OK - so here's what happened in the end.
Put in on a scope and couldn't get it to produce any weirdness, it starts to clip with anything more than 0.5v on the input at full volume.
I reduced the transformer from 30-0-30 to 20-0-20 and removed the current limiting. I checked the the output trannies had the diode protection in and they did. I added external diodes from c-e too - though I doubt it did anything.
It does now work and although it's not really powerful enough, at least it doesn't blow up. Time to move on as I don't see a way of getting this reliable at the full 30-0-30 voltage. I have the sub amp from Elliot sound products on order.
Thanks for everyone's help on this - it's been an interesting experience..
Dan
Put in on a scope and couldn't get it to produce any weirdness, it starts to clip with anything more than 0.5v on the input at full volume.
I reduced the transformer from 30-0-30 to 20-0-20 and removed the current limiting. I checked the the output trannies had the diode protection in and they did. I added external diodes from c-e too - though I doubt it did anything.
It does now work and although it's not really powerful enough, at least it doesn't blow up. Time to move on as I don't see a way of getting this reliable at the full 30-0-30 voltage. I have the sub amp from Elliot sound products on order.
Thanks for everyone's help on this - it's been an interesting experience..
Dan
Dan,
I am glad - but excuse me for adding one more comment. If you then did get it to work with the scope and all, did you also look at a square wave? When you said it clipped I presume you meant a sine wave.
The problem of blowing the transistors may very well show up as overshoot on the edges of a square wave, ringing or some other anomaly. That is the advantage of looking at a square wave as an "early warning". Overshoot will indicate a supersonic peak in the amplitude/frequency response somewhere, and from there one can carefully tailor by small changes in the feedback phase capacitor, etc. Hope you can try this (at low output).
Regards
I am glad - but excuse me for adding one more comment. If you then did get it to work with the scope and all, did you also look at a square wave? When you said it clipped I presume you meant a sine wave.
The problem of blowing the transistors may very well show up as overshoot on the edges of a square wave, ringing or some other anomaly. That is the advantage of looking at a square wave as an "early warning". Overshoot will indicate a supersonic peak in the amplitude/frequency response somewhere, and from there one can carefully tailor by small changes in the feedback phase capacitor, etc. Hope you can try this (at low output).
Regards
Based on the circuit design of this amp, and the possibility of high frequency oscillation blowing the output transistors, can anyone suggest some cap changes that might make it less likely to oscillate? Bearing in mind I wanted to use this as a basic sub amp and couldn't give a monkey's about anything much over ~100Hz...
Thanks a million,
Dan
Thanks a million,
Dan
I know and if it wasn't for me having the assistance of our electronics technicians at work, I'd assume it was entirely my fault. But... we've rebuilt the whole thing and it still doesn't work right.
With an input and speaker connected (nothing playing - just connected), it gets to a certiain point in the volume and the voltage across R19 (bias resistor) goes from 15mv or so right up to 0.5v - on the 30v supply this will blow the output pair, on the 20v it just causes a hum.
Weird!
With an input and speaker connected (nothing playing - just connected), it gets to a certiain point in the volume and the voltage across R19 (bias resistor) goes from 15mv or so right up to 0.5v - on the 30v supply this will blow the output pair, on the 20v it just causes a hum.
Weird!
Dan,
Something is very wrong here, now that you state voltages. The voltage across R7 should be some 600mV for normal operation. 15mV definitely means that the driver T6 is in cut-off, with the output sitting close to + rail. Did you start with the power stage current adjuster RV1 at its lowest value, increasing slowly until the correct current is drawn (which you measure how?) While the output transistors go when in fact the voltage across R7 is correct, I suspect that the output stage might be drawing far too much "zero" current.
Then (and I think this is separate), the effect of the volume control, if it is connected at the input, is that at zero resistance (tap at common), the input base is shorted to common a.c. wise. But this is where the use of R3 connected to the feedback point can cause trouble. With some resistance in the volume control the said base is no longer at common, but quickly closer to the feedback point than common. This should not normally worry matters, but it can in a certain type of circuit. (I am not necessarily judgemental about the Velleman circuit, but something is funny, and one cannot give precise advice without tracing matters with a scope - sorry. I have never built the circuit, especially with Darlingtons.)
But (pweh! this is getting complicated!) you should have a d.c. potential at the output emitters' junction of about half your rail voltage, or in this case close to 0 and a voltage across R7 of about 600mV. Not to underestimate you, but are you quite sure that component values are correct? (Locally I sometimes have difficulty because the body colour of some makes of resistors so affects the ring colours as to confuse me and I have to measure.)
Let me stop here. I hope you can check the voltages, etc. Also, you do need to know if there is h.f. oscillation (instability) at any stage. But also, one must be careful to measure inside a feedback loop with long meter leads; it can cause oscillation and give wrong figures and ruin things ....... sometimes nothing is easy with transistors!
Regards
Something is very wrong here, now that you state voltages. The voltage across R7 should be some 600mV for normal operation. 15mV definitely means that the driver T6 is in cut-off, with the output sitting close to + rail. Did you start with the power stage current adjuster RV1 at its lowest value, increasing slowly until the correct current is drawn (which you measure how?) While the output transistors go when in fact the voltage across R7 is correct, I suspect that the output stage might be drawing far too much "zero" current.
Then (and I think this is separate), the effect of the volume control, if it is connected at the input, is that at zero resistance (tap at common), the input base is shorted to common a.c. wise. But this is where the use of R3 connected to the feedback point can cause trouble. With some resistance in the volume control the said base is no longer at common, but quickly closer to the feedback point than common. This should not normally worry matters, but it can in a certain type of circuit. (I am not necessarily judgemental about the Velleman circuit, but something is funny, and one cannot give precise advice without tracing matters with a scope - sorry. I have never built the circuit, especially with Darlingtons.)
But (pweh! this is getting complicated!) you should have a d.c. potential at the output emitters' junction of about half your rail voltage, or in this case close to 0 and a voltage across R7 of about 600mV. Not to underestimate you, but are you quite sure that component values are correct? (Locally I sometimes have difficulty because the body colour of some makes of resistors so affects the ring colours as to confuse me and I have to measure.)
Let me stop here. I hope you can check the voltages, etc. Also, you do need to know if there is h.f. oscillation (instability) at any stage. But also, one must be careful to measure inside a feedback loop with long meter leads; it can cause oscillation and give wrong figures and ruin things ....... sometimes nothing is easy with transistors!
Regards
Dan,
To follow - still trying to help: To get one thing right at a time (and if you cannot check at the output for absence of oscillation), it might help to decrease the feedback temporarily by making the 3K3 resistor (R8 if I recall correctly) larger - say about 15K - 22K, and remove the 47pF (temporarily) until the d.c. matters are in order. But - sorry to repeat - you do need to see whether there is instability (oscillation). The presence of that can mess up all d.c. measurements. It is simply a no-go to keep blowing things up. Can you get hold of an oscilloscope?
Regards.
To follow - still trying to help: To get one thing right at a time (and if you cannot check at the output for absence of oscillation), it might help to decrease the feedback temporarily by making the 3K3 resistor (R8 if I recall correctly) larger - say about 15K - 22K, and remove the 47pF (temporarily) until the d.c. matters are in order. But - sorry to repeat - you do need to see whether there is instability (oscillation). The presence of that can mess up all d.c. measurements. It is simply a no-go to keep blowing things up. Can you get hold of an oscilloscope?
Regards.
Johan,
There seems to be some confusion here - I was talking about the mv across R19 - which is set as recommended (10-15mv) not across R7 - which I admit I haven't measured.
I will (again) double check all the values, but I have done this already - relying on a meter rather than my ability to read the bands correctly.
I can make the temporary changes you suggest - is it ok to run it up like that or is there a risk - ie if it improves matters, why not leave it with the modifications?
If you need to check the schematic it's page 19 of this pdf:
k8060 manual
I can bring the amp to work and connect it to a scope - what, where and how would you check as we've already done some basic comparisons of input and output waves (from a signal generator) but admittedly should have set it up so that it would 'go crazy' to see what was happening.
Thanks again!
Dan
There seems to be some confusion here - I was talking about the mv across R19 - which is set as recommended (10-15mv) not across R7 - which I admit I haven't measured.
I will (again) double check all the values, but I have done this already - relying on a meter rather than my ability to read the bands correctly.
I can make the temporary changes you suggest - is it ok to run it up like that or is there a risk - ie if it improves matters, why not leave it with the modifications?
If you need to check the schematic it's page 19 of this pdf:
k8060 manual
I can bring the amp to work and connect it to a scope - what, where and how would you check as we've already done some basic comparisons of input and output waves (from a signal generator) but admittedly should have set it up so that it would 'go crazy' to see what was happening.
Thanks again!
Dan
I've used the one provided - a little plastic single turn effort - 1k.
Exactly as in this picture (only black!)
picture of k8060
I can try something different of course. Should it still be 1k (as provided in the kit parts) as I notice in the circuit that's it down as 500R?
Dan
Exactly as in this picture (only black!)
picture of k8060
I can try something different of course. Should it still be 1k (as provided in the kit parts) as I notice in the circuit that's it down as 500R?
Dan
It's not a massive deal, but ideally if a 1k pot has been used that would mean the other resistors each side need to be increased (about doubled) as well. It wouldn't hurt to try a 500 ohm pot if the other resistors are as per schematic, but I'd be a little surprised if it was the cause of the problems.
Hi Dan,
Oh boy!
What am I doing in the middle of the night - and not for the first time, I saw when rereading all the posts. My apologies.
Dan, I am going to model and analise this circuit on Spice - owe you that much (that is p. 19, k6080 manual). It is a little unsatisfactory suggesting one thing after another and not seeing the effect of every change, with the danger of destroying yet more components. Such a simulation will not include casual capacitances, but will give an idea of where any sensitivities lie, if any. There are quite a number of capacitors that can influence stability, should that be a factor (i.e. C1, C2, C3, C4, C5, C6-R10, etc.)
I hope to be able to get back to you tomorrow night (it is now 22:20 here), if someone else does not come up with wisdom in the meantime. Let us hope that something will come up.
Regards.
Oh boy!
What am I doing in the middle of the night - and not for the first time, I saw when rereading all the posts. My apologies.
Dan, I am going to model and analise this circuit on Spice - owe you that much (that is p. 19, k6080 manual). It is a little unsatisfactory suggesting one thing after another and not seeing the effect of every change, with the danger of destroying yet more components. Such a simulation will not include casual capacitances, but will give an idea of where any sensitivities lie, if any. There are quite a number of capacitors that can influence stability, should that be a factor (i.e. C1, C2, C3, C4, C5, C6-R10, etc.)
I hope to be able to get back to you tomorrow night (it is now 22:20 here), if someone else does not come up with wisdom in the meantime. Let us hope that something will come up.
Regards.
AMAZING:
I tried a 500k pot, but it wouldn't give me the values I nees to get the bias right, so I went back to a new 1k, 10 turn type.
Miracle, it now works! I've been able to use the 30-0-30 supply too! The little platic trim pot must have had a problem, although it measures ok? I cannot believe after all this hassle and all the re-stuffing etc. that the pot was the cause, but it is now working absolutely fine - or at least as well as a simple kit like this can.
Sooo many thanks to all for your input, but especially to Richie for hitting the nail on the head. It makes sense in the end as the pot was the only bit not changed or replaced from original.
Cheers,
Dan
I tried a 500k pot, but it wouldn't give me the values I nees to get the bias right, so I went back to a new 1k, 10 turn type.
Miracle, it now works! I've been able to use the 30-0-30 supply too! The little platic trim pot must have had a problem, although it measures ok? I cannot believe after all this hassle and all the re-stuffing etc. that the pot was the cause, but it is now working absolutely fine - or at least as well as a simple kit like this can.
Sooo many thanks to all for your input, but especially to Richie for hitting the nail on the head. It makes sense in the end as the pot was the only bit not changed or replaced from original.
Cheers,
Dan
Dan!!
OK - I went out before reading of your success, so I suppose to relate the results of my promised analysis does not make much sense now. (I was lead astray by your finding about 15mV across R19, which indicated that your bias per RV1 was set somewhere near correct.)
Anyway, I did a few hours of analysis on the circuit including some worst case, which I do not suppose will do much harm relating it here - it may be of value to others. There are certain things, I fear, that are not kosher, even though it will "play". This has mostly to do with stability under different load conditions, especially with some capacitance, and improved distortion performance, although the latter is quite low. Also apology if this is going to be a little long.
To begin with, stability is to be reckoned with because of the inordinately high NFB factor of 52dB (400x). As a result compensation sets in after 3 KHz, which leaves quite a reactive interaction at higher frequencies; something I try to avoid.
The most serious matter is perhaps the very variable input impedance with frequency. We observed correctly that R3 acts in an amplified way, being connected to the NFB return point, but such an arrangement has a penalty - it is NFB-dependant. In this case the mid-frequency impedance is about 46K, but going right down to <3K round about 20Hz as a result of NFB phase change because of the values of C9 and particularly C11. This would manifest as an audible lack of deep bass as soon as one feeds the input via a volume control introducing some series kilo-ohms. (In the analysis my generator internal impedance was 2K.)
The best way to improve this is to increase the impedance of the whole network. A substantial increase in C11 only (to a lesser extent C9) would do the trick, but leaves one with several thousand uF; not very practical. (Also keep in mind that R3 and R8 must be equal to ensure any sort of balance in the currents of T1, T2, something that will affect the zero-currents in T7, T8 - they are quite sensitive to this.)
I ended up with R3=R8=33K, C11=25uF - 33uF (not more), R2=1K, C3=6,8pF, and C9=470nF (an electrolytic here is not the best). For some relief from early phase shift caused by C1, that value could become C1=12 - 15pF. The input changes will give an input impedance of 47K, rolling off to no less than 38K at 15 Hz.
It was mentioned that some equality in the collector currents of T1, T2 be obtained. (See the power amplifier design discussion by Douglas Self.) I fear that in this circuit Ic(T2) was about 70% of Ic(T1). I changed R9 to 4,7K to get within 4%. [The problem with a non-adjustable balanced input circuit like this is that T1 and T2 must be equal. Either you select them (which the manufacturers may have done) or you arrange for adjustment. Simple circuits can be over-simplified.)]
Finally, I wondered about the fairly large C4, C5 shunting a high-impedance circuit. Under all circumstances they adversely affected pulse response; I removed them. Perhaps someone else can enlighten me about what their purpose might have been. Overshoot also cleared up when I removed R4 (make a short), and connected about 10nF over C, E of T6.
This may sound like a redesign of something intended to be simple; it in fact simplifies it even more. I could really not find justification for the original values, both as far as stability and pulse response (related) are concerned. The intention was not redesign, simply a cleaning-up. Many other refinements are possible, as most will realise. The amplifier still does not like capacitive loads - for that the usual simple inductance in series with the load is required. But it is now stable up to no load - at least in analysis. (As stated before casual constructional capacitances cannot be taken into account.)
Also, though the Zobel network (C6-R10) showed some advantage on analysis, I have found in the past that not all circuits like that in practice. It has to do with the layout, practical loudspeaker impedance, etc.
So, this naturally puts you under no obligation, Dan; the analysis was good exercise. I am happy for your success after such frustration. You would have learnt that it is often the simplest things that bogs one down. I have had many such experiences.
Good luck!
OK - I went out before reading of your success, so I suppose to relate the results of my promised analysis does not make much sense now. (I was lead astray by your finding about 15mV across R19, which indicated that your bias per RV1 was set somewhere near correct.)
Anyway, I did a few hours of analysis on the circuit including some worst case, which I do not suppose will do much harm relating it here - it may be of value to others. There are certain things, I fear, that are not kosher, even though it will "play". This has mostly to do with stability under different load conditions, especially with some capacitance, and improved distortion performance, although the latter is quite low. Also apology if this is going to be a little long.
To begin with, stability is to be reckoned with because of the inordinately high NFB factor of 52dB (400x). As a result compensation sets in after 3 KHz, which leaves quite a reactive interaction at higher frequencies; something I try to avoid.
The most serious matter is perhaps the very variable input impedance with frequency. We observed correctly that R3 acts in an amplified way, being connected to the NFB return point, but such an arrangement has a penalty - it is NFB-dependant. In this case the mid-frequency impedance is about 46K, but going right down to <3K round about 20Hz as a result of NFB phase change because of the values of C9 and particularly C11. This would manifest as an audible lack of deep bass as soon as one feeds the input via a volume control introducing some series kilo-ohms. (In the analysis my generator internal impedance was 2K.)
The best way to improve this is to increase the impedance of the whole network. A substantial increase in C11 only (to a lesser extent C9) would do the trick, but leaves one with several thousand uF; not very practical. (Also keep in mind that R3 and R8 must be equal to ensure any sort of balance in the currents of T1, T2, something that will affect the zero-currents in T7, T8 - they are quite sensitive to this.)
I ended up with R3=R8=33K, C11=25uF - 33uF (not more), R2=1K, C3=6,8pF, and C9=470nF (an electrolytic here is not the best). For some relief from early phase shift caused by C1, that value could become C1=12 - 15pF. The input changes will give an input impedance of 47K, rolling off to no less than 38K at 15 Hz.
It was mentioned that some equality in the collector currents of T1, T2 be obtained. (See the power amplifier design discussion by Douglas Self.) I fear that in this circuit Ic(T2) was about 70% of Ic(T1). I changed R9 to 4,7K to get within 4%. [The problem with a non-adjustable balanced input circuit like this is that T1 and T2 must be equal. Either you select them (which the manufacturers may have done) or you arrange for adjustment. Simple circuits can be over-simplified.)]
Finally, I wondered about the fairly large C4, C5 shunting a high-impedance circuit. Under all circumstances they adversely affected pulse response; I removed them. Perhaps someone else can enlighten me about what their purpose might have been. Overshoot also cleared up when I removed R4 (make a short), and connected about 10nF over C, E of T6.
This may sound like a redesign of something intended to be simple; it in fact simplifies it even more. I could really not find justification for the original values, both as far as stability and pulse response (related) are concerned. The intention was not redesign, simply a cleaning-up. Many other refinements are possible, as most will realise. The amplifier still does not like capacitive loads - for that the usual simple inductance in series with the load is required. But it is now stable up to no load - at least in analysis. (As stated before casual constructional capacitances cannot be taken into account.)
Also, though the Zobel network (C6-R10) showed some advantage on analysis, I have found in the past that not all circuits like that in practice. It has to do with the layout, practical loudspeaker impedance, etc.
So, this naturally puts you under no obligation, Dan; the analysis was good exercise. I am happy for your success after such frustration. You would have learnt that it is often the simplest things that bogs one down. I have had many such experiences.
Good luck!
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.