G.Kleinschmidt said:OK, I’ll stir the pot.
Here........
http://www.diyaudio.com/forums/showthread.php?postid=1344023#post1344023
.......is the partial schematic of an amp design of mine with the VAS voltage diode clamped with a circuit that tracks the rails, thus preventing the driver and output stages from saturating. I also like to keep the voltage headroom conservative to prevent fT droop at low Vce as this allows a higher unity gain loop frequency. The circuit also provides a clipping detect output.
The circuit also has a diode clamp (unlike D.Self’s I-limiting transistor) to limit the VAS current.
Attached below is the clipping performance at 100kHz. The input stage LTP’s are being driven well into cut-off here. The LTP’s however, are a very fast stage of the amplifier and the recovery from overdrive is measurable in nano seconds.
So, the question begs:
Why should we care if the input LTP(s) is(are) driven into cut-off during clipping?
I agree Glen. In most cases there is no need to worry about the input LTPs being driven into cutoff during clipping.
Cheers,
Bob
JPV said:
I have often asked the question, "have we left the realm of high fidelity if our amplifier is clipping?". The answer may not be as simple as one might think; indeed, clipping does happen. However, many of us do not have a good sense of how often it happens. Very few consumer amplifiers have clip indicators; I wish they did.
I also wish that clipping was more of a problem with today's music. Why, you ask? Unfortuantely, new recordings with minimal compression and terrific dynamic range are getting more scarce every day. It is a sad turn of events.
Cheers,
Bob
Bob Cordell said:
Indeed: http://www.rollingstone.com/news/story/17777619/the_death_of_high_fidelity
Jan Didden
JPV said:
I wish reviewers really knew whether or not their review amps were clipping during their listening tests.
My Athena active louspeakers have four 125-watt MOSFET power amplifiers in each channel. One behind the tweeter, one behind the midrange, and one behind each of two pairs of woofers. I have measured brief 30-volt peaks at the tweeter terminals on a Rickey lee Jones snare drum thwack.
Cheers,
Bob
Hi, Edmond,
Sorry about that
I just feel that this CFB clamp is very interesting concept, many people should see this brainchild of yours.
I don't quite understand it, I hope that the pencil drawing above can represent the core of your idea.
I feel that it can change the sound of transistor+feedback amps alot (accompanied with good output stage and small phase shift of the whole amp). I will try it with the real components soon. I still think about how to implement it for ordinary VFB amps. I'm thinking about your suggestions to move the sensor to pre-driver (instead of VAS), also thinking about injecting the error signal to emitor junction of the differential pair (instead to base of inverting input).
Sorry about that
I just feel that this CFB clamp is very interesting concept, many people should see this brainchild of yours. I don't quite understand it, I hope that the pencil drawing above can represent the core of your idea.
I feel that it can change the sound of transistor+feedback amps alot (accompanied with good output stage and small phase shift of the whole amp). I will try it with the real components soon. I still think about how to implement it for ordinary VFB amps. I'm thinking about your suggestions to move the sensor to pre-driver (instead of VAS), also thinking about injecting the error signal to emitor junction of the differential pair (instead to base of inverting input).
FCB clamp
Hi David,
I also think that many people should see this brainchild of me. Therefore it's so important that they see the original version first, the more so as most other variants are missing an essential feature.
At the moment, the schematic of the CFB clamp has vanished from my website for unclear reasons, but I will replace it the next day.
In the past, you asked me several relevant questions about this clamp. So, do you agree if I put them also on my website, in a kind of Q and A section?
Cheers, Edmond.
lumanauw said:Hi, Edmond,
Sorry about that
I don't quite understand it, I hope that the pencil drawing above can represent the core of your idea.
I feel that it can change the sound of transistor+feedback amps alot (accompanied with good output stage and small phase shift of the whole amp). I will try it with the real components soon. I still think about how to implement it for ordinary VFB amps. I'm thinking about your suggestions to move the sensor to pre-driver (instead of VAS), also thinking about injecting the error signal to emitor junction of the differential pair (instead to base of inverting input).
Hi David,
I also think that many people should see this brainchild of me. Therefore it's so important that they see the original version first, the more so as most other variants are missing an essential feature.
At the moment, the schematic of the CFB clamp has vanished from my website for unclear reasons, but I will replace it the next day.
In the past, you asked me several relevant questions about this clamp. So, do you agree if I put them also on my website, in a kind of Q and A section?
Cheers, Edmond.
Re: Re: Re: clamp
Well, I think that I have demonstrated that it works quite well.
But if you think that NFB clamping is not viable for a VFB amp and such "brute force" methods are not to your taste, then what alternative scheme would you suggest?
One thing that my rail-tracking clamp addresses that I mentioned in my first post, that hasn't been discussed and is relevant to the topic of controlling an amps clipping behaviour, is how close the output voltage is allowed to swing towards the rails.
The skyrocketing Cin of power MOSFET's and BJT's (especially the RET's below 5Vce) is something to be avoided.
How close to the rail's do you like to go?
Cheers,
Glen
Edmond Stuart said:
Well, I think that I have demonstrated that it works quite well.
But if you think that NFB clamping is not viable for a VFB amp and such "brute force" methods are not to your taste, then what alternative scheme would you suggest?
One thing that my rail-tracking clamp addresses that I mentioned in my first post, that hasn't been discussed and is relevant to the topic of controlling an amps clipping behaviour, is how close the output voltage is allowed to swing towards the rails.
The skyrocketing Cin of power MOSFET's and BJT's (especially the RET's below 5Vce) is something to be avoided.
How close to the rail's do you like to go?
Cheers,
Glen
Bob Cordell said:
Thank you for this information. I am designing 100 to 150 watts in thsi 4 ways system and you confirm the good choice. It is surprising that Linkwitz is recommending lower power ( 60 watts) beeing enough, we need enough current AND enough voltage in a transient way not average power alone.
On this topic, it is well known that the transistor physical equations are highly non linear and what we are forced to do is playing with linear circuit approximations. Then, to investigate the non linearities, we use circuit element variation with signal level ( Ft, C, Early voltage, transfer function, saturation, clipping ..) Depending on the element we analyze and the test signal used we create a name of the distortion(TIM, THD, PIM IIM CCITT SMPTE, ..) and we would like to come with a small set of single figures characterising correctly the aural effect and justifying the toplogy of the circuit. This is obviously oversimplification but the subject is complex.
I would suggest to use a transient test to investigate non linearities.
The swept signe wave impulse I described here above is very selective. It allows to generate out of band distortion spectral elements in a transient way.
The test signal is easy to generate in LTspice and you can then play with amplitude and frequency band.
A figure like the rapport of output out of band energy to output in band energy is a very good indication of non linearities and perhaps better related to aural effects than THD.
I have an excel file with the spectrum of the test signal but it is to large. I can mail it. The analitycal development ot the spectrum is existing but is very complex.
JPV
Re: Re: Re: Re: clamp
Hello Glen,
What's the trade-off here? For a given power amp, with a given supply, if you limit the min Vce (Vds) to 5V, that then means that the amp clips at the Vce=5V point. OTOH, you could accept that between Vce=5 and, say, Vce=2 (and corresponding Vds of course) distortion would rapisly rise, but you do have some additional headroom before clipping (when the distortion REALLY rises).
So, I guess what I'm saying is, do you think that it is so worthwhile to limit min Vce that clipping is to be preferred?
Jan Didden
G.Kleinschmidt said:
Hello Glen,
What's the trade-off here? For a given power amp, with a given supply, if you limit the min Vce (Vds) to 5V, that then means that the amp clips at the Vce=5V point. OTOH, you could accept that between Vce=5 and, say, Vce=2 (and corresponding Vds of course) distortion would rapisly rise, but you do have some additional headroom before clipping (when the distortion REALLY rises).
So, I guess what I'm saying is, do you think that it is so worthwhile to limit min Vce that clipping is to be preferred?
Jan Didden
Re: Re: Re: Re: Re: clamp
Hi Jan.
The trade-off is that if you let the output get really close to the rails, the amplifier has to be compensated more heavily to maintain unconditional stability when the amplifier clips (due to the bandwidth of the OPS suddenly going down the tube).
That means less HF loop gain and more HF THD at all power levels.
Cheers,
Glen
janneman said:
Hi Jan.
The trade-off is that if you let the output get really close to the rails, the amplifier has to be compensated more heavily to maintain unconditional stability when the amplifier clips (due to the bandwidth of the OPS suddenly going down the tube).
That means less HF loop gain and more HF THD at all power levels.
Cheers,
Glen
Re: Re: Re: Re: Re: Re: clamp
Hm. Yes. I always thought that when you allow low Vce, the amp would already start to 'soft clip' as it were, from itself. The reduction of BW in that region would, I thought, make it more stable instead of less, because the unity gain point would be reached earlier. Of course there is also the additional phase shift so maybe the effect is not that simple.
Conditional stability often shows up when the amp comes out of clip, but I guess that would not change whether you would let it clip at Vce=5 or Vce=2, for example?
Jan Didden
G.Kleinschmidt said:
Hm. Yes. I always thought that when you allow low Vce, the amp would already start to 'soft clip' as it were, from itself. The reduction of BW in that region would, I thought, make it more stable instead of less, because the unity gain point would be reached earlier. Of course there is also the additional phase shift so maybe the effect is not that simple.
Conditional stability often shows up when the amp comes out of clip, but I guess that would not change whether you would let it clip at Vce=5 or Vce=2, for example?
Jan Didden
Re: Re: Re: Re: Re: Re: Re: clamp
The OPS does contribute additional phase shift at low Vce, and that generally lessens the phase margin, not improves it (in my experiments anyway).
Those few extra volts can make a huge difference. Below 5V is when most devices abruptly start suffering major fT droop (with most RET's you also have to avoid >5A Ic, because it has the same effect).
Cheers,
Glen
janneman said:
The OPS does contribute additional phase shift at low Vce, and that generally lessens the phase margin, not improves it (in my experiments anyway).
janneman said:
Those few extra volts can make a huge difference. Below 5V is when most devices abruptly start suffering major fT droop (with most RET's you also have to avoid >5A Ic, because it has the same effect).
Cheers,
Glen
Re: Re: Re: Re: Re: Re: Re: clamp
Hi Jan,
In case of BJT output devices, I'm not sure whether Vce=5V vs Vce=2V really matters, but I do know, in case of MOSFET, that it does matter.
An early version of our PGP amp, for example, got unstable when the output exceeded a certain level. But we were able to remedy this evil by an additional lead-lag compensation.
Also, all my simulations confirmed that you're asking for trouble if you let the amp clip at a low value of Vds or Vgd, say less the 3V.
Cheers, Edmond.
janneman said:
Hi Jan,
In case of BJT output devices, I'm not sure whether Vce=5V vs Vce=2V really matters, but I do know, in case of MOSFET, that it does matter.
An early version of our PGP amp, for example, got unstable when the output exceeded a certain level. But we were able to remedy this evil by an additional lead-lag compensation.
Also, all my simulations confirmed that you're asking for trouble if you let the amp clip at a low value of Vds or Vgd, say less the 3V.
Cheers, Edmond.
Re: Re: Re: Re: clamp
Hi Glen. I agree. Note that, to my knowledge, Tom Holman was the first to use the rail-tracking clamp in his Apt 1 power amplifier (see my post #2154 above). It is definitely a good technique.
I also agree that it does not always make sense to try to mine the last couple of volts of output swing possible. Not only is the higher device capacitance a concern, but also one is swinging perilously close to the nastiness residing on the main rails, the least of which is some 120 Hz sawtooth hum.
As I see it, the little bit of extra power dissipation in the power amplifier that results from providing that couple of volts of extra margin is just part of the price of achieving a good-sounding, well-behaved amplifier.
To put it in perspective, a 100 watt amplifier might have nominal main rails of, say, 45 volts. To provide two more extra volts of margin, one would have to design it with 47 volt rails. The amplifier's output stage power dissipation under any given output power condition would then go up by about 4.4%, necessitating a 4.4% larger heat sink and a 4.4% larger power transformer. This is not an unreasonable price to pay for a better-behaved amplifier that will likely be less sensitive to the garbage on the rails.
In fairness, one can look at it another way. If one did nothing to the power supply and enforced 2 extra volts of margin, the maximum power output might go down by about 8.8%, turning a 100 watt amplifier into a 91 watt amplifier.
Cheers,
Bob
Cheers,
Bob
G.Kleinschmidt said:
Hi Glen. I agree. Note that, to my knowledge, Tom Holman was the first to use the rail-tracking clamp in his Apt 1 power amplifier (see my post #2154 above). It is definitely a good technique.
I also agree that it does not always make sense to try to mine the last couple of volts of output swing possible. Not only is the higher device capacitance a concern, but also one is swinging perilously close to the nastiness residing on the main rails, the least of which is some 120 Hz sawtooth hum.
As I see it, the little bit of extra power dissipation in the power amplifier that results from providing that couple of volts of extra margin is just part of the price of achieving a good-sounding, well-behaved amplifier.
To put it in perspective, a 100 watt amplifier might have nominal main rails of, say, 45 volts. To provide two more extra volts of margin, one would have to design it with 47 volt rails. The amplifier's output stage power dissipation under any given output power condition would then go up by about 4.4%, necessitating a 4.4% larger heat sink and a 4.4% larger power transformer. This is not an unreasonable price to pay for a better-behaved amplifier that will likely be less sensitive to the garbage on the rails.
In fairness, one can look at it another way. If one did nothing to the power supply and enforced 2 extra volts of margin, the maximum power output might go down by about 8.8%, turning a 100 watt amplifier into a 91 watt amplifier.
Cheers,
Bob
Cheers,
Bob
Re: Re: Re: Re: Re: clamp
Bob,
Sure, but the impact of a few extra volts of supply on the SOA (in case of BJTs) would probably be a bit higher than the 4.4% number suggests.
But I agree, it's part of the price of better sound. And that 8.8% is just 0.8dB, although Marketing would call it 'a whopping, almost 10 Watts more power'
Jan Didden
Bob Cordell said:
Bob,
Sure, but the impact of a few extra volts of supply on the SOA (in case of BJTs) would probably be a bit higher than the 4.4% number suggests.
But I agree, it's part of the price of better sound. And that 8.8% is just 0.8dB, although Marketing would call it 'a whopping, almost 10 Watts more power'
Jan Didden
Re: Re: Re: Re: Re: Re: clamp
OK.
BTW, as for the CFB-clamp, my website is updated (but still not finished), see: www.data-odyssey.nl ->currents project -> output stage -> details.
Cheers, Edmond.
G.Kleinschmidt said:Huh?
Ain't got one of those. I'm talking about my "brute force" clamp, man. Schematic posted earlier.
Cheers,
Glen
EDIT:
Oops. In my post 2188, the second paragraph in the quotation box is the one I was replying to.
OK.
BTW, as for the CFB-clamp, my website is updated (but still not finished), see: www.data-odyssey.nl ->currents project -> output stage -> details.
Cheers, Edmond.