A small "riddle"... - diyAudio
Go Back   Home > Forums > Amplifiers > Solid State

Solid State Talk all about solid state amplification.

Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Reply
 
Thread Tools Search this Thread
Old 18th February 2006, 11:40 AM   #1
MikeB is offline MikeB  Germany
diyAudio Member
 
MikeB's Avatar
 
Join Date: Apr 2004
Location: GŁtersloh
Cool A small "riddle"...

I was researching a bit and stumbled over that...
Attached is a test circuit with 2 configurations, the left has the cascode in input-ltp referenced to the emitters of the input-ltp, the right has this cascode referenced to gnd. The left circuit shows a THD of 0.00003%, the right one has 0.03%. The difference is factor 1000 ! How is that possible ?

Input was 2v 20khz, output 4v. Both have OLG of 3800 at 20khz. Without the cascode THD is 0.007%.

Mike
Attached Images
File Type: png riddle1.png (6.9 KB, 977 views)
  Reply With Quote
Old 18th February 2006, 12:09 PM   #2
diyAudio Member
 
jan.didden's Avatar
 
Join Date: May 2002
Location: Great City of Turnhout, Belgium
Blog Entries: 7
Mike,

Probably it is a simulator artifact. I have had cases that output signal went down from several volts to just a few millivolts when referencing the supply of a circuit to some output (bootstrapping the supply). In very similar cases is worked OK. What sim do you use, Pspice?
In any case, the HD figures of a sim are notoriously inconsistent anyway. What are the differences with an FFT?

Jan Didden
__________________
I won't make the tactical error to try to dislodge with rational arguments a conviction that is beyond reason - Daniel Dennett
Check out Linear Audio Vol 7!
  Reply With Quote
Old 18th February 2006, 12:22 PM   #3
MikeB is offline MikeB  Germany
diyAudio Member
 
MikeB's Avatar
 
Join Date: Apr 2004
Location: GŁtersloh
Hi Jan, yes, i use PSpice. I do not use the THD-function of pspice, it's way to unreliable. I always look to the FFT. It seems that the effect is real, it is just caused by an effect not expected in this scale, making the nfb inefficient.

Mike
  Reply With Quote
Old 18th February 2006, 12:57 PM   #4
Eva is offline Eva  Spain
diyAudio Member
 
Eva's Avatar
 
Join Date: Oct 2003
Location: Near the sea
Send a message via MSN to Eva
I have been experimenting with cascodes, current sources and LTPs in real circuits with real transistors and I have been able to observe that kind of effects, altough not so strong.

All the trouble comes from some assumptions that we make when designing circuits but that are not true with real components. Three of these assumptions are: Constant Vbe, constant hFE and constant capacitances.

For example, I have been able to measure Vbe changes of a few milivolts associated with Vce changes of a few volts. High gain circuits are evil because they amplify also all these unwanted erratic signals due to Vbe, hFE and Cbc modulation. That's why keeping Vce constant, or at least, keeping the AC component of Vce small and strictly proportional to the audio signal, reduces distortion not only in simulation but also in real life (altough not to 0.0000003% ).

Both in my LTP and its associated current source I was able to measure Ic modulation due to the Vce sweep caused by the input signal and the power supply ripple voltage. Even after taming most of the error component coming from the PSU and isolating the error component coming from the input signal voltage sweep, the high content of 2nd harmonic was still quite evident at first sight (even with the oscilloscope in the 1mV/div range and picking up a lot of RF noise that blurred the waveform).


Anyway, the worst modulation by far is the one coming from power supply ripple (the one from real power supplies, not from PSpice voltage sources ). I have found really really difficult to reduce it, but this is the key to good PSRR and "clean" circuit behaviour.
  Reply With Quote
Old 18th February 2006, 01:07 PM   #5
diyAudio Member
 
Tube_Dude's Avatar
 
Join Date: Mar 2002
Location: Aveiro-Portugal
Default Re: A small "riddle"...

Quote:
Originally posted by MikeB
Attached is a test circuit with 2 configurations, the left has the cascode in input-ltp referenced to the emitters of the input-ltp, the right has this cascode referenced to gnd. The left circuit shows a THD of 0.00003%, the right one has 0.03%. The difference is factor 1000 ! How is that possible ?

Input was 2v 20khz, output 4v. Both have OLG of 3800 at 20khz. Without the cascode THD is 0.007%.

The left input LTP is effectively bootstrapped , this decrease the Miller capacitance of the LTP input transistors , and remember that you are measuring distortion at 20 KHz.

Certainly at 1KHz , the difference in distortion will not be so dramatic.
__________________
Jorge
  Reply With Quote
Old 18th February 2006, 01:13 PM   #6
Eva is offline Eva  Spain
diyAudio Member
 
Eva's Avatar
 
Join Date: Oct 2003
Location: Near the sea
Send a message via MSN to Eva
I did most of my measurements at 1Khz because at 10Khz supply ripple caused by tone bursts was too small. At 10Khz and above capacitances are expected to have greater impact than Vbe and hFE modulation, but the latter effects are still present.
  Reply With Quote
Old 18th February 2006, 01:28 PM   #7
MikeB is offline MikeB  Germany
diyAudio Member
 
MikeB's Avatar
 
Join Date: Apr 2004
Location: GŁtersloh
Yes, that's right, this effect is frequency dependent and caused by Vce variations that are barely existing in the left version. Gladly pspice models the dynamic behaviour of bjts. It's frightening how easily a circuit can be mis designed causing "horribly high" distorted trebles.

BUT, it is still fascinating where the problem occurs, obviously the openloop distortion can't be increased by factor 1000 by that. (that would be 57% )

Mike
  Reply With Quote
Old 18th February 2006, 03:21 PM   #8
diyAudio Moderator Emeritus
 
lumanauw's Avatar
 
Join Date: Oct 2002
Location: Bandung
Send a message via Yahoo to lumanauw
Hi, Mike,

How are you

I really amazed by NP's patent 4,107,619. It is dated 1978, look at the drawing at the first page.

What amazed me more, is that it seems NP knows distortions that can be eliminated by feedback, and which ones cannot. So he cleverly advoids the distortion cause that cannot be eliminated by feedback. Distortions caused by non-linear internal transistor capacitance, for example, cannot be eliminated by feedback (I think....)
  Reply With Quote
Old 18th February 2006, 03:39 PM   #9
MikeB is offline MikeB  Germany
diyAudio Member
 
MikeB's Avatar
 
Join Date: Apr 2004
Location: GŁtersloh
Yes, in fact this is a distortion that can not be compensated by feedback. Applying more feedback even worse things as it does not reduce anything, but the increased openloop gain will amplify these effects. I believe this is the trap where many made the experience that more feedback sounds worse (except the stability thing).

Why ?

The distortion is created outside the feedback loop...
The signal is already distorted at the input of the amp, there is no way to compensate THAT with feedback.
This is caused by dynamic miller capacitance, creating a reactive input distorting the signal BEFORE it enters the amp. The same thing happens at the feedback node, resulting in non functional feedback plus the distortions generated at input. The amp itself works correctly, but gets fed with defect signals. That should be one reason why jfets sounds different, they create different distortions.

The simplest way out is to increase supply voltage, reduce swing at input and use good bjts. Or use correct working cascodes, they can nearly kill this effect. (reduce it by factor 1000) These cascodes make the amp also tolerant to high impedance source.
This cascoding can be simply done with jfets around the input devices.

Mike
Attached Images
File Type: png riddle2.png (5.2 KB, 722 views)
  Reply With Quote
Old 18th February 2006, 03:43 PM   #10
Eva is offline Eva  Spain
diyAudio Member
 
Eva's Avatar
 
Join Date: Oct 2003
Location: Near the sea
Send a message via MSN to Eva
Distortions that taint the error signal calculation (the difference between the actual output and the desired value) cannot be eliminated by feedback.

In other words: Input differential stage should be as linear as possible (and should feature high PSRR). Once the error signal is precisely calculated, high open loop gain will do the rest (provided that a reliable path for the amplified error signal to reach the output exists).

EDIT: I have just fallen in love with these JFET cascodes!
  Reply With Quote

Reply


Hide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
Need your opinion, how small gain can I use with LM3886 (how close to "flat amp")? danielwritesbac Chip Amps 20 30th November 2008 12:27 AM
Riddle: Drawing: Horn ca. 1974 8" lowther/coral h 100 x w 40 x d 25 cm speaker up KlausenDK Full Range 0 15th January 2008 02:05 PM
looking for a small 2" or 3" driver with poor bass response space-cake Full Range 3 27th December 2007 01:16 AM
suspending a small tweeter in front of woofer for "coaxial" effect? bikehorn Multi-Way 3 2nd January 2006 02:04 PM
Thiele-Small R_ms (mechanical losses) vs. "precise" sound? capslock Multi-Way 7 16th September 2002 09:32 AM


New To Site? Need Help?

All times are GMT. The time now is 08:11 PM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright ©1999-2014 diyAudio

Content Relevant URLs by vBSEO 3.3.2