Hi all
A few years ago I built an amp very heavily inspired by Douglas Self's "Blameless" designs. The performance was very good except that the two-transistor VAS would saturate heavily on clipping and take a long time to recover. Long enough that if it was driven into heavy clipping at 20kHz, the duty cycle would be seriously off and the DC offset protection would shut the amp down.
I believe most of Self's amps probably do this and he either didn't notice or dismissed it with the comment that "This is not a treatise on fuzz-boxes."
Recently I realised that the two-transistor VAS can also be Baker clamped very easily by adding a single diode across the compensation capacitor. I tested this on my amp and it worked a treat. The positive side already clipped reasonably well, so this essentially fixed the whole issue.
The diode might be expected to cause a small increase in THD at high frequencies due to its non-linear capacitance, but I can live with that. I used a BAV20 which has a capacitance of maybe 1pF with a reasonable reverse voltage, comparable to the capacitances of the other transistors at the VAS node, and Cdom is 100pF anyway. I measured THD-20 unloaded after adding the diode and it increased from maybe 0.005% to 0.006%.
So now you can listen to audiophile-grade music with huge dynamics, or plug in your Les Paul and chug away. 🙂
A few years ago I built an amp very heavily inspired by Douglas Self's "Blameless" designs. The performance was very good except that the two-transistor VAS would saturate heavily on clipping and take a long time to recover. Long enough that if it was driven into heavy clipping at 20kHz, the duty cycle would be seriously off and the DC offset protection would shut the amp down.
I believe most of Self's amps probably do this and he either didn't notice or dismissed it with the comment that "This is not a treatise on fuzz-boxes."
Recently I realised that the two-transistor VAS can also be Baker clamped very easily by adding a single diode across the compensation capacitor. I tested this on my amp and it worked a treat. The positive side already clipped reasonably well, so this essentially fixed the whole issue.
The diode might be expected to cause a small increase in THD at high frequencies due to its non-linear capacitance, but I can live with that. I used a BAV20 which has a capacitance of maybe 1pF with a reasonable reverse voltage, comparable to the capacitances of the other transistors at the VAS node, and Cdom is 100pF anyway. I measured THD-20 unloaded after adding the diode and it increased from maybe 0.005% to 0.006%.
So now you can listen to audiophile-grade music with huge dynamics, or plug in your Les Paul and chug away. 🙂
Attachments
Last edited:
Why not just filter out and block any frequencies above the normal extended audio range for the human ear?
Well, I put a RC filter on the input anyway to roll off ultrasonic frequencies somewhat, because of that infamous TIM/slew rate limiting business.
But clipping is a separate issue. The clipping behaviour would be as nasty with 19kHz as it was with 21kHz. Self argues that clipping behaviour is irrelevant because a hi-fi amp shouldn't be driven into clipping anyway, but in practice it probably will now and again, so it should handle it gracefully if the design allows that without too much trouble.
But clipping is a separate issue. The clipping behaviour would be as nasty with 19kHz as it was with 21kHz. Self argues that clipping behaviour is irrelevant because a hi-fi amp shouldn't be driven into clipping anyway, but in practice it probably will now and again, so it should handle it gracefully if the design allows that without too much trouble.
There has been discussion of this type of clamp in the DX blame MkIII thread. From what I remember it helped but there was still a little rail sticking.
However, I used it in my prototype CFA and in real life it works well with no visible rail sticking.
However, I used it in my prototype CFA and in real life it works well with no visible rail sticking.
Yes, I can still see a tiny "pip" on the positive rail but the negative one is a heck of a lot better than it was before. 🙂
A 1pF capacitor added in parallel to a 100pF capacitor should not affect the distortion of the amplifier.
I suspect the slight increase in distortion you have measured is due to some other mechanism.
Leakage current through the diode may be a plausible cause and explain the small change. Does LT spice show an effect if the leakage current is increased?
I suspect the slight increase in distortion you have measured is due to some other mechanism.
Leakage current through the diode may be a plausible cause and explain the small change. Does LT spice show an effect if the leakage current is increased?
Not only did I ( others ,too) experience the "bad" clip - "sticking" , but also a bit of ringing as the high gain pair recovered from saturation.
The BAV series of diodes are a good choice. The low capacitance/reverse leakage is the key to no increased distortion.
This post - http://www.diyaudio.com/forums/solid-state/192431-diyab-amp-honey-badger-20.html#post3627597
is where I resolved this behavior.
This post is the real amp doing the real thing 😀 ...
http://www.diyaudio.com/forums/solid-state/211905-diyab-amp-honey-badger-build-thread-12.html#post3625548
VERY ugly , some will overdrive their amp.
The diode does the trick. 🙂
OS
The BAV series of diodes are a good choice. The low capacitance/reverse leakage is the key to no increased distortion.
This post - http://www.diyaudio.com/forums/solid-state/192431-diyab-amp-honey-badger-20.html#post3627597
is where I resolved this behavior.
This post is the real amp doing the real thing 😀 ...
http://www.diyaudio.com/forums/solid-state/211905-diyab-amp-honey-badger-build-thread-12.html#post3625548
VERY ugly , some will overdrive their amp.
The diode does the trick. 🙂
OS
Last edited:
Definitely worth the extra component.
^As above.
Clipping can be really ugly with that sort of VAS - total saturation.
^As above.
Clipping can be really ugly with that sort of VAS - total saturation.
fwiw a quick napkin calculation of the relationship between the reference sensitivity of the typical home speaker (<90dB/1w/1m) and the typical average power level required to drive said system to >90dB SPL at the listening position, WRT the dynamic range of *better* quality recordings, or DIY recordings with >20db above average level dynamic range (to peaks) shows that one is very likely to clip peaks on an amp of 200wrms or less. In my experience that is the case.
Looking back at OStripper's thread, the Stochino amp was mentioned... whoa! A lot of diodes in that one! 😀
http://www.diyaudio.com/forums/attachments/solid-state/370514d1378876280-diyab-amp-honey-badger-build-thread-pa_stochino-2011-06-21-sch.pdf
Looking back at OStripper's thread, the Stochino amp was mentioned... whoa! A lot of diodes in that one! 😀
http://www.diyaudio.com/forums/attachments/solid-state/370514d1378876280-diyab-amp-honey-badger-build-thread-pa_stochino-2011-06-21-sch.pdf
AndrewT: 0.005% at 20kHz is getting on for the limits of my measuring equipment so I couldn't say conclusively that the distortion got worse on adding the diode. I'm just happy that it didn't get obviously worse. 🙂
Anyway, the issue is that the 1pF of capacitance is not constant, it depends on the instantaneous voltage across the diode, and at high frequencies the global feedback is rolled off and the feedback is mostly local through Cdom, of which the diode now constitutes 1%. There are plenty of other nonlinearities to worry about in the Lin/Blameless circuit though!
Leakage current could be an issue, as the BAV20 can leak up to 1uA when it gets hot, but I think of this as a DC phenomenon and there is lots of global feedback at DC. It might cause a slight change in the DC offset.
Bear: It was you going on about clipping that reminded me to try this. 🙂
Anyway, the issue is that the 1pF of capacitance is not constant, it depends on the instantaneous voltage across the diode, and at high frequencies the global feedback is rolled off and the feedback is mostly local through Cdom, of which the diode now constitutes 1%. There are plenty of other nonlinearities to worry about in the Lin/Blameless circuit though!
Leakage current could be an issue, as the BAV20 can leak up to 1uA when it gets hot, but I think of this as a DC phenomenon and there is lots of global feedback at DC. It might cause a slight change in the DC offset.
Bear: It was you going on about clipping that reminded me to try this. 🙂
Last edited:
- Status
- Not open for further replies.