Eva said:
anatech said:
Hi EVA & CHRIS,
I think you both misunderstood the schematic.....
ONLY THE SUPPLY RAILS OF THE OUTPUT DEVICES ARE SWITCHED NOT THE WHOLE DAMN AMP WITH FRONTEND IS SWITCHED....
IN the drawing the "AMP" symbol denotes the output stage ONLY not the whole amp......
EVA->
I have compared my switching circuit with QSC PLX 3402& Crest PRO9200 amplifiers which were Class-H.....
And I am happy to say that this circuit performs much better than these amps....
Much Less Switching Noise due to Schottky diodes...Soft recovery characteristics.....and better HF response ...
And also very simple!
If you have any excellent suggesstions then its great to see an improvement in this design.....
regards,
K a n w a r
Hi Workhorse
The typical “switching noise” of class-H we see on scope, it is maybe not only diode switch noise…
Think about the instantaneous beta variation of the poor output transistors submitted to brusque 50V variation in their VCE !
And about the NFB reaction to compensate for these…
I have seen the photo-coupler driven switch implemented on Camco and RAM amplifiers and the result is near the same of the typical class-H amps.
Alias, they are little schematic difference between Camco and QSC-PLX
except that Camco substitutes the comparator on the schematic you have posted (post-94) by one Photo-coupler. The “noise” on both amplifiers is very similar...
The typical “switching noise” of class-H we see on scope, it is maybe not only diode switch noise…
Think about the instantaneous beta variation of the poor output transistors submitted to brusque 50V variation in their VCE !
And about the NFB reaction to compensate for these…
I have seen the photo-coupler driven switch implemented on Camco and RAM amplifiers and the result is near the same of the typical class-H amps.
Alias, they are little schematic difference between Camco and QSC-PLX
except that Camco substitutes the comparator on the schematic you have posted (post-94) by one Photo-coupler. The “noise” on both amplifiers is very similar...
Hi Toino,TOINO said:
No, I have used both Schottky and regular Fast recovery Diodes....and the result is very much better in case with Schottkys..much lower noise due to their soft recovery....
TOINO said:
Yeah I know The "EARLY EFFECT" due to instantaneous change in Vds Mosfets in output in my case....But it has very little effect in comparision with vertical mosfet at outputs....as these are not bipolar but unipolar in conduction nature..
TOINO said:
I have seen the photo-coupler driven switch implemented on Camco and RAM amplifiers and the result is near the same of the typical class-H amps.
Alias, they are little schematic difference between Camco and QSC-PLX
CAMCO and Ram Audio amplifier uses Simple Optocoupler , Whereas I used an Integerated Photocoupled Gate Driver from Toshiba TLP250 ..which is much simpler to drive...
QSC and Crest uses Comparator+PP driver to drive the switch....
Noise maybe similar but not compared to my version has low SW noise which inturns has a definative edge over other PREVAILING Class-H designs......
If you Drive any of these amp with 20KHZ Sine wave in to 2 ohm full load and Observe the output on Scope.....you will be amazed to see the Tier Transitional Glitches and hand full of spikes......
And Niether of them uses any Schottkys...
Cheers,
K a n w a r
Hi Workhorse
The effect I am talking about is not only the "EARLY EFFECT" but the gain variation in the transistor because of the VCE variation.
It is why the cascade circuit is so linear: they are no VCE variations. The Carver circuits that Eva and Anatech talking about, is a kind of power cascade with a switch on the top. Anyway it produces lots of switching noise very visible on scope. Have discovered, when repairing one, that the internal Supply wires has a big influence on noise; twisting some of this wires attenuates the noise. Decoupling also helps a lot. Maybe part of noise is instantaneous EMC??
But I realise that your circuit is a different story because it is Mosfet output. Must confess that never see any commercial mosfet class-H.
The RAM uses also the Photo-driver HCPL3120.
Please, when you have measurements and scope photos, share with us.
I am curious about your mosfet class-H..
The effect I am talking about is not only the "EARLY EFFECT" but the gain variation in the transistor because of the VCE variation.
It is why the cascade circuit is so linear: they are no VCE variations. The Carver circuits that Eva and Anatech talking about, is a kind of power cascade with a switch on the top. Anyway it produces lots of switching noise very visible on scope. Have discovered, when repairing one, that the internal Supply wires has a big influence on noise; twisting some of this wires attenuates the noise. Decoupling also helps a lot. Maybe part of noise is instantaneous EMC??
But I realise that your circuit is a different story because it is Mosfet output. Must confess that never see any commercial mosfet class-H.
The RAM uses also the Photo-driver HCPL3120.
Please, when you have measurements and scope photos, share with us.
I am curious about your mosfet class-H..
Hi TOINO,
If you look at the later Carver designs where the lower two supplies track the audio signal, you will not see glitches.
The earlier ones that just swtched full on and off, yes. You could hear that.
Kanwar,
The Lightstar is another approach. The 125 V supplies are continously varied from around 12.5 VDC all the way up to full. PWM at high frequencies and filtering. Efficient and no glitches at all. His best sounding amplifier to that time.
-Chris
If you look at the later Carver designs where the lower two supplies track the audio signal, you will not see glitches.
The earlier ones that just swtched full on and off, yes. You could hear that.
Kanwar,
The Lightstar is another approach. The 125 V supplies are continously varied from around 12.5 VDC all the way up to full. PWM at high frequencies and filtering. Efficient and no glitches at all. His best sounding amplifier to that time.
-Chris
When the rails are switched at a controlled slope of say 15V/us (the maximum slope of a 240Vpp 20Khz sine) diodes have plenty of time to recover and no glitches are produced. Also, the amount of EMI radiated is ridiculously small.
I have used that approach to apply low frequency PWM signals directly to motor windings and big heaters while switching over conducting diodes without any EMI filter.
That capture shows a 120V rail being dropped softly to ground at 15V/us:
I have used that approach to apply low frequency PWM signals directly to motor windings and big heaters while switching over conducting diodes without any EMI filter.
That capture shows a 120V rail being dropped softly to ground at 15V/us:
An externally hosted image should be here but it was not working when we last tested it.
Hi Anatech
As far as I remember, you are right about the glitches on Carver. They are class-G and class-H mixed.
Lightstar is not the only one working with tracking PSU. I have repaired LAB gruppen (Martin audio OEM version) who works in the same technique. It is amazing the power it delivers for such a compact circuitry! And the power supply is Fly-back regulated 4Kw (maybe good for EVA investigateļ).
I remember also one Yamaha P7000 but it works at much lower frequency (very big coil) with a self oscillating tracking PSU driven by the amp current consumption.
Sure the Lightstar (never see one) is different kind of technology. I have analysed the Bob patent and it linearises the tracking PSU sensing the output filter capacitor current. It is curious that the quality of the Tracking PSU must be the some or better than the analogue amplifier; otherwise it doesn¡¦t work properly. I am tented to say that in a good design we could take of the analogue amplifier because the PSU is as much as necessary to do the amplifierƒº
On my efforts to understand this subject of tracking PSU I have found ¡§Thousands¡¨ of patents:
4430625
4472687
4507619
5075634
5200711
5347230
5396194
5450037
5510753
5543753
5606289
5825248
5834977
5886506
5937074
5990751
6031746
6091292
Must of them are the some. Someone must be jocking!
As far as I remember, you are right about the glitches on Carver. They are class-G and class-H mixed.
Lightstar is not the only one working with tracking PSU. I have repaired LAB gruppen (Martin audio OEM version) who works in the same technique. It is amazing the power it delivers for such a compact circuitry! And the power supply is Fly-back regulated 4Kw (maybe good for EVA investigateļ).
I remember also one Yamaha P7000 but it works at much lower frequency (very big coil) with a self oscillating tracking PSU driven by the amp current consumption.
Sure the Lightstar (never see one) is different kind of technology. I have analysed the Bob patent and it linearises the tracking PSU sensing the output filter capacitor current. It is curious that the quality of the Tracking PSU must be the some or better than the analogue amplifier; otherwise it doesn¡¦t work properly. I am tented to say that in a good design we could take of the analogue amplifier because the PSU is as much as necessary to do the amplifierƒº
On my efforts to understand this subject of tracking PSU I have found ¡§Thousands¡¨ of patents:
4430625
4472687
4507619
5075634
5200711
5347230
5396194
5450037
5510753
5543753
5606289
5825248
5834977
5886506
5937074
5990751
6031746
6091292
Must of them are the some. Someone must be jocking!
The oscilloscope is a Hameg HM-407 and the (slow!!) software that comes with it may be installed on a PC to allow high resolution screen captures to be transferred via RS-232 and pasted on your favourite picture editing software. If you search among my posts, you will see dozens of these captures.
The joke is the US patent system. They will approve anything if you put enough numbers on the check
Concerning tracking power supplies, they have to be almost as good as full range class D amplifiers, but there is a little problem with that because the output filter introduces a delay in the passband and there is no way to eliminate it, and sensing capacitor current allows to get better phase margin near (or even above) filter resonance but does not solve the delay either. Don't ask me how I know all that (I'm currently listening to a class-D prototype that plays well up to 10Khz without load dependent peaking, despite its output filter resonating at 3.6Khz... It's not intended for audio, though, but it's a funny way to test it)
Concerning tracking power supplies, they have to be almost as good as full range class D amplifiers, but there is a little problem with that because the output filter introduces a delay in the passband and there is no way to eliminate it, and sensing capacitor current allows to get better phase margin near (or even above) filter resonance but does not solve the delay either. Don't ask me how I know all that (I'm currently listening to a class-D prototype that plays well up to 10Khz without load dependent peaking, despite its output filter resonating at 3.6Khz... It's not intended for audio, though, but it's a funny way to test it)
There are two nested control loops. One loop continuously adjusts duty cycle to force a certain current across the inductor. The other loops tells how much current has to flow through the inductor in order to obtain the desired output voltage. In these circumstances the LC system has no chance to resonate, well, actually you can see it resonating only when the amplifier clips and the control loop loses control over it (funny).
Theoretically, that produces 90 degree phase margin for all frequencies. In practice, it doesn't because the current amplifier starts to roll-off and shows increasing phase lag above Fs/(2*pi) (and my Fs gets as low as 28Khz so I have to use capacitor current as an extra aid to improve phase margin).
That is nothing new. It has been employed in SMPS control for ages, and there are a lot of expired patents on it
Theoretically, that produces 90 degree phase margin for all frequencies. In practice, it doesn't because the current amplifier starts to roll-off and shows increasing phase lag above Fs/(2*pi) (and my Fs gets as low as 28Khz so I have to use capacitor current as an extra aid to improve phase margin).
That is nothing new. It has been employed in SMPS control for ages, and there are a lot of expired patents on it
TOINO said:Hi Workhorse
The effect I am talking about is not only the "EARLY EFFECT" but the gain variation in the transistor because of the VCE variation.
It is why the cascade circuit is so linear: they are no VCE variations. The Carver circuits that Eva and Anatech talking about, is a kind of power cascade with a switch on the top. Anyway it produces lots of switching noise very visible on scope. Have discovered, when repairing one, that the internal Supply wires has a big influence on noise; twisting some of this wires attenuates the noise. Decoupling also helps a lot. Maybe part of noise is instantaneous EMC??
But I realise that your circuit is a different story because it is Mosfet output. Must confess that never see any commercial mosfet class-H.
The RAM uses also the Photo-driver HCPL3120.
Please, when you have measurements and scope photos, share with us.
I am curious about your mosfet class-H..
Hi Toino,
Yes you were right Cascoded series transistors are more linear, thats the reason why Class-G is very Linear than Class-H....But Class-G is also very much costly because it employs double the number of output devices required in Class-H....
I have seen Carvers many of them employ both Class-G/H simuntaneously and increases the complexities in the circuit also....
IMHO Gain variations due to instantaneous change in Vds of Mosfets have a little effect[assuming total gate charge to remain uniform/Ciss does vary]...but in bipolars yes its of a much significance..
I would definately post my results when i would finished this project....but still more work is required to be done....Its the first ever All Mosfet output/Switch amp in the world alone using vertical N-Channel devices....
I am now planning to go for a 3-Tier +-50V, +-100V , +-150V version for obtaining 2000W@2 Ohms with just 5 pairs of IRFP360 Mosfets....
Hi Chris,anatech said:
The Level of complexity in Lightstar is very high and I think nowadays one can make a good Class-D amp much better than opting for Tracking power supplies rails....And in my case i have succeeded in that topology as well....the Class-D
BTW: How about a different section must be enlisted for "Pro-Audio" user segment apart from Solid state....in this forum...
I have both Class-D and Class-H/G in my hands...but you know I LOVE to experiment and explore more...
Cheers,
K a n w a r
Class G is not more linear. The drive current required from the VAS is doubled when the upper half of the output stage is working, thus introducing a strong odd harmonic content in the open-loop THD figures of the amplifier.
In comparison, class H only produces momentary glitches and gain changes in the range of say +-20% (while in class G gain changes by 100%!!)
In comparison, class H only produces momentary glitches and gain changes in the range of say +-20% (while in class G gain changes by 100%!!)
Eva said:
Its certainly not true Eva,
I have seen several Class-G designs in which upper Tier Devices are powered from the output of lower Tier devices...and not from the Vas...
Secondly Class-G is More linear because the upper Tier Devices Actually conduct in Linear fashion in accordance with the signal and also forms a cascode...whereas in Class-H the upper Tier switches the whole rail voltage just like a Square Wave type...
Class-G is costly whereas Class-H is a very low cost solution....
You can't reliably drive the upper side of a class G output stage from the output of the own amplifier. It will either oscillate (locally or globally) or require tremendous amounts of frequency compensation for stability (and very low feedback), as the main pole comes from the output devices and it would be applied twice!! (despite miller VAS capacitors are used to mask that pole most of the time).
Also, class H is not a low cost solution because in a proper implementation with controlled slopes the "switching" devices may be subject to considerable dissipation sometimes and should be capable of handling it.
Also, class H is not a low cost solution because in a proper implementation with controlled slopes the "switching" devices may be subject to considerable dissipation sometimes and should be capable of handling it.
Hi Kanwar,
Circuit complexity is relative with regard to the Lightstar. That amp has complex supplies, and amplifier stages! I am sure the down converters are very worthwhile as you still have a linear output stage. Eva could probably design the down converters easily.
-Chris
Circuit complexity is relative with regard to the Lightstar. That amp has complex supplies, and amplifier stages! I am sure the down converters are very worthwhile as you still have a linear output stage. Eva could probably design the down converters easily.
New software will bring some changes to the forum. Nothing right now.
-Chris