Looking for maximal slew-rate reminds the name of Stochino :
High-speed 100W amplifier design
http://www.diyaudio.com/forums/solid-state/157-giovanni-stochinos-ultra-fast-amp.html
Worth to mention the clever Sansui diamond input stage, using a single LTP at the input :
US Pat 4,229,705
It has been said that pushing the game of high slew-rate too far was disastrous for this company, their last ultra-fast Mos amps were unstable and often crashed.
High-speed 100W amplifier design
http://www.diyaudio.com/forums/solid-state/157-giovanni-stochinos-ultra-fast-amp.html
An externally hosted image should be here but it was not working when we last tested it.
Worth to mention the clever Sansui diamond input stage, using a single LTP at the input :
US Pat 4,229,705
It has been said that pushing the game of high slew-rate too far was disastrous for this company, their last ultra-fast Mos amps were unstable and often crashed.
I saw that one. It's more complicated than I want to analyze at the moment. What is with all the diodes?
I've been building up a classic LTP and VAS from scratch using the recommended methods, and I've got it up to 120V/us. THD levels are suffering though. Open loop gain is only 52dB. LTP bias current is 2mA and VAS is 9.5mA, so I think there's still wiggle room for more improvement.
I still have yet to build out the output section. That might present some surprises.
Oh yeah the LTP tail is just a resistor. I thought I try it without a current source, since I saw comments about some people preferring the tone of a simple resistor over a CCS in the LTP.
The current feedback amps seem to achieve hundreds of volts per usec with ease. How do they sound in comparison with classic topologies? The only way I would get to hear the difference for myself is to build some and have a listen. I'd have to build some better monitors though.
If they sound harsher in the highs, I'm thinking that's because of the media rather than any drawback from a fast amp. 44.1khz or even 48khz is too low of a sampling rate. The fast amps just make that noticable, instead of glazing over the harshness of an insufficient data rate as a slow amp would do, acting as kind of secondary aliasing filter.
I've been building up a classic LTP and VAS from scratch using the recommended methods, and I've got it up to 120V/us. THD levels are suffering though. Open loop gain is only 52dB. LTP bias current is 2mA and VAS is 9.5mA, so I think there's still wiggle room for more improvement.
I still have yet to build out the output section. That might present some surprises.
Oh yeah the LTP tail is just a resistor. I thought I try it without a current source, since I saw comments about some people preferring the tone of a simple resistor over a CCS in the LTP.
The current feedback amps seem to achieve hundreds of volts per usec with ease. How do they sound in comparison with classic topologies? The only way I would get to hear the difference for myself is to build some and have a listen. I'd have to build some better monitors though.
If they sound harsher in the highs, I'm thinking that's because of the media rather than any drawback from a fast amp. 44.1khz or even 48khz is too low of a sampling rate. The fast amps just make that noticable, instead of glazing over the harshness of an insufficient data rate as a slow amp would do, acting as kind of secondary aliasing filter.
Last edited:
the normalization to amplifier Vpeak is the only fair way to compare
that would be the slope of a ~80 kHz full amplitude sine
the excess margin over 20 kHz is sometimes considered a proxy for normal level in-band linearity - but that depends on topolgy, compensation - various linearized diff pair or 2-pole compensation change the relations
that would be the slope of a ~80 kHz full amplitude sine
the excess margin over 20 kHz is sometimes considered a proxy for normal level in-band linearity - but that depends on topolgy, compensation - various linearized diff pair or 2-pole compensation change the relations
Maybe we can split this into two
1. Slewing amplifiers, of which the classic Cdom compensated VFA is the best example. My e-Amp and Ovation 250 are of this type. Stochino's amp actually falls into this class - he achieves high slew rates because he uses high LTP currents and low values for Cdom.
2. Non slewing types, of which the CFA is the best example. The sx and nx-Amp are examples. We could also stretch this to include MIC comp'd designs as well (my view only on this one)
In type 1 you will always get the amplifier to slew at some HF frequency determined by Cdom and the available LTP current. In type 2, you don't get slewing, just a reduction in amplitude as you increase frequency.
1. Slewing amplifiers, of which the classic Cdom compensated VFA is the best example. My e-Amp and Ovation 250 are of this type. Stochino's amp actually falls into this class - he achieves high slew rates because he uses high LTP currents and low values for Cdom.
2. Non slewing types, of which the CFA is the best example. The sx and nx-Amp are examples. We could also stretch this to include MIC comp'd designs as well (my view only on this one)
In type 1 you will always get the amplifier to slew at some HF frequency determined by Cdom and the available LTP current. In type 2, you don't get slewing, just a reduction in amplitude as you increase frequency.
Last edited:
I haven't bothered to Analyze the whole thing either, but in general......
It's interesting to look at what happens when an amp is clipping. In most amps, the output voltage is limited by a transistor somewhere saturating. Then during clipping the input stage is overloaded, which probably results in some other transistors getting saturated as well.
The problem is that transistors (especially power transistors) tend to be quite slow to come out of saturation. So when the input signal drops back down below the clipping level, the output is slow to respond and stays stuck at the maximum output voltage too long. the result can look like the first pic below.
The cure for this is to use fast diodes to prevent the transistors from saturating. Looking at the circled parts in the 2'nd pic below:
In the part on the left, D1 and D2 prevent Q3 from saturating. Without the diodes, it would saturate when Q2 is conducting much more current than Q1.
Similarly, on the right D3 and D4 prevent Q5 from saturating.
Attachments
godfrey,
What would be the advantage of using all those diodes instead of setting up the input and voltage gains stage to avoid clipping under normal operation?
Bonsai,
With regard to your nx amp, you bootstrapped the input section to increase its
input impedance and used a CFP in the bias servo to increase it's gain?
What would be the advantage of using all those diodes instead of setting up the input and voltage gains stage to avoid clipping under normal operation?
Bonsai,
With regard to your nx amp, you bootstrapped the input section to increase its
input impedance and used a CFP in the bias servo to increase it's gain?
Last edited:
With regard to the schematic for the nx in the pdf, R1 C1 R4 R11 looked at first glance to resemble a bootstrap, although now that I look closer at it, it looks more like a filter arrangement for the adjustable off set.
It looks like Q3 and Q20 form a complementary feedback pair for the bias servo. Why would you want to increase the gain there, or is there some other reason for that?
Well done on the write ups. I see you've read and explored a lot of the things that interested me as well, like the SS relay for over current protection. I have been wanting to build a small amp to power some reference monitors for mixing down music. I thin something like your sx amp would be perfect.
Are you going to provide PCBs or gerber files for the layout for that amp?
Last edited:
The F5 is an amp.
I mean just the input stage, like this kind of thing:
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.