Can you post the link to the Mirand A2 Sonnya?
(NB - still have to order those devices from you - Wednesday I took the day off to work on a PCB - I'll do it then!)
Click on last line of Sonnya comment. The amp you ask about is said to be a CFA topology. No schematic shown.
-RM
Last edited:
I've just deleted a whole pile of drivel, and some other content (because it was collateral). Waly Public warning. Keep up the trolling and you will get bin time (or worse). If you have nothing technical to add to this thread DO NOT POST. See the rules. http://www.diyaudio.com/forums/site-announcements/167561-diyaudio-rules.html taking particular note of rule 2.
Also take note of this clause in the rules:
You have all been warned before don't make us take drastic action!
As i see it.
The worst thing for BJT's is the Ic * Vce.
We will start out with SSA frontend, then TSSA and VSSA frontend. Then we will end up with the classic diamond buffer.
They all requires a great deal of work to perform well. The enemies are negative tempco of 2.3mV/K, Early voltage and Cob.
In any of them with no cascode you will have (Just an example) Vce of 40V. and an Ic of 4mA. This will produce an dissipation of 160mW.
If we take different packages:
SOT-23 => 568K/W (No copper)
SOT-223 => 166K/W (No copper)
TO-92 => 250K/W.
If we multiply the dissipation factor with the dissipation we get.
SOT-23 => 90,88K
SOT-223 => 26,56K.
TO-92 => 40K
So during the heat process (We asume constant room temp.) the Vbe voltage will drop:
SOT-23 => 209mV
SOT-223 => 61mV
TO-92 => 92mV
If we look on the SSA with no cascode driven by a CCS providing a "constant" Ie current, we will have (feedback network 1K||50R => 21x gain) a voltage drop across feedback network of 209mV (it is worst case with SOT-23) at least, meaning that the Ic will go up by at least 4,3mA. This is what MiiB talks about.
Of course with TO-92 it is down to 1.9mA but in my terms it is still a lot as we talk 50% increase in Ic during heatup process.
With cascode and Vce of 3.2volt (2 LED's) it is down to 12.8mW meaining 16.7mV (SOT-23) and 7,36mV (TO-92). This will give an Ic increase of 355uA (8,8%) and 156uA (3,9%). It is more acceptable but could be improved.
VSSA has an AC coupling of the 50R feedback resistor. This means that the Vbe change only looks into 1K.
Without cascode and Vce of 40volt, this will give an Ic increase of 209uA (5.2%) and 92uA (2.3%). It is a clear improvement above SSA.
With cascode and Vce of 3.2volt (2 LED's) it is down to 12.8mW meaining 16.7mV (SOT-23) and 7,36mV (TO-92). This will give an Ic increase of 16,7uA (0,4%) and 7,36uA (0,2%). It is more acceptable but could be improved.
The TSSA design has CFP input stage meaning that the input pair which is connected as the VSSA only has an Ic of 0.65mA. with Vce of 40V it will dissipate 26mW without cascode and with cascode 2.1mW.
For SOT-23 it will mean:
Without cascode : 14.8K=> Delta Vbe = 33,64mV => 33,64uA => 0.84%
With cascode : 1.2K => Delta Vbe = 2,74mV => 2,74uA => 0,065%
Clearly a lot better than SSA and VSSA.
Another benefit of the CFP in the TSSA input stage is that Re get reduced meaning more openloop gain.
In the diamond buffer all transistors needs to be placed close to each other. Again if it get cascoded we will talk similar drift values as the TSSA.
A drawback of the Diamond buffer is that we have an RE of 4.7 - 47R. This clearly reduces openloop gain as the feedback network is part of the loop.
Benefit of the TSSA and Diamond buffer is that we have two gain stages where the SSA and VSSA only have one. So any fluctation in Ic because of regulation will be less with diamond buffer and TSSA.
So in that case some designers chooses to use C version of BC546/556 BC550/560 BC846/856 BC850/860.
But by raising hFE the Earlyvoltage drops. Early voltage determens the hOE of the transistor. Low early voltage means low output impedance and has a negative influence of the distortion. That is why hawksford cascode is a good choice as the Vce is "constant" .
So my advice if i should give any. Wear cascode!
The worst thing for BJT's is the Ic * Vce.
We will start out with SSA frontend, then TSSA and VSSA frontend. Then we will end up with the classic diamond buffer.
They all requires a great deal of work to perform well. The enemies are negative tempco of 2.3mV/K, Early voltage and Cob.
In any of them with no cascode you will have (Just an example) Vce of 40V. and an Ic of 4mA. This will produce an dissipation of 160mW.
If we take different packages:
SOT-23 => 568K/W (No copper)
SOT-223 => 166K/W (No copper)
TO-92 => 250K/W.
If we multiply the dissipation factor with the dissipation we get.
SOT-23 => 90,88K
SOT-223 => 26,56K.
TO-92 => 40K
So during the heat process (We asume constant room temp.) the Vbe voltage will drop:
SOT-23 => 209mV
SOT-223 => 61mV
TO-92 => 92mV
If we look on the SSA with no cascode driven by a CCS providing a "constant" Ie current, we will have (feedback network 1K||50R => 21x gain) a voltage drop across feedback network of 209mV (it is worst case with SOT-23) at least, meaning that the Ic will go up by at least 4,3mA. This is what MiiB talks about.
Of course with TO-92 it is down to 1.9mA but in my terms it is still a lot as we talk 50% increase in Ic during heatup process.
With cascode and Vce of 3.2volt (2 LED's) it is down to 12.8mW meaining 16.7mV (SOT-23) and 7,36mV (TO-92). This will give an Ic increase of 355uA (8,8%) and 156uA (3,9%). It is more acceptable but could be improved.
VSSA has an AC coupling of the 50R feedback resistor. This means that the Vbe change only looks into 1K.
Without cascode and Vce of 40volt, this will give an Ic increase of 209uA (5.2%) and 92uA (2.3%). It is a clear improvement above SSA.
With cascode and Vce of 3.2volt (2 LED's) it is down to 12.8mW meaining 16.7mV (SOT-23) and 7,36mV (TO-92). This will give an Ic increase of 16,7uA (0,4%) and 7,36uA (0,2%). It is more acceptable but could be improved.
The TSSA design has CFP input stage meaning that the input pair which is connected as the VSSA only has an Ic of 0.65mA. with Vce of 40V it will dissipate 26mW without cascode and with cascode 2.1mW.
For SOT-23 it will mean:
Without cascode : 14.8K=> Delta Vbe = 33,64mV => 33,64uA => 0.84%
With cascode : 1.2K => Delta Vbe = 2,74mV => 2,74uA => 0,065%
Clearly a lot better than SSA and VSSA.
Another benefit of the CFP in the TSSA input stage is that Re get reduced meaning more openloop gain.
In the diamond buffer all transistors needs to be placed close to each other. Again if it get cascoded we will talk similar drift values as the TSSA.
A drawback of the Diamond buffer is that we have an RE of 4.7 - 47R. This clearly reduces openloop gain as the feedback network is part of the loop.
Benefit of the TSSA and Diamond buffer is that we have two gain stages where the SSA and VSSA only have one. So any fluctation in Ic because of regulation will be less with diamond buffer and TSSA.
So in that case some designers chooses to use C version of BC546/556 BC550/560 BC846/856 BC850/860.
But by raising hFE the Earlyvoltage drops. Early voltage determens the hOE of the transistor. Low early voltage means low output impedance and has a negative influence of the distortion. That is why hawksford cascode is a good choice as the Vce is "constant" .
So my advice if i should give any. Wear cascode!
Nice to have data.
What was the test procedure?
Number of tests per person?
Statistical confidence level?
Best wishes
David
RNMarsh (4280)
Then see what part of the topology or parameter of the topology is responsible for the sound difference. Can you talk to these points for us?
The output stage with a corrector Hawksford operates error signal which is equal to:
Rout * Iout
With precise balance corrector bridge Rout tends to zero, and hence tends to zero and the error signal.
However, between the amplifier and the speakers there is a speaker cable and a crossover that also have resistance. And as if there was little output impedance of the amplifier signal errors due to voltage drop across the speaker wires and elements crossover gets more. These distortions increase even more because of back EMF. Therefore, to minimize distortions in the real electro- heads must have a negative output impedance of the amplifier is equal to the total resistance of the cables and crossover elements. Error corrector Hawksford at small detuning allows to obtain the necessary negative output impedance due to positive current feedback output. Despite the slight increase in distortion of the sound quality of the amplifier without overall (global) negative feedback with the output negative impedance is considerably improved, it is more realistic.
Best regards
Petr
Then see what part of the topology or parameter of the topology is responsible for the sound difference. Can you talk to these points for us?
The output stage with a corrector Hawksford operates error signal which is equal to:
Rout * Iout
With precise balance corrector bridge Rout tends to zero, and hence tends to zero and the error signal.
However, between the amplifier and the speakers there is a speaker cable and a crossover that also have resistance. And as if there was little output impedance of the amplifier signal errors due to voltage drop across the speaker wires and elements crossover gets more. These distortions increase even more because of back EMF. Therefore, to minimize distortions in the real electro- heads must have a negative output impedance of the amplifier is equal to the total resistance of the cables and crossover elements. Error corrector Hawksford at small detuning allows to obtain the necessary negative output impedance due to positive current feedback output. Despite the slight increase in distortion of the sound quality of the amplifier without overall (global) negative feedback with the output negative impedance is considerably improved, it is more realistic.
Best regards
Petr
You may want to look here:
http://www.cordellaudio.com/papers/MOSFET_Power_Amp.pdf
It is the paper on a MOSFET power amplifier with error correction that I did in the early 80s. It discusses the Hawksford Error Correction (HEC) that it uses and the tunability of it.
Cheers,
Bob
DC coupling from input to output, like SM11S1, but with unbalanced input. I use Onsemi input and output devices.
The shown distortion is simulated at 100W/8ohm output power.
Sajti
MiiB (4288)
Petr
Could you post a schematic of your tunable error-correction OPS..??
I have seen this done on an amplifier with good results.
You can type in the search radiohibby 04 2013 - is the beginning of the article
ending in the next magazin radiohobby 05 2013
or look here
Ðàäèîõîááè ¹4 (àâãóñò 2013) » Ñêà÷àòü òîððåíò (page 50)
Ðàäèîõîááè ¹5 (îêòÿáðü 2013) » Ñêà÷àòü Ýëåêòðîííûå Æóðíàëû (page 51)
besr regards
Petr
Petr
Could you post a schematic of your tunable error-correction OPS..??
I have seen this done on an amplifier with good results.
You can type in the search radiohibby 04 2013 - is the beginning of the article
ending in the next magazin radiohobby 05 2013
or look here
Ðàäèîõîááè ¹4 (àâãóñò 2013) » Ñêà÷àòü òîððåíò (page 50)
Ðàäèîõîááè ¹5 (îêòÿáðü 2013) » Ñêà÷àòü Ýëåêòðîííûå Æóðíàëû (page 51)
besr regards
Petr
OS, could you post pics of these examples or even *.ASCs?Without taking the PSRR or thermal stability into consideration , the 2 simplest
circuits ... 1 VFA , 1 CFA.
The VFA would most likely be a single differential with a bootstrapped
or current sourced VAS. If bootstrapped, the extra transistor could be a LED CCS
(for the LTP).
There .. 4 whole devices.
The simplest CFA would be the VSSA ... 2 for the input pair / 2 for the VAS.
Again , 4 devices with a number of resistors and capacitors for support.
With 4 devices each , the VFA is a "toy" .... .05 -.1 % THD , no Superior PSRR.
On the other hand , a 4 device CFA (like the VSSA) ... especially if
crudely regulated , can provide low double digit PPM.
Performance wise , at this absolute minimum ... the CFA is WAY ahead
of the VFA. I'm a "VFA man" ... but I admit this.
With an 8-10 device CFA/VFA "shootout" , the field is leveled. One topology
with slightly lower THD ... one with a faster response. They actually tend
to "merge" (performance wise) at this device level.
It's difficult to understand what you are talking about without a picture .. worth a thousand words etc ..
I'm not familiar in MM7025, and I didn't find the shematic of it. However I keep searching...
Sajti
Although CFAs were already in use by Marantz, by 2006 all their equipment featured CFAs including analog stages in their digital equipment.
The same has happened with NAD with the exception of the multi channel AV recievers.
This is the very simple version of the Marantz CFA. It was derived from the multichannel home movie amplifiers...
Sajti