Member
Joined 2009
Paid Member
By the way, I wasn't concerned about hijacking, the more interest in helping me develop TGM the better for my rate of learning. But I guess it would be less confusing and more democratic with a fresh start.
A completely flexible approach sounds good, but to avoid a mess you have to introduce some constraints - I assume we're still talking Class AB boostrap style amplifier ?
Do you mean to include we should consider a MOSFET output option, I am thinking about this more and more since it is less complex for newbie from a thermal management perspective. It also means that high current circuits can be better isolated from low current circuits (like my TGM pcb - shameless plug on my part !). Reliability and safety is a very good point too, I include speaker/amp protection in this.
A completely flexible approach sounds good, but to avoid a mess you have to introduce some constraints - I assume we're still talking Class AB boostrap style amplifier ?
Do you mean to include we should consider a MOSFET output option, I am thinking about this more and more since it is less complex for newbie from a thermal management perspective. It also means that high current circuits can be better isolated from low current circuits (like my TGM pcb - shameless plug on my part !). Reliability and safety is a very good point too, I include speaker/amp protection in this.
No complexity here , same as TGM 1 or the DX (both are good amps)
same 7 transistor design as the RCA. Just add a big PS (flat 6A bridge and 6800-10kuf caps) , A alternative to the chip amps for the "first timers". 🙂 Possibly a special CCS to keep the original 55 "sound". A smaller amp like this will pop the rail fuses before it takes out the speaker... usually.
That is a good idea , depending on what type of mosfet. Vertical devices would be a "drop in replacement" for they have nearly the same thermal comp. as BJT's , just a slight change to the Vbe.
Lateral devices are much more expensive and just require a trimmer for bias. Both mosfet variants would need gate protection and gatestoppers (same as basestoppers) , so no reduction in parts count. What you WOULD get is a 2 OP device TGM that would do 100W-8R/200W-4R at 50-0-50Vdc.
I would of gone this route already , but my NJW BJT's are $1.40 each vs. IRF240/9240 (vertical mosfets)at $4.40 each.
OS
same 7 transistor design as the RCA. Just add a big PS (flat 6A bridge and 6800-10kuf caps) , A alternative to the chip amps for the "first timers". 🙂 Possibly a special CCS to keep the original 55 "sound". A smaller amp like this will pop the rail fuses before it takes out the speaker... usually.
That is a good idea , depending on what type of mosfet. Vertical devices would be a "drop in replacement" for they have nearly the same thermal comp. as BJT's , just a slight change to the Vbe.
Lateral devices are much more expensive and just require a trimmer for bias. Both mosfet variants would need gate protection and gatestoppers (same as basestoppers) , so no reduction in parts count. What you WOULD get is a 2 OP device TGM that would do 100W-8R/200W-4R at 50-0-50Vdc.
I would of gone this route already , but my NJW BJT's are $1.40 each vs. IRF240/9240 (vertical mosfets)at $4.40 each. OS
Member
Joined 2009
Paid Member
ostripper said:
Lumab, OS,
My thinking is not a Cascode. Maybe I don't understand, perhaps the Cascode is the same thing as I am saying ? What I am looking at is a current sink under the VAS. I don't want to isolate the VAS from the sonic benefits of the bootstrap so I figure the current sink has to live down there at the bottom out of the way. It will try to control the current flow through the VAS but it's a couple of junctions away from the bootstrap node. Am I embarrassing myself here ?
Don't feel bad , I knew much less when I built my first ones (still don't know enough).
What you are doing is not really isolating the VAS device from the bootstrap , just linearizing it, current wise . Since the main VAS device is not doing the "work" anymore, memory and miller effects are more constant.
read this article.. http://peufeu.free.fr/audio/memory/
(part 5 - VAS)
A current sink under the VAS might not work for number of reasons.
OS
What you are doing is not really isolating the VAS device from the bootstrap , just linearizing it, current wise . Since the main VAS device is not doing the "work" anymore, memory and miller effects are more constant.
read this article.. http://peufeu.free.fr/audio/memory/
(part 5 - VAS)
A current sink under the VAS might not work for number of reasons.
OS
Member
Joined 2009
Paid Member
ostripper said:
I agree - to be attractive it must be better than the chip amps as they are much easier to construct. My experience as a noob starting from scratch was that there's a lot of hidden work. The chasis, the power supply, materials selection - you don't get to the fun part too quickly. If you can make this easier it would be good.
Having a personality behind an amplifier helps. When you build a DX you aren't just building a DX, you are building a Carlos DX ! (or an AKSA, a supersym, a Frugal, a Patchwork... etc.). It can't be too dry, must be some magic.
You can get close with this one , but the constructor will have to bias the Op stage. One could just use a diode bias string like the RCA , but the magic of the discreets would be lost. 40+ parts is more than 15+ (LM3886 chip amp).
🙂 This is harder than designing the amp itself. It is THE true classic amp , the one that fender adapted for its SS guitar line and all the Japanese made variants. Maybe EASYAMP classic, or NOOBWORX 50, D - TOY. I am not a marketer. 🙁 OS
Member
Joined 2009
Paid Member
ostripper said:
Noobworx - perfect !!!!!
=====
TGM: Today we have to get mean with TGM1. Time to push it outside of the comfortable area. I dug out some Latin American music, bass heavy. I also dug out some Heavy Metal. Both of them deserving a fairly high volume.
Now we find things are different. The TGM1 does not have the bass presence of the Bryston, it's quite noticeable. I swapped around channels in case it was a L vs R thing, but the bass performance moved with the amp not with the channel. The Bryston maintained it's slightly cleaner presentation overall but with the added bonus of effortless bass. The TGM couldn't do that, I had to crank up the volume a long way (and move away from the speakers!); it felt like it had run out of steam.
Perhaps 24V rails with big music into big speakers at party volume levels are not ideal bedfellows. At comfortable listening levels the TGM1 was my preferred choice and even for the Latin you'd be very happy. So TGM1 is still my first choice for the HT system.
(note to myself: next project use minimum 36V rails).
Bigun said:
Your finding what I found with my first self build amp that the amp quickly runs out of steam on transients.
And I was using +/-45 volt rails.
On my next amp I increased the rails to +/-60 volts and that improved things quite a bit.
Music has a vast dynamic range and the transients are easily clipped if care is not taken.
Member
Joined 2009
Paid Member
ostripper said:
Well, all these comments got me thinking.
I still got to thinking about improving the VAS (reducing distortion), it's well trodden ground so here goes...
As I understand, there are a few things for me to look at:
1) Cascode: I poked my nose into D.Self on the web to look at the Cascoded VAS. It looks like a nice step forward. There is the question of buffering it from the non-linear load of the output but even without, it's still effective in reducing distortion. What I've learned is that you acknowledge the inherent non-linearity of the VAS device and try to reduce the distortion by increasing the local negative feedback. So we do this by increasing the effective collector resistance. The additional cascode device pins the VAS collector voltage.
2/ PSRR: We get noise off the -ve rail (I saw this with my X-talk simulations). So we can try to isolate the VAS emitter off the -ve rail. One option is a constant voltage source (ahaha, I get to put a device below the VAS afterall !) which we can implement with an LED and BJT.
3/ Differential Drive: This one is a bit more tricky. I found this in D. Self. The VAS is driven in a quasi-differential fashion. The collector from the input device of the LTP goes to the base as usual, but we also take a wire from the feedback device collector and apply this to the VAS emitter. This also drives up the gain. It also seems not to require a buffer which the Cascode method recommends. This sounds interesting because it may be possible to do this without adding any more active devices to TGM1 (the example in D.Self adds an extra device though) and this appeals to me.
Hi Nigel,
That's an interesting comment, I have neglected to look at the clipping behaviour. Right now I have the feedback set up to produce fairly low gain, a 1V input should give me a 15V output so I am limiting things.
Definitely , 40V would be perfect for deep bass at moderate levels on medium speakers (8" 2-ways).
I see you have been reading the "bible" (D. self). 😀 , one drawback to the cascode and the CVS is that you lose voltage swing (it will run out of steam sooner) ,but you will have less THD and PSRR.
The differential drive sounds interesting , with the low rails of the TGM , a mod like this would not be as dangerous an exercise as with my 77v- 10+a rails. If you read to the end of self's VAS chapter ,you also see the balanced VAS , he says it has too much complexity ??
Another thing to consider is that reducing distortion by the above methods is comparable to actually building a full blown blameless. You will alter the character (sound) of the bootstrap. I built a few blameless's , I think their sound is flat and sterile (here I am in agreement with carlos).
OS
Member
Joined 2009
Paid Member
I'm not planning to go down the route of a balanced VAS, it's digging a bigger hole in terms of complexity. Also it it would probably lead to making a more symmetrical design overall and this isn't the direction I have taken with TGM. Furthermore, a fully balanced VAS seems a well travelled route whereas the quasi-differential looks like a path less travelled and that makes it more interesting for me. My next step in this regard is to look at some simulations; this will help me understand this quasi-balanced approach before I go tinkering with the hardware. And definitely it is a goal to retain the sonic character of the bootstrap.
Member
Joined 2009
Paid Member
Update:
Following advice I am now also changing capacitors (having already changed some resistors) to see how things can be optimized for TGM1.
I have focussed on C5 (compensation at VAS). I have to say right off the bat that I found the changes hard to quantify, hard to be sure of myself because each change took some minutes and my brain-ears were losing their memory between. I used my Bryston on L channel, TGM1 on R channel to compare and try to hold a consistent course.
stock: 68p - so far I like this sound, not as precise/open as the Bryston, bass is less firm but mids and highs are nice
change to: 22pF - didn't like this, bass sounded less but probably because highs were more (Carlos warned us of how changes in trebble can trick our perception of bass). It sounded somehow flat, not engaging, like a bedside radio (well not that bad actually !)
change to 44pF - the magic came back, the amp sounded more open, bass still good and trebles still smooth, it's as if the amplifier found where it's comfortable.
change to 112pF - I didn't like this, the magic was less again.
Conclusion - I have left it with 44pF and listened for a longer time. It seems improved but how much I can't quantify as my ears simply struggle with the changes. With Chesky CD (christy baron) it is simply very satisfying (Zinfandel?).
I am getting very impatient to add the 2nd channel...but have to fix the hum first.
Following advice I am now also changing capacitors (having already changed some resistors) to see how things can be optimized for TGM1.
I have focussed on C5 (compensation at VAS). I have to say right off the bat that I found the changes hard to quantify, hard to be sure of myself because each change took some minutes and my brain-ears were losing their memory between. I used my Bryston on L channel, TGM1 on R channel to compare and try to hold a consistent course.
stock: 68p - so far I like this sound, not as precise/open as the Bryston, bass is less firm but mids and highs are nice
change to: 22pF - didn't like this, bass sounded less but probably because highs were more (Carlos warned us of how changes in trebble can trick our perception of bass). It sounded somehow flat, not engaging, like a bedside radio (well not that bad actually !)
change to 44pF - the magic came back, the amp sounded more open, bass still good and trebles still smooth, it's as if the amplifier found where it's comfortable.
change to 112pF - I didn't like this, the magic was less again.
Conclusion - I have left it with 44pF and listened for a longer time. It seems improved but how much I can't quantify as my ears simply struggle with the changes. With Chesky CD (christy baron) it is simply very satisfying (Zinfandel?).
I am getting very impatient to add the 2nd channel...but have to fix the hum first.
My experience of the VAS capacitor was that too little C allowed oscilation on the output and too much made the amp lose higher frequencies.
If your losing bass then look at the input filter if there is one.
Also look at the capacitor in the feedback divider chain and consider increasing its value.
If your losing bass then look at the input filter if there is one.
Also look at the capacitor in the feedback divider chain and consider increasing its value.
Gareth,
that capacitor is a nuisance and should not be used to tune the sound. Also, there`s nothing magic about the bootstrap.
that capacitor is a nuisance and should not be used to tune the sound. Also, there`s nothing magic about the bootstrap.
Member
Joined 2009
Paid Member
Hi Nigel (you're up late !),
Indeed, I don't have particularly generous caps in those locations. The HT project has a sub so the amp doesn't have to deliver below 80Hz. However, the thing is, it's not how low the bass goes but how much impact it has (some people call it 'slam' ?). It isn't a problem as such but I would like to learn what aspects of the design impact 'slam' rather than simply the bass extension.
Indeed, I don't have particularly generous caps in those locations. The HT project has a sub so the amp doesn't have to deliver below 80Hz. However, the thing is, it's not how low the bass goes but how much impact it has (some people call it 'slam' ?). It isn't a problem as such but I would like to learn what aspects of the design impact 'slam' rather than simply the bass extension.
Member
Joined 2009
Paid Member
Lumba Ogir said:
Lumba, good to see you here !
I remember you have a dislike for this capacitor but to be honest, it didn't seem to sound as nice when I dropped the value of it. At the end, it's always there inside the transistor anyhow.
What would you adjust to optimize the sound of TGM1 ?
Gareth,
to get better bass (and overall) performance, the power supply capacitance that needs to be increased (not the rail voltage). The capacitor bank is to be counted in hundred thousands of microfarads.
to get better bass (and overall) performance, the power supply capacitance that needs to be increased (not the rail voltage). The capacitor bank is to be counted in hundred thousands of microfarads.
- Status
- Not open for further replies.