pietjers said:
I like the output stage you posted.
You don't "KNOW" that an output stage has very low distortion until you've measured it. If you just assume that any old emitter follower output stage has very low distortion you are being very naive. This has nothing to do with your assertion that there is often discrepency between perceived sound quality and what is measured.
BTW, you would be amazed at what you see when you propoerly and thoroughly measure an amplifier. Most people don't properly and thoroughly measure an amplifier, and this is part of the problem (but certainly not all of the problem).
Cheers,
Bob
Hawksford discusses some subtle cascode issues:
http://www.essex.ac.uk/dces/research/audio_lab/malcolmspubdocs/J10 Enhanced cascode.pdf
but with more emphasis on VAS/collector output
a careful reading will improve your understanding of cascode opteration and circuit options
the "feedback" topology he shows was in fact described by Baxandall and could give some output cascode advantages without as much output Vswing penalty
http://www.essex.ac.uk/dces/research/audio_lab/malcolmspubdocs/J10 Enhanced cascode.pdf
but with more emphasis on VAS/collector output
a careful reading will improve your understanding of cascode opteration and circuit options
the "feedback" topology he shows was in fact described by Baxandall and could give some output cascode advantages without as much output Vswing penalty
Peter,pietjers said:
Hhm, just a feeling from the stomach, so to say... I would perhaps rearrange the cascoding more horizontally than vertically. That is, each stage cascoded with the corresponding cascode transistors, not with the complete stage as a whole. Like cascoding predrivers with predrivers, drivers with drivers and outputs with outputs. This would shift the working conditions (and the timings issues) a bit (and would save two resitors). Hard to say if it would make things better or worse or maybe it's just insignificant...
@jcx: excellent Hawksford paper you linked to, thanks (I thought I had them all -- the relevant ones to amp design --, but obviously I missed this one).
- Klaus
I am playing with two types of cascoded output Designs for Headphone use however it is simple to upscale to what you need taking into account the Voltage limitations of the J-FETS used in the casscode current sources.
Type 1 is similar to what (Lumba Ogir) Purposed I did not use any emitter resistors on the Drivers and Pre-drivers and this works good. A constant current source (Not shown) As purposed in one of Mr Passes patents between the Collectors of the lower cascode transistors can be used to ensure glitch free start up but not required.
Type 2 is a more conventional cascode arrangement providing cascoding of the Drivers and pre-drivers to each other. this type is so far as shown proving to have intermittent stability issues, However when it is working it has both better sound and lower measured THD and IMD performance than type 1. Both Circuits show very low distortion and i use the type 1 as a standalone Line driver stage with no gain or overall feedback (I consider Emitter resistors local feedback and thus any circuit using them really can not claim totally feedback free) Pay close attention to the different locations of C8 in the Type 1 and type 2 designs as this location was chosen empirically to make stable. Has more benefit on type 1 topology than type 2 and i believe once the type 2 is made unconditionally stable C8 can be eliminated.
Type 1 is similar to what (Lumba Ogir) Purposed I did not use any emitter resistors on the Drivers and Pre-drivers and this works good. A constant current source (Not shown) As purposed in one of Mr Passes patents between the Collectors of the lower cascode transistors can be used to ensure glitch free start up but not required.
Type 2 is a more conventional cascode arrangement providing cascoding of the Drivers and pre-drivers to each other. this type is so far as shown proving to have intermittent stability issues, However when it is working it has both better sound and lower measured THD and IMD performance than type 1. Both Circuits show very low distortion and i use the type 1 as a standalone Line driver stage with no gain or overall feedback (I consider Emitter resistors local feedback and thus any circuit using them really can not claim totally feedback free) Pay close attention to the different locations of C8 in the Type 1 and type 2 designs as this location was chosen empirically to make stable. Has more benefit on type 1 topology than type 2 and i believe once the type 2 is made unconditionally stable C8 can be eliminated.
coloradosound said:
Schematics for the TA-N7, TA-N7B can be found in this thread:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=51798&perpage=25&pagenumber=1
Bob,
I did a quick read of the part about distortion measurement on your web-site. It takes quite a lot of experience and specialized equipment to analyse the distortion behaviour of an amplifier stage. What can one do with only a function generator and an oscilloscope?
Klaus,
Thank you for your great idea!
I think this is what you had in mind.
By fotopeter
I did a quick read of the part about distortion measurement on your web-site. It takes quite a lot of experience and specialized equipment to analyse the distortion behaviour of an amplifier stage. What can one do with only a function generator and an oscilloscope?
Klaus,
Thank you for your great idea!
I think this is what you had in mind.
An externally hosted image should be here but it was not working when we last tested it.
By fotopeter
-------------------------------------------------------------------------
pietjers wrote:
I did a quick read of the part about distortion measurement on your web-site. It takes quite a lot of experience and specialized equipment to analyse the distortion behaviour of an amplifier stage. What can one do with only a function generator and an oscilloscope?
-------------------------------------------------------------------------------------
a function generator, while not a low distortion oscillator can be used with a dual trace oscope to function very similar to a distortion analyzer. set the vertical mode to ADD, and the ch2 mode to INVERT. apply the amp output to ch1, and the amp input to ch2. input a sine wave from tthe function generator. switch the scope vertical mode to ch2 and adjust the gain for full screen deflection. then switch the vertical mode to ch1 and adjust for full deflection. now switch back to add mode and observe the distortion residual. it's not perfect because of the glitches in the function generator sine wave, but it's close
pietjers wrote:
I did a quick read of the part about distortion measurement on your web-site. It takes quite a lot of experience and specialized equipment to analyse the distortion behaviour of an amplifier stage. What can one do with only a function generator and an oscilloscope?
-------------------------------------------------------------------------------------
a function generator, while not a low distortion oscillator can be used with a dual trace oscope to function very similar to a distortion analyzer. set the vertical mode to ADD, and the ch2 mode to INVERT. apply the amp output to ch1, and the amp input to ch2. input a sine wave from tthe function generator. switch the scope vertical mode to ch2 and adjust the gain for full screen deflection. then switch the vertical mode to ch1 and adjust for full deflection. now switch back to add mode and observe the distortion residual. it's not perfect because of the glitches in the function generator sine wave, but it's close
Thanks Philip, (nice stages you have there). A bigger difference seems to be that you derive the cascode voltages from the output. Maybe one could get away with lower/no cascode emitter resistors with a input derive topology (as they defeat part of the cascode action depending on standing/max current ratio in the predriver/driver)....
- Klaus
- Klaus
Peter, you might also want to look into a soundcard based measurement solution, besides the scope nulling trick as described by "the Uncle" -- or hardware variant (nulling circuit), or a distortion magnifier (a specialized notch filter) together with a good generator.
With a common soundcard (24bits, 96kHz or 192kHz) and generator/analyzer software you can get quite a lot of resolution and bandwith for the given investment cost. I can get some 90dB reliable range at least for HD and IMD measurents, at 40kHz bandwith (24/96 card). With specialized measurements one can go even lower. It's not as convenient to handle as a hardware setup and you can easily blow channels with overvoltage if you don't clamp/protect inputs plus there ar some other issues (grounding etc), but I feel it gives the best results for the money, especially when you build your own driver and probe adapter electronics for the various measurement tasks (differential probe/drive, current sensors etc).
- Klaus
With a common soundcard (24bits, 96kHz or 192kHz) and generator/analyzer software you can get quite a lot of resolution and bandwith for the given investment cost. I can get some 90dB reliable range at least for HD and IMD measurents, at 40kHz bandwith (24/96 card). With specialized measurements one can go even lower. It's not as convenient to handle as a hardware setup and you can easily blow channels with overvoltage if you don't clamp/protect inputs plus there ar some other issues (grounding etc), but I feel it gives the best results for the money, especially when you build your own driver and probe adapter electronics for the various measurement tasks (differential probe/drive, current sensors etc).
- Klaus
KSTR said:
The Right Mark Audio Software based system that is also free is intended to test sound cards however with some Creativity and research on this topic (Google) you can measure amplifiers and sutch providing an FFT plot of the Distortion picture. Right Mark Audio
PPL,
It’s funny how similar your designs and my design ideas (are planning to give them a try in real life) are. The fun is that I hadn’t opened the shortcut in your post #51 before I posted my second hand drawn schematic. Saw it after posting when I was re reading the last posts.
I have an old Tektronics AA501 Distortion Analyser plug in. But the pushbutton switches are quite bad and not easy to get okay with contact cleaner. This analyser has a function output showing filtered distortion. But it’s very important to have a low distortion generator because normal generators have so much distortion that it will dominate the measurement. Must have also a Tek low distortion generator somewhere……., will perform a search soon.
Most difficult part will probably be to get everything stable!! I have built already the inner part of the cascode (the triple darlington part) and the only way to get it really stable is be means of the resistor at the input. From 2k and up it’s stable. Below 2k and especially below 1k the circuit starts to oscillate. I have tried about everything one can think of to control HF stability, but always the input resistor turned out to be the best and most effective solution.
What I tried (al the time with a square wave signal of a few V and a frequency of about 200kHz at the input to be able to closely observe ringing and oscillation behaviour):
• Adding base stoppers. Did not really improve stability. These resistors are only effective in the VHF region I think. The drawback of base stoppers (especially at the output devices) is that the output impedance of the circuit rises (want to use it without feedback).
• Adding capacitors from the base of the pre-drivers, drivers and output devices to supply voltage (tried it one by one). Found that their values are quite critical. Too small, little or no improvement. To big => instability. When about optimal still certain circumstances where the idling current starts to quickly run away for unknown reasons.
• Adding small capacitors from input to ground is a little playing with fire. Can also easily cause instability.
• Adding a small capacitor from output to input. Didn’t add much stability either. Perhaps I should have added a series resistor with the capacitor here.
But however I will try to get it up and running!
If you have suggestions, please let me know. I have already heard the suggestion to add resistors between the upper and lower half of the cascode, so I have added those.
Peter
It’s funny how similar your designs and my design ideas (are planning to give them a try in real life) are. The fun is that I hadn’t opened the shortcut in your post #51 before I posted my second hand drawn schematic. Saw it after posting when I was re reading the last posts.
I have an old Tektronics AA501 Distortion Analyser plug in. But the pushbutton switches are quite bad and not easy to get okay with contact cleaner. This analyser has a function output showing filtered distortion. But it’s very important to have a low distortion generator because normal generators have so much distortion that it will dominate the measurement. Must have also a Tek low distortion generator somewhere……., will perform a search soon.
Most difficult part will probably be to get everything stable!! I have built already the inner part of the cascode (the triple darlington part) and the only way to get it really stable is be means of the resistor at the input. From 2k and up it’s stable. Below 2k and especially below 1k the circuit starts to oscillate. I have tried about everything one can think of to control HF stability, but always the input resistor turned out to be the best and most effective solution.
What I tried (al the time with a square wave signal of a few V and a frequency of about 200kHz at the input to be able to closely observe ringing and oscillation behaviour):
• Adding base stoppers. Did not really improve stability. These resistors are only effective in the VHF region I think. The drawback of base stoppers (especially at the output devices) is that the output impedance of the circuit rises (want to use it without feedback).
• Adding capacitors from the base of the pre-drivers, drivers and output devices to supply voltage (tried it one by one). Found that their values are quite critical. Too small, little or no improvement. To big => instability. When about optimal still certain circumstances where the idling current starts to quickly run away for unknown reasons.
• Adding small capacitors from input to ground is a little playing with fire. Can also easily cause instability.
• Adding a small capacitor from output to input. Didn’t add much stability either. Perhaps I should have added a series resistor with the capacitor here.
But however I will try to get it up and running!
If you have suggestions, please let me know. I have already heard the suggestion to add resistors between the upper and lower half of the cascode, so I have added those.
Peter
An externally hosted image should be here but it was not working when we last tested it.
what does yout input/VAS stage look like? i've found that if the input and VAS stages are too fast for the output stage, the design is uncontrollably unstable. also a triple darlington can be a difficult beast to tame. miller capacitance is junction capacitance multiplied by beta. so let's say your individual transistor betas are something typical like this:
predriver 200
driver 100
output 75
that's a total beta of:
200x100x75=1,500,000 !!!!!
so whatever junction capacitances you have, mainly the output devices, gets multiplied by one and a half million, and output device Cbco is about 100pf on average, some higher, some lower. a juncyion capacitance of 100pf in the output stage becomes a miller capacitance of 150uf with that kind of beta multiplication.
predriver 200
driver 100
output 75
that's a total beta of:
200x100x75=1,500,000 !!!!!
so whatever junction capacitances you have, mainly the output devices, gets multiplied by one and a half million, and output device Cbco is about 100pf on average, some higher, some lower. a juncyion capacitance of 100pf in the output stage becomes a miller capacitance of 150uf with that kind of beta multiplication.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.