personally, I only square wave test at 1W (2.8Vrms) into a purely resistive 8R load (no parallel cap) at 1k and 20k Hz. Despite the fact that no music signal contains square waves, the sharp transitions make apparent potential stability concerns that might be missed on a 'scope when looking at a sine wave or music signal.
I've never considered driving squares into a reactive load as a valid or useful test.
Does anybody else have an opinion?
Thanks for taking the time to run these tests.
All the sine waves appear extremely good. In terms of the square waves, well, they are horrible, lots of ringing and overshoot, but that is to be expected when driving squares into a reactive load.
Personally, I only square wave test at 1W (2.8Vrms) into a purely resistive 8R load (no parallel cap) at 1k and 20k Hz. Despite the fact that no music signal contains square waves, the sharp transitions make apparent potential stability concerns that might be missed on a 'scope when looking at a sine wave or music signal.
I've never considered driving squares into a reactive load as a valid or useful test.
Does anybody else have an opinion?
- Home
- Amplifiers
- Solid State
- Very simple quasi complimentary MOSFET amplifier