Hi all,
Thanks for all the helpfull comments and suggestions so far!
I thought I'd let you know what I'll do with them:
My plan is to first *finish* an amplifier (casing and all I mean), so I'll integrate the suggestions that are easiest to incorporate first:
- Lower value for R56 (100 instead of 1200)
- base resistors for Q9/Q109
- BC550C/BC560C in LTPs (to be honest: schematic is inaccurate since I already used those. The reason I put a cascode in the LTP in the first place was the slightly over 40V power supply, a little bit too close for comfort to the limits of the BC550 & BC560...)
I don't need alterations around ccs LED and cascode zener for reasons of reducing hum, that's already non-existent anyway ('pre-stage' and 'power-stage' are already somewhat decoupled by a LPF of R40/C42, R42/C43), so I'll leave those suggestions for v2.0
I already tried the lower R56 value (very easy modification of course) and was pleasantly surprised by the effect
Much more detailed! Switch from OpAmp based to solid state yielded much more details in mid & high. This simple change resulted in even more details but predominantly in 'lower' regions, e.g. cello and the lower keys of the piano ('better defined attacks of strings' is the best way I can think of to describe it). Some recordings are also less 'harsh' than before.
For the exisiting prototype the addition of base resistors for Q9/Q109 can be retrofitted relatively easy by cutting two tracks and drilling four holes. Since a decent amplifier needs two channels (either in the same case or separate, but I don't want to start that discussion here) it's very easy to include it in a new PCB as well (minor mod).
All other suggestions require (and certainly deserve!) more investigation & design.
So why finish an amplifier first? Well, I don't expect to ever finish with tweaks and improvements and it would be a shame if that would prevent me from ever finishing a complete amplifier.
Also my ultimate goal is a bi-amped system, actively if possible, but that's in a far and distant future. Since that implies a second amplifier I already have a valid excuse to look into & try the other suggestions.
Which brings me to the final (last but certainly not least) reason: The SWMBO factor (or HWMBO in my case). At least finishing a complete amplifier is a very persuasive manner of justifying more time & money spent on this subject ("What? are you telling me that you don't want to enjoy better music?")
Next steps:
- build the other channel.
- put it in a nice 19" case (hmm, that makes it sound so easy)
- try not to forget to take pictures along the way
- enjoy/work on v2.0/think of preamp (all at the same time)
Wow, what a hobby!
Remco
P.S.: Sorry for the long post, but I feel that the thoughtful and helpful suggestions at least deserve some explanation as to what I'll do with them.
Thanks for all the helpfull comments and suggestions so far!
I thought I'd let you know what I'll do with them:
My plan is to first *finish* an amplifier (casing and all I mean), so I'll integrate the suggestions that are easiest to incorporate first:
- Lower value for R56 (100 instead of 1200)
- base resistors for Q9/Q109
- BC550C/BC560C in LTPs (to be honest: schematic is inaccurate since I already used those. The reason I put a cascode in the LTP in the first place was the slightly over 40V power supply, a little bit too close for comfort to the limits of the BC550 & BC560...)
I don't need alterations around ccs LED and cascode zener for reasons of reducing hum, that's already non-existent anyway ('pre-stage' and 'power-stage' are already somewhat decoupled by a LPF of R40/C42, R42/C43), so I'll leave those suggestions for v2.0
I already tried the lower R56 value (very easy modification of course) and was pleasantly surprised by the effect
Much more detailed! Switch from OpAmp based to solid state yielded much more details in mid & high. This simple change resulted in even more details but predominantly in 'lower' regions, e.g. cello and the lower keys of the piano ('better defined attacks of strings' is the best way I can think of to describe it). Some recordings are also less 'harsh' than before.
For the exisiting prototype the addition of base resistors for Q9/Q109 can be retrofitted relatively easy by cutting two tracks and drilling four holes. Since a decent amplifier needs two channels (either in the same case or separate, but I don't want to start that discussion here
) it's very easy to include it in a new PCB as well (minor mod).All other suggestions require (and certainly deserve!) more investigation & design.
So why finish an amplifier first? Well, I don't expect to ever finish with tweaks and improvements and it would be a shame if that would prevent me from ever finishing a complete amplifier.
Also my ultimate goal is a bi-amped system, actively if possible, but that's in a far and distant future. Since that implies a second amplifier I already have a valid excuse to look into & try the other suggestions
.Which brings me to the final (last but certainly not least) reason: The SWMBO factor (or HWMBO in my case). At least finishing a complete amplifier is a very persuasive manner of justifying more time & money spent on this subject ("What? are you telling me that you don't want to enjoy better music?"
)Next steps:
- build the other channel.
- put it in a nice 19" case (hmm, that makes it sound so easy)
- try not to forget to take pictures along the way
- enjoy/work on v2.0/think of preamp (all at the same time)
Wow, what a hobby!
Remco
P.S.: Sorry for the long post, but I feel that the thoughtful and helpful suggestions at least deserve some explanation as to what I'll do with them.
Re: Images... ...1 of 2
I'll stick to about 40V... Never let the magic smoke escape!, you should know better than that
Nice pictures!
Cool to see the phase shift increase between 10 & 100 kHz .
(and the 'wobbles' on the peaks of the triangle @100kHz)
If I manage to get my hands on a digital camera you'll see 'real' scope pictures (including the 'Tektronix 434 storage oscilloscope' I mean).
Were you already able to listen to it and find out how it sounds?
(Just curious, since you apparently use fets)
dkemppai said:
Cool, here are a few images. (I hope you can find that camera to snap a few to post also )
The measurements are pretty self explanatory, as to frequency, phase, and gain (can be calculated from peak to peak) Blue trace is output signal into 8 ohm 100W non inductive resistor. Yellow trace is input into amplifier.
I was going to grab a few more, but I left it run at high power a little too long... ...POP! SNAP! followed by orange flame from the fets, and lots of magic smoke. So, until I rebuild (In about an hour) you'll have to settle for the 100Khz and single 10Khz traces...
Also, the amp was running at +- 50 Volt rails, +- 70 volts seems to improve linearity at high frequencies.
I'll stick to about 40V... Never let the magic smoke escape!, you should know better than that
Nice pictures!
Cool to see the phase shift increase between 10 & 100 kHz .
(and the 'wobbles' on the peaks of the triangle @100kHz)
If I manage to get my hands on a digital camera you'll see 'real' scope pictures (including the 'Tektronix 434 storage oscilloscope' I mean
).Were you already able to listen to it and find out how it sounds?
(Just curious, since you apparently use fets)
Re: Re: Images... ...1 of 2
This is the second set of Fets I've burned up, so there's lots of magic smoke on the walls to help me think
(Still can't figure out why my wife bought me that fire extinguisher, fets ALWAYS self extinguish Trust me, I know)
Yes, I use fets for the output devices, but the rest of the amp is BJT. No, I have not had a chance to listen to it. Before I hook speakers to it the bias voltage for the output fets needs some temp compensation. It also needs the integrator (Although, the trim resistor I'm using for DC offset seems to work well). I've only spent a few days doing the design/test, so there's lots that needs to be done. I've got my first amp sitting here also, and I plan to do some 'side by side' comparisons.
You say you've got a board designed? How much for parts for your amp? Are they avaliable from common supplers in the US? It would be nice to hear how yours sounds. It would also be nice to see some pics of your board and stuff, as you get them!
Also, Looking at your schematic (Nice layout), You may try a resistor from base to emitter on both Q9 and Q109. Something in the range of 470 to 100 Ohms. Darlington pairs often use them to help with the slew rate...
-Dan
Rambi said:
I'll stick to about 40V... Never let the magic smoke escape!, you should know better than that
Were you already able to listen to it and find out how it sounds?
(Just curious, since you apparently use fets)
This is the second set of Fets I've burned up, so there's lots of magic smoke on the walls to help me think
(Still can't figure out why my wife bought me that fire extinguisher, fets ALWAYS self extinguish
Trust me, I know)Yes, I use fets for the output devices, but the rest of the amp is BJT. No, I have not had a chance to listen to it. Before I hook speakers to it the bias voltage for the output fets needs some temp compensation. It also needs the integrator (Although, the trim resistor I'm using for DC offset seems to work well). I've only spent a few days doing the design/test, so there's lots that needs to be done. I've got my first amp sitting here also, and I plan to do some 'side by side' comparisons.
You say you've got a board designed? How much for parts for your amp? Are they avaliable from common supplers in the US? It would be nice to hear how yours sounds. It would also be nice to see some pics of your board and stuff, as you get them!
Also, Looking at your schematic (Nice layout), You may try a resistor from base to emitter on both Q9 and Q109. Something in the range of 470 to 100 Ohms. Darlington pairs often use them to help with the slew rate...
-Dan
[
I don't think so, the diodes (not zeners) are just there to prevent the input from rising above or below the supply voltages of the opamp (not very likely, but you never know).
Since the integrator is connected to the input side of the LTPs it needs to be a 'negative' integrator as well (i.e. positive input voltage causes negative going ramp on output of integrator).
Besides, connecting it the wrong way around would cause a DC offset, not cancel it, that would be noticeable soon enough
Remco
timc said:
I don't think so, the diodes (not zeners) are just there to prevent the input from rising above or below the supply voltages of the opamp (not very likely, but you never know).
Since the integrator is connected to the input side of the LTPs it needs to be a 'negative' integrator as well (i.e. positive input voltage causes negative going ramp on output of integrator).
Besides, connecting it the wrong way around would cause a DC offset, not cancel it, that would be noticeable soon enough
Remco
Re: Re: Re: Images... ...1 of 2
Sanken transistors are the most expensive parts (for populating the board, i.e. not counting heatsinks, casing, toroids, etc.)
BD911/BD912 (which I use for prototypes) are a few euros per piece if I buy them in a store, but on-line store like Reichelt seems to have them for about 50 eurocents (roughly the same in dollars).
Remco
Most stuff is 'ordinary' and shouldn't be hard to get in the US. Don't know about the BC type transistors, should be no problem to replace them with other types of small signal transistors and maybe MJE340/350 for the BD139/140 types?dkemppai said:
Sanken transistors are the most expensive parts (for populating the board, i.e. not counting heatsinks, casing, toroids, etc.)
BD911/BD912 (which I use for prototypes) are a few euros per piece if I buy them in a store, but on-line store like Reichelt seems to have them for about 50 eurocents (roughly the same in dollars).
Remco
grimberg said:
http://www.tek.com/cgi-bin/wtx.cgi?...es/escope/overview/index.html&FrameSet=Escope
A bit spendy, but nice! I've also used the TDS1000, TDS2000 and TDS200, and TDS400 series. All very nice. (TDS1000 Starts at about $1000)
-Dan
I have a question about the servo op amp, U1. There is only a capacitor in the feedback loop. If the op amp has a dc output offset voltage (and they all do) it is possible for a dc voltage to develop across the integration capacitor and ultimately build up so large as to saturate the op amp. Typically, a large value resistor is paralled across the capacitor to provide a path for the dc to discharge. You have a pretty large Rin right now, and the LF411 is not exactly a low offset voltage op amp, so I am concerned about this charge buildup.
If this is a problem, you may need to lower R80 to maybe around a 10 - 100K and parallel C80 with 1 to 10 M. The output network may have to be rescaled as well.
If this is a problem, you may need to lower R80 to maybe around a 10 - 100K and parallel C80 with 1 to 10 M. The output network may have to be rescaled as well.
Re: Images... ...1 of 2
Very dangerous to drive the amp at very high frequency with no or little load, even dangerous with load. I have made the same mistake myself. +-47 V, BD317, BD318. No load, 200 kHz = smoke! The output transistors started to conduct at the same time. Talk about fireworks in the fuses!
dkemppai said:
Very dangerous to drive the amp at very high frequency with no or little load, even dangerous with load. I have made the same mistake myself. +-47 V, BD317, BD318. No load, 200 kHz = smoke! The output transistors started to conduct at the same time. Talk about fireworks in the fuses!
Re: Re: Images... ...1 of 2
Not really true. I've been running the amp at HF for a couple of days now, load and no load. As long as voltages do not go over specified ratings of parts, no harm is done. Under certian situations, the lead inductance can generate large voltage spikes, but I hardly suspect this was the problem at 100Khz.
In my case, the problem was not the frequency, but rather the heat generated. I was dumping about 70W into an 8 ohm load for an extended period of time with poor heatsinking. I knew heat was a problem, so kept high power tests to short bursts. (All, but the last one, of course)
-Dan
peranders said:
Not really true. I've been running the amp at HF for a couple of days now, load and no load. As long as voltages do not go over specified ratings of parts, no harm is done. Under certian situations, the lead inductance can generate large voltage spikes, but I hardly suspect this was the problem at 100Khz.
In my case, the problem was not the frequency, but rather the heat generated. I was dumping about 70W into an 8 ohm load for an extended period of time with poor heatsinking. I knew heat was a problem, so kept high power tests to short bursts. (All, but the last one, of course)
-Dan
Peranders is right
Operating an audio amplifier at high frequency and max power for long periods can result in destruction of the output transistors.
This is caused by cross conduction of output transistors, because of the relatively long time needed to evacuate charge carriers in the base-emitter region.
Amplifiers with modified output drive like this one (R56, C51) are more tolerant, because of the accelerated evacuation of the charge carriers by the RC network. This is also depending on particular characteristics of the output transistors, high Ft and low Ccb are essential.
Regards, Pierre Lacombe.
Operating an audio amplifier at high frequency and max power for long periods can result in destruction of the output transistors.
This is caused by cross conduction of output transistors, because of the relatively long time needed to evacuate charge carriers in the base-emitter region.
Amplifiers with modified output drive like this one (R56, C51) are more tolerant, because of the accelerated evacuation of the charge carriers by the RC network. This is also depending on particular characteristics of the output transistors, high Ft and low Ccb are essential.
Regards, Pierre Lacombe.
Re: Peranders is right
Hi,
I realize that cross conduction can be a problem, but I'm pretty sure that this is not so in my particular case. The output devices get just as hot at 1Khz as 100Khz. under similar loads. (Again because of the small temporary heatsink). Under no load situations HF and LF the devices stay cool.
However, I guess, if you consider that the fets may need more VGS to conduct current into the load than under no load, there is more charge in the gate, and thus more to remove. Not sure how more current affects the levels of gate charge with the input/output capacitance.
I would suspect that with the gate charge and input/output capacitance, the problem should also be as apparent on the rising egde of the waveform as well as the falling one, showing itself as a limiting of rising as well as falling slew rate.
Overall, my gut feeling is this isn't the case, but I'm open to the suggestion. (And more comments)
-Dan
P.Lacombe said:
Hi,
I realize that cross conduction can be a problem, but I'm pretty sure that this is not so in my particular case. The output devices get just as hot at 1Khz as 100Khz. under similar loads. (Again because of the small temporary heatsink). Under no load situations HF and LF the devices stay cool.
However, I guess, if you consider that the fets may need more VGS to conduct current into the load than under no load, there is more charge in the gate, and thus more to remove. Not sure how more current affects the levels of gate charge with the input/output capacitance.
I would suspect that with the gate charge and input/output capacitance, the problem should also be as apparent on the rising egde of the waveform as well as the falling one, showing itself as a limiting of rising as well as falling slew rate.
Overall, my gut feeling is this isn't the case, but I'm open to the suggestion. (And more comments)
-Dan
Jeff R said:
In this case that won't matter since the (inverting) integrator feeds back to the input of the amplifier; It's a closed loop. A dc buildup by the integrator will be amplified (by the amplifier) and fed back to the input of the integrator, negating the buildup (since the integrator itself is inverting).
Of course an offset of the opamp might cause the output of the amplifier to be not exactly 0V. But I'm not interested in 'absolute zero', I just want to compensate for drift by temperature changes and imbalances in the circuit (although it probably won't compensate for totally unmatched components).
Although there's no connection drawn in the schematic, it is connected by the label 'IN', which is also placed at the top of C61. I used labels here to keep the schematic relatively clean. Same holds for the feedback (NFB labels, between R54 & R55 and at the base of Q4).
Remco
Rambi said:
While true, the general rule in audio design seems be to use local feedback rather than overall global feedback. By using a paralleled resistor you can negate most of the charge build up right there without having to go through the feedback. I have no idea if the effects are audible - probably not - but then who would have thought that different brands of metal film resistors would sound different?
Just something to consider. Good luck!
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.