Hi All,
Two years ago my wife bought me a set of KEF Q4’s for my birthday, and I have been wanting to build my self a decent stereo amplifier my speakers.
A couple of months ago I bought my self G. Randy Slone’s book (High Power Audio Amplifier Construction Manual). The information in this book thought me a lot about designing amplifiers with all the formulas. What I don’t like about the book is that there are uncertain factors like the gm of the input stage or the RC (stability components on the feedback rail).
Anyway I used the formulas to design an amplifier and the schematic attached is the result. I did do intensive simulation using Electronics Worksbench’s MultiSIM version 8 and the amplifier seems to be stable. The current theoretical specifications according to the application are as follows.
Input Impedance: 22K
Input Sensitivity: 1v RMS
Output Power: 60 Watts RMS into 6 Ohm (19.2v RMS)
Bandwidth: 2 Hz – 160 kHz (-3 dB)
Distortion: 2nd Harmonic (380 uV; 0.0019% @ 6 Ohms full drive @ 20 kHz)
3rd Harmonic (4.1 mV; 0.021% @ 6 Ohms full drive @ 20 kHz)
If some of you experts out there can please comment on my design I will really appreciate very much. Any feedback whether good or bad is most welcome as well as any pointers for the do’s and don’ts will be appreciated.
Thanks,
CorrieB
Two years ago my wife bought me a set of KEF Q4’s for my birthday, and I have been wanting to build my self a decent stereo amplifier my speakers.
A couple of months ago I bought my self G. Randy Slone’s book (High Power Audio Amplifier Construction Manual). The information in this book thought me a lot about designing amplifiers with all the formulas. What I don’t like about the book is that there are uncertain factors like the gm of the input stage or the RC (stability components on the feedback rail).
Anyway I used the formulas to design an amplifier and the schematic attached is the result. I did do intensive simulation using Electronics Worksbench’s MultiSIM version 8 and the amplifier seems to be stable. The current theoretical specifications according to the application are as follows.
Input Impedance: 22K
Input Sensitivity: 1v RMS
Output Power: 60 Watts RMS into 6 Ohm (19.2v RMS)
Bandwidth: 2 Hz – 160 kHz (-3 dB)
Distortion: 2nd Harmonic (380 uV; 0.0019% @ 6 Ohms full drive @ 20 kHz)
3rd Harmonic (4.1 mV; 0.021% @ 6 Ohms full drive @ 20 kHz)
If some of you experts out there can please comment on my design I will really appreciate very much. Any feedback whether good or bad is most welcome as well as any pointers for the do’s and don’ts will be appreciated.
Thanks,
CorrieB
Hi,
transistor choice seems too varied.
All the BC 550/560 could be the same grade, use c grade rather than a or b, unless Sloan has given a reason for low gain types.
Some of the other types could also be made similar or even the same but I don't have data on them.
Are you making the PCB yourself?
Then include base resistors on all the LTP inputs (4 off).
R19 C3 has the correct RC time constant for 1uS. but the small resistor value (low noise) and large capacitor value (poor high frequency) do not get optimum. You are trying to attenuate all above audio frequencies, including RF. a small cap does this better and since it's effect is virtually out of the audio band a high audio quality cap is not so important. an NP0 (=C0G) ceramic might be ok or a 680pF MKS, MKT, PES, PP etc would all do the job. If you adopt a bulky cap then allow space for a tiny RF cap in parallel. Be prepared to experiment between 0.5uS and 1.5uS to suit your speakers/ears/sources. Some even suggest pushing this to 0.3uS.
The DC block is right on the limit of my usual recomendation (80 to 100mS). But the NFB DC block is set to 264mS. This is lower than half an octave below the input DC blocking filter so ok, but the PSU filter needs to be a further half octave lower again and that target becomes nearly impossible to meet for 6ohms if you aim for 264*1.41=372mS. You will require +-62mF of smoothing capacitance per channel (a bit bulky and expensive). It will give extended bass but might interfere with mid and treble quality.
The FB point is shown between the emitter resistors, move the tapping point to the link between r24 and r25.
Are you planning to mount the Thiel network on the speakers terminals? If not then you MUST ensure that the ground return does not contaminate the other PCB grounding and cause oscillation.
Add a cap in parallel to r17 in case you need to adjust stability and/or output overshoot into slightly capacitive loading.
Consider adding a pair of series diodes (1.4V) across C7 to prevent reverse DC across the cap in event that the output offset goes high on fuse blowing or transistor failure. The cap voltage rating must be either rail voltage or add another pair of diodes or single Zener to protect the caps for offset the other way.
C9 may sound better if it is replaced with a polypropylene (or other high audio quality cap). But probably a lot smaller, try 1uF.
Leave space for on board decoupling caps from rail to power ground. Both electrolytic (100uF to 1mF) AND Hi-K ceramic 10nF to 100nF.
Make sure your audio ground and power ground are kept completely separate on the PCB.
Some of your grounds carry only DC currents and some carry low value AC currents and some medium value AC currents. All these are best returned to a PCB signal star ground. The high current grounds and those that see half wave AC currents must be combined separately from the clean ground.
That's enough to think about for now.
transistor choice seems too varied.
All the BC 550/560 could be the same grade, use c grade rather than a or b, unless Sloan has given a reason for low gain types.
Some of the other types could also be made similar or even the same but I don't have data on them.
Are you making the PCB yourself?
Then include base resistors on all the LTP inputs (4 off).
R19 C3 has the correct RC time constant for 1uS. but the small resistor value (low noise) and large capacitor value (poor high frequency) do not get optimum. You are trying to attenuate all above audio frequencies, including RF. a small cap does this better and since it's effect is virtually out of the audio band a high audio quality cap is not so important. an NP0 (=C0G) ceramic might be ok or a 680pF MKS, MKT, PES, PP etc would all do the job. If you adopt a bulky cap then allow space for a tiny RF cap in parallel. Be prepared to experiment between 0.5uS and 1.5uS to suit your speakers/ears/sources. Some even suggest pushing this to 0.3uS.
The DC block is right on the limit of my usual recomendation (80 to 100mS). But the NFB DC block is set to 264mS. This is lower than half an octave below the input DC blocking filter so ok, but the PSU filter needs to be a further half octave lower again and that target becomes nearly impossible to meet for 6ohms if you aim for 264*1.41=372mS. You will require +-62mF of smoothing capacitance per channel (a bit bulky and expensive). It will give extended bass but might interfere with mid and treble quality.
The FB point is shown between the emitter resistors, move the tapping point to the link between r24 and r25.
Are you planning to mount the Thiel network on the speakers terminals? If not then you MUST ensure that the ground return does not contaminate the other PCB grounding and cause oscillation.
Add a cap in parallel to r17 in case you need to adjust stability and/or output overshoot into slightly capacitive loading.
Consider adding a pair of series diodes (1.4V) across C7 to prevent reverse DC across the cap in event that the output offset goes high on fuse blowing or transistor failure. The cap voltage rating must be either rail voltage or add another pair of diodes or single Zener to protect the caps for offset the other way.
C9 may sound better if it is replaced with a polypropylene (or other high audio quality cap). But probably a lot smaller, try 1uF.
Leave space for on board decoupling caps from rail to power ground. Both electrolytic (100uF to 1mF) AND Hi-K ceramic 10nF to 100nF.
Make sure your audio ground and power ground are kept completely separate on the PCB.
Some of your grounds carry only DC currents and some carry low value AC currents and some medium value AC currents. All these are best returned to a PCB signal star ground. The high current grounds and those that see half wave AC currents must be combined separately from the clean ground.
That's enough to think about for now.
AndrewT,
Thanks for all the valuable information!
Do you have some kind of base line theory/recommendation that I might use for designing audio amplifiers? I am not an expert when it comes to electronics but I really want to learn!!! Where or with what kind of design/kit do you recommend I begin with?
Thanks
CorrieB
Thanks for all the valuable information!
Do you have some kind of base line theory/recommendation that I might use for designing audio amplifiers? I am not an expert when it comes to electronics but I really want to learn!!! Where or with what kind of design/kit do you recommend I begin with?
Thanks
CorrieB
The schematic looks good to me. The only thing I would really change is the multitude of transistors being used.
For current sources, mirrors, i would just use BC546B/556B in place of 2SA733/2SC945.
The VBE multiplier (Q15) I would probably use MJE340 or BD139 here so that it can be easily attached to the heatsink for thermal tracking.
I would not use MJE340/350 as drivers - I would instead use drivers such as MJE15034/5 or MJE15032/3.
The use of medium power 2SA916/2SC1941 as VAS buffers (Q11/Q12) strikes me as pointless - again BC546/556 would do here. The NEC transistors look quite good for the VAS though, as they are fairly fast and have a low Cob. If sourcing them is an issue, BD139/140 would do.
edit: I say all this assuming it is easier to get european devices where you are
For current sources, mirrors, i would just use BC546B/556B in place of 2SA733/2SC945.
The VBE multiplier (Q15) I would probably use MJE340 or BD139 here so that it can be easily attached to the heatsink for thermal tracking.
I would not use MJE340/350 as drivers - I would instead use drivers such as MJE15034/5 or MJE15032/3.
The use of medium power 2SA916/2SC1941 as VAS buffers (Q11/Q12) strikes me as pointless - again BC546/556 would do here. The NEC transistors look quite good for the VAS though, as they are fairly fast and have a low Cob. If sourcing them is an issue, BD139/140 would do.
edit: I say all this assuming it is easier to get european devices where you are
They would probably work but the MJ15003/4 pair are a lot older and slower devices. If you want TO-3 packaged devices, I would recommend MJ21193/4 but be aware these are still not as good as the MJL3281/1302 devices.
If you find japanese devices easier to get you could consider 2SA1943/2SC5200 but be aware - there are LOTS of counterfeits of these!
If they ship to your country, consider using the OnSemi samples programme - this would get you guaranteed genuine transistors for low cost.
If you find japanese devices easier to get you could consider 2SA1943/2SC5200 but be aware - there are LOTS of counterfeits of these!
If they ship to your country, consider using the OnSemi samples programme - this would get you guaranteed genuine transistors for low cost.
Hi,
I know I ask a lot of questions but where I am staying in my country we don’t have that many experts on the subject as here on this wonderful place of information.
From a design point of view this schematic is dual differential input stage with the required VAS for the input stage. If I go and design a single differential stage with a differential VAS will the sound difference from the amplifier be perceivable to my ear or will I only gain quality from a measurement point of view?
Thanks,
CorrieB
I know I ask a lot of questions but where I am staying in my country we don’t have that many experts on the subject as here on this wonderful place of information.
From a design point of view this schematic is dual differential input stage with the required VAS for the input stage. If I go and design a single differential stage with a differential VAS will the sound difference from the amplifier be perceivable to my ear or will I only gain quality from a measurement point of view?
Thanks,
CorrieB
Slone doesn't really encourage the use of a differential VAS. He seems to like the Darlington VAS the most, and says that a cascode VAS may be very good. He uses output buffering between the VAS and OPS sometimes. But differential VAS... are you sure that's what you want?corrieb said:
If you want to try just a non-symmetrical design, with a single-ended differential input stage and a conventional Darlington VAS, you can check the threads I'd started based on Slone's textbook designs.
TCPIP,
I have tried a single differential with a darlington VAS with two pole compensation worked out according to Slone's formulas and the for some reason the BE voltage drop across the VAS transistor was never more than 300 mV! and always experienced a voltage off set of more than 3 volts on the output.
When I remove the two pole compensation and the darlington configuration everything worked fine.
Thanks,
Corrie
I have tried a single differential with a darlington VAS with two pole compensation worked out according to Slone's formulas and the for some reason the BE voltage drop across the VAS transistor was never more than 300 mV! and always experienced a voltage off set of more than 3 volts on the output.
When I remove the two pole compensation and the darlington configuration everything worked fine.
Thanks,
Corrie
Unfortunately this design is completely wrong.
Read this:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=16796&highlight=
Read this:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=16796&highlight=
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.