Reinventing the N-channel wheel? - Page 5 - 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 16th February 2006, 09:33 PM   #41
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
Quote:
Originally posted by Ultima Thule


No, it's just the strong neonic collors in here schematic that makes it complex!

Eva,

OMG, what are you doing..
Ic Q17, 18 doesn't look very funny..

I have thought myself a lot regarding what I call "dynamic-class-A" (you call it pseudo-class-A, don't know what is the correct name for it), I believe the transistion from the sustain to active region must be made softer to prevent such current jumps.

Cheers Michael

I'm just experimenting, measuring, playing and enjoying

For example, I have been looking at VAS emitter current waveforms (in the real prototype) with Q18 only used as a driver, and with Q42 as a driver cascoded with Q18 (Vce=5V). The latter case requires a more sine-like VAS current waveform to produce a sine amplifier output, thus it's clearly more linear and it wins.

I have been also looking at LTP collector current waveforms (also in the real prototype) with a single transistor VAS (Q15) and with a cascoded VAS (Q15 and Q40) with Vce=5V. Here, the difference is even more evident, the "gain" of the non-cascoded VAS is clearly modulated by Vce, while the cascoded VAS has a higher gain and requires a much more sine-like drive current in order to produce sine output.

So I have adopted both cascodes.

Ic of Q17 and Q18 just reflects the extreme non-linearity of MOSFETs (even considering that that my PSpice simulates constant capacitances). There are essentially two ways to linearize the amplifier, one is throwing the damn MOSFETs in some SMPS where they should have been always sitting and use some Sanken or Toshiba bipolars, and the other is finding a circuit having a non-linearity complimentary to the quadratic law of MOSFETs I could produce a more smooth transition, but it would be only cosmetic, because the transfer function of the MOSFETs would be still here and the harmonics generated would be lower (currently they are essentially above 10Khz).

Also, with bipolars the transition from the sustain region to the active region is much less troublesome and may be sharp because they behave more or less like constant current gain devices. I could drive the MOSFETs with symmetric gate waveforms around a center bias voltage (like I did in the schematic in page 1) but then there is a big issue: Vgs threshold not only changes with temperature but also with Vds!! This means that the bias is only constant when the zero-current transition happens at 0 volts. When driving reactive loads, there will be both excessive cross-conduction and crossover distortion due to a lack of bias depending on the voltage level where the zero-current transition happens (this is one of the reasons why people has to use a dozen TO-247 MOSFETs to make just a 500W amplifier, and that's something that I'm trying to avoid).
  Reply With Quote
Old 16th February 2006, 10:58 PM   #42
dimitri is offline dimitri  United States
diyAudio Member
 
dimitri's Avatar
 
Join Date: Oct 2002
Location: retired
Send a message via ICQ to dimitri
hey eva, could we continue in private?
could you pls contact me?
Attached Images
File Type: gif mail.gif (1.0 KB, 748 views)
  Reply With Quote
Old 16th February 2006, 11:09 PM   #43
diyAudio Member
 
Join Date: Jun 2004
Location: HKSAR
Hi eva,

Could you kindly post your schematic 'black on white', It's awfully difficult for old chaps like me to read it 'color on black'

Thanks a lot
  Reply With Quote
Old 18th February 2006, 11:54 AM   #44
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
dimitri:

I have mailed you, but I got no reply


bigpanda:

Is this easier to read?
Click the image to open in full size.

This is a fixture of my amplifier without global negative feedback. The signal coming from the output is bypassed with 100uF to ground that in turn forms a 0.16Hz low-pass filter with the 10K resistor. Only output DC level is corrected. The gain of the input LTP has been reduced by increasing the value of the degeneration resistors to 560 ohms, and the same has been done to the VAS forcing it to show unity current gain by increasing C8 to 470uF and loading its output with 22K to ground. With these modifications, an open loop gain similar to the closed loop value (30dB) is achieved.

This is how the actual prototype (configured as the previous schematic shows) puts a 10Khz sine wave at almost full power into a 2.5 ohm load (approx 70W). CH1 is input and CH2 is output. Note the ridiculous phase shift and the absence of visible distortion (PSpice predicts 0.6%):
Click the image to open in full size.

When zooming with the oscilloscope, some mild crossover distortion may be observed, but there are not any spikes, even with bias levels as low as 10mA. The simple comparison of both waveforms also shows traces of some 2nd harmonic (asymmetry). I have measured the output impedance to be between 0.26 and 0.27 ohms, dominated by the 0.25ohm current sense resistors (and also the own lousy contacts of the breadboard), so the output stage itself does an excelent job by converting the 22K input impedance from the LTP to a few miliohms while taming all the MOSFET non-linearity (I'm currently using a pair of IRF540).

I'm considering placing some jumpers in the final PCB to allow to quickly change between an amplifier with 140dB open loop gain at 1Khz and an amplifier with 30dB of gain and no global feedback

I have to admint that I find quite difficult to distinguish both configurations by ear (but it has been demonstrated that most people will fail 50% of the times in a serious A/B test anyway ). The non-global-feedback version may interact a bit with the speaker (single 6" driver in a sealed box) by slightly attenuating midrange, but this is due to the 0.25 ohm output impedance.

p.s.: I have been too lazy to test the pseudo-class-A version with real components, but the non-global-feedback version was just much simpler to try
  Reply With Quote
Old 18th February 2006, 12:34 PM   #45
dimitri is offline dimitri  United States
diyAudio Member
 
dimitri's Avatar
 
Join Date: Oct 2002
Location: retired
Send a message via ICQ to dimitri
eva, pls check your mail

Quote:
with 140dB open loop gain at 1Khz and an amplifier with 30dB of gain
so you lowered the gain by 110 db, lets compose a gain budget
- lowered transconductance of the first stage from 68 Ohm to 560 Ohm -20dB
- lowered VAS load from 1M to 22K -35dB
- before the input stage load was 68*hFE, now it is 68, assuming hFE=200 -46dB
-20-35-46=-100dB

ok, close by to 110dB

the input stage gain is -20dB, do you need it at all?
  Reply With Quote
Old 19th February 2006, 02:31 AM   #46
diyAudio Member
 
Workhorse's Avatar
 
Join Date: Dec 2004
Hi EVA,

Your circuit looks over complicated...very much..
Here is mine...a simple one

regards,
K a n w a r
Attached Images
File Type: jpg ampdiy.jpg (44.4 KB, 821 views)
  Reply With Quote
Old 19th February 2006, 08:39 AM   #47
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
As SKA, your VAS-less circuit relies somewhat on constant gate capacitances (an additional gain stage makes it far better than SKA) and very strongly on constant Vgs/Id characteristics of the output devices across the entire output voltage swing for proper dynamic bias setting. But these characteristics are far from constant (simulation won't show it but real life does). Also, your circuit relies strongly on proper thermal feedback from the heatsink for bias setting, but you don't show how this is achieved in your schematic. That's a complex task that will require a dozen or so of components, thus making it less simple.

Also, your circuit is conceptual because it neither shows transistor models, nor component values, nor the required frequency compensation networks in the gate drive cells, nor input stage cascoding, nor the required thermal feedback from the heatsink, while mine reflects a true working and ready-to-build prototype. This makes me wonder seriously whether you really have the thing working or you are just dealing with something purely conceptual, without any analysis of real prototype behaviour. Could you post pictures of something capable of producing sound without smoke? Could you post oscilloscope captures showing amplifier output at high frequencies and current waveforms from each output device showing no cross-conduction?

On the other hand, my circuit does not rely at all on transfer characteristics of output devices thus it's much more robust and fault-tolerant:

- It even works fine with a IRF540 100V MOSFET in one of the rails and a SKP10N60 600V IGBT in the other. Since each output device will have a gate-drive cell, no matching is required. You toss random devices on the board and the circuit works fine.

- There is no need to readjust the bias when output devices are changed, the circuit is able to keep the bias constant independently of gate thresholds and without any thermal feedback from the heatsink thanks to the current feedback output stage. You toss random devices on the board and the bias is still kept at the same value it was adjusted last time.

- As my previous post shows, it produces low distortion when operating without any global feedback, because the "complex" output stage linearizes the output MOSFETs/IGBTs/whatever to such an extent that it makes them appear to the front-end as ideal adjustable current sources. You toss random devices on the board, and the real circuit still produces less than 1% distortion and sounds fine with all the 90dB of global feedback removed (you can see the oscilloscope capture in my previous post).

Your circuit is simple, but it will misbehave or destroy itself in any of the previous circumstances. Simplicity is not always a good thing in PA amplifiers (except when it allows you to make more money by selling an unreliable thing claiming it to be reliable).

Have you tried it driving 10Khz full power into 2.5 ohms without any global feedback? I bet your output waveform is going to resemble very little to the input sine
  Reply With Quote
Old 19th February 2006, 02:10 PM   #48
diyAudio Moderator Emeritus
 
lumanauw's Avatar
 
Join Date: Oct 2002
Location: Bandung
Send a message via Yahoo to lumanauw
Hi, EVA,

Nice design (although your colorfull cct is hard to see )
From your last schematic, you seems to make EC for each side. There are 3 loops in the whole amp. 1 loop is the main loop, by differential input pair. Upper side is controlled by CF loop by Q17. Lower loop by another CF loop by Q42, with folded cascode by Q22.

I want to ask about the final stage. If 1N4148 and BC's are having different B-E drop, would this force 1 side will have to saturate?

Why don't you make something like this?

How does it sounds?
Attached Images
File Type: jpg rf2.jpg (23.0 KB, 667 views)
  Reply With Quote
Old 19th February 2006, 02:30 PM   #49
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
Yes, you are right about the three amplifiers in one

At the beginning I also had doubts about the totem pole drivers biased with just two diodes. However, my pairs of 1N4148 biased to 3.3mA and my BC546B/BC556B combo resulted in a nice match because the transistors have a slightly higher Vbe than the diodes, so they always end up biased with just 100uA to 500uA, at the threshold of class-AB for small signal devices and with perfect thermal compensation.

I tried 1N4148 from two manufacturers and got the same results. The transistors are all On-Semi, though. I could insert a small resistor between the emitters just to be in the safe side, but I have very little space in the breadboard. I may put it in the PCB.

All my 1N4148s measure approx 630mV for 1mA, while the transistors show approx 680mV Vbe for 1mA Ic (at the same temperature).

Also, if you take a look at the datasheets of the transistors, you will see that the Vbe of the BC546B increases almost 80mV from Ic=1mA to Ic=10mA and the Vbe of the BC556B increases 50mV (for Vce=5V). That in fact works as if there were 8 and 5 ohm emitter resistors respectively, giving a great margin for diode tolerance and ensuring that excessive cross-conduction will hardly happen. In fact, I'm considering adding a small resistor in series with the diodes to intentionally increase cross-conduction and get better Ft and hFE.

To me it doesn't sound, it just amplifies the input waveform (Sound is a loudspeaker thing).
  Reply With Quote
Old 19th February 2006, 03:03 PM   #50
diyAudio Moderator Emeritus
 
lumanauw's Avatar
 
Join Date: Oct 2002
Location: Bandung
Send a message via Yahoo to lumanauw
Quote:
In fact, I'm considering adding a small resistor in series with the diodes to intentionally increase cross-conduction and get better Ft and hFE.
How about using 3 diode drop, and small R for each emitors? This will bias the BC's all the time.
In my experience, VBE of 1 kind of transistor will vary much. I've found difference of 0V61 to 0V67 for the same transistor type from the same manufacturer.

One more question. Why you use so much "soft clipping" cct ? Like Q12, Q45, Q46, D23 (not clear, upper left of D13)?
  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
Geared wheel for Phillips CD850-MK2 Maurice Digital Source 112 2nd November 2011 05:50 PM
Ferris Wheel Project mattthegamer463 Everything Else 14 31st May 2009 08:37 PM
New kind of feedback or just re-inventing the wheel? Circlotron Solid State 51 9th March 2003 08:02 AM


New To Site? Need Help?

All times are GMT. The time now is 04:48 AM.


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