Re: SC ULD2
Ultra-LD Mk.2 200W Power Amplifier, August 2008: See Part 2 in September 2008 (page 70) for minor circuit modifications. There are also some corrections: the 470uF 63V capacitor on the -55V rail is shown with reversed polarity on the circuit diagram (Fig.1), ie, its negative lead should go to the -55V rail. As well, the voltages marked across the 0.1 ohm resistors are out by a factor of 10. They should be 7-10mV, not 70-100mV.
In addition, the 68 ohms resistor in series between the +55V rail and Q7’s collector has been reduced to 47 ohms, while a 22k resistor has been added in series between Q8’s collector and ground (this change has been incorporated onto the PC board patterns sent to the parts retailers).
Finally, the following changes should be made to the parts list: add 1 x 22k 0.25W resistor, add 1 x 47 ohms 0.25W resistor, add 1 x 10 ohms 0.25W resistor, delete 1 x 68 ohms 0.25W resistor and change the two 470 ohms 5W test resistors to 68 ohms 5W. There should also be 5 x 100nF 63V MKT polyester capacitors (not four). (09/08)
That's a good idea - I hadn't realised they had an errata section. As it turns out what's there is only what's in part II of the article. Here's what the errata says:sandyK said:
Ultra-LD Mk.2 200W Power Amplifier, August 2008: See Part 2 in September 2008 (page 70) for minor circuit modifications. There are also some corrections: the 470uF 63V capacitor on the -55V rail is shown with reversed polarity on the circuit diagram (Fig.1), ie, its negative lead should go to the -55V rail. As well, the voltages marked across the 0.1 ohm resistors are out by a factor of 10. They should be 7-10mV, not 70-100mV.
In addition, the 68 ohms resistor in series between the +55V rail and Q7’s collector has been reduced to 47 ohms, while a 22k resistor has been added in series between Q8’s collector and ground (this change has been incorporated onto the PC board patterns sent to the parts retailers).
Finally, the following changes should be made to the parts list: add 1 x 22k 0.25W resistor, add 1 x 47 ohms 0.25W resistor, add 1 x 10 ohms 0.25W resistor, delete 1 x 68 ohms 0.25W resistor and change the two 470 ohms 5W test resistors to 68 ohms 5W. There should also be 5 x 100nF 63V MKT polyester capacitors (not four). (09/08)
Re SCULD Amp
Sounds as if you are new at this so my advice is as follows, build the amp
as per SC/JCAR Kit with no mods , power up & check voltages across the .1ohm Resistors & see what you get , as this will all depend on the variations in the transistor sets, when i first powered mine up the max voltage i could get was
2mV thats when i tried the mod with the pot as per john clarks suggestion & that was
it , make sure you set the pot to zero first then let the amp warm up 10-15mins
then slowly adjust pot for 7-10mv across resistors.
I found it very stable once set , checked mine again tonight all OK, average of
10mv across all resistors thats after 3months of fairly heavy use every day.
Thats the only mod i would do , just run the amp for a few weeks & see how you like it before attempting anything else.
I can send you some pics of where to mount the pot if you like just e-mail me.
mick@plus8.com
Sounds as if you are new at this so my advice is as follows, build the amp
as per SC/JCAR Kit with no mods , power up & check voltages across the .1ohm Resistors & see what you get , as this will all depend on the variations in the transistor sets, when i first powered mine up the max voltage i could get was
2mV thats when i tried the mod with the pot as per john clarks suggestion & that was
it , make sure you set the pot to zero first then let the amp warm up 10-15mins
then slowly adjust pot for 7-10mv across resistors.
I found it very stable once set , checked mine again tonight all OK, average of
10mv across all resistors thats after 3months of fairly heavy use every day.
Thats the only mod i would do , just run the amp for a few weeks & see how you like it before attempting anything else.
I can send you some pics of where to mount the pot if you like just e-mail me.
mick@plus8.com
Guys,
Do not believe me that this OPS is not well designed ... please do yourself the following test:
Run the amp very hard, just below the clipping level, in 8ohm resistive load, for about 10 minutes. In the same time measure the voltage drop over one Re.
Remove the input signal and watch the voltage variation over Re.
Please report here the results.
I bet that SC never measure THD and adjust Re voltage drop for minimum distortion value
Regards,
Mihai
Do not believe me that this OPS is not well designed ... please do yourself the following test:
Run the amp very hard, just below the clipping level, in 8ohm resistive load, for about 10 minutes. In the same time measure the voltage drop over one Re.
Remove the input signal and watch the voltage variation over Re.
Please report here the results.
I bet that SC never measure THD and adjust Re voltage drop for minimum distortion value
Regards,
Mihai
It is not a "bad" design , just over simplified. As it is just
a "blameless" with NJL's , I think a minimalist approach was
the point.
I use the same devices without the integrated
diode (njw0281/0302) , and use the same internal mur120 as a
separate component. The better technique would be to
have the diodes bias an active transistor Vbe. (attached)
OPS Re's of .22R - .33R also improve thermal stability.
It might run for a long time as it is , but never give consistant
performance (thd wise).
OS
Attachments
In the case of this SC amp, the mur120's are in the devices,
you would use them.. as that is the point of the NJL OP.
The diode /mje340 circuit I posted will track to 1-2mv on SMALL
heatsinks and you can set the tracking coefficient with
the ratio or the bias resistors.By changing the R on the diode "side" of the circuit you can track ANY HS ,big or small.
On the SC amp , the only adjustment of the Vbias coefficient is to
remove or add a diode from the "chain".
Whether you have to do this will be largly dependant on the
C/per Watt of your heatsink. With the ideal sized HS , (the
one SC designed the amp ON
) , the 4 diodes might beadequate. with a larger one- less change per C (3 diodes, etc.)
here is the thread...
http://www.diyaudio.com/forums/showthread.php?s=&threadid=71534&perpage=25&highlight=thermaltrak&pagenumber=4
Doug Self argues this point out with the "boys"..
OS
Some progress... Had another try today, starting afresh with a whole new kit. I've put it together as per SC's article, except I've added the trimpot to try to get the right voltages across the 0R1's.
Unfortunately, the 1st time I powered it up I had the wrong polarity from the PSU, and fried a resistor.
I replaced that, waited for it to warm up a bit, then adjusted the trimpot to get about 12mV across Q12 and Q13's resistors. However, at this point, the voltage across Q14's resistor is 21mV, and across Q15 it's only 4mV.
Adjusting the trimpot doesn't increase the voltage across Q15's resistor - although it goes up a bit for a few seconds, it then drops back to 4mV.
Is this normal? Or does it mean I've fried Q15, or stuffed something else up?
Thanks for everyone's help so far - it's great to be progressing!
Unfortunately, the 1st time I powered it up I had the wrong polarity from the PSU, and fried a resistor.
I replaced that, waited for it to warm up a bit, then adjusted the trimpot to get about 12mV across Q12 and Q13's resistors. However, at this point, the voltage across Q14's resistor is 21mV, and across Q15 it's only 4mV.Adjusting the trimpot doesn't increase the voltage across Q15's resistor - although it goes up a bit for a few seconds, it then drops back to 4mV.
Is this normal? Or does it mean I've fried Q15, or stuffed something else up?
Thanks for everyone's help so far - it's great to be progressing!
SC ULD 200W AB Amp
Jeremy brought it around yesterday and we examined it in detail.
Offset at +/-20V on each rail was around -1.4V, with large currents drawn indicating output device failure. There was no sign of malfunction at the CCSs, or the VAS, or the LTP.
I found a flaw in the biasing system, already identified by John Clarke in the Errata columns of the mag, IIRC.
There are two pairs of 3281D/1302D outputs, the TO3-P devices with five leads incorporating a control diode. All four of these diodes are connected in series as the bias generator. No bias trimmer is provided for on the board, and Jeremy had added a 200R pot in series with the diode string.
The four emitter resistors are small, only 0.1 ohm. This makes bias adjustment tetchy, and increases the risk of thermal runaway. Anything under 0.22R is, in my experience, potentially troublesome.
A pnp and npn output had blown in previous attempts to get it running. These were removed, though we didn't have replacements.
Jeremy spotted before I did that removing the faulty outputs also removed the diodes, breaking the bias string and leading to rail to rail bias. I'm so used to a single BD139 Vbe multiplier! We soldered a couple of 4148s across the broken string, but quiescent current was then excessive (we used a variac for safety).
Bridged one diode, and tried again, all systems go, set it to 27mV on the trimmer, which corresponds to 270mA of quiescent, about three times that recommended. We adjourned for lunch. Bad decision!
On return an hour later, both remaining outputs had fried, taking out fuses and base stoppers, thermal runaway. Combined dissipation would have started at 30 watts on a 8" x 3" heatsink, too much. Two of the internal diodes in the outputs are not enough to pull back runaway, all four are required.
This points up some interesting lessons. These transistors have tightly controlled tempcos, and the stage as designed needs four diodes in the string to control the output devices, even just two. Furthermore, it's folly to expect accurate bias control by simply using a bias string without a trimpot. On top of that, 0.1R emitter resistors give very thin margins should the devices be subjected to thermal strain; currents shoot up very quickly.
Since the value of the emitter resistor is ameliorated by global feedback, I can see no benefit in going below 0.22R. I'm told there are sonic benefits, but in this design they come at a cost. Damping factor on the ULD 200W is probably better than 200 anyway. Certainly there are power losses with higher emitter resistors, but this can be offset by using three pairs rather than two for 56V rails.
Conclusion: these new transistors are a new technology and their use in audio amps can only improve as more people use them and adjust to their quirks. I believe they should be used with no less than 0.22R emitter resistors, though I've yet to prove this empirically. I will suggest to Jeremy on Thursday when he calls that we do just this......
Cheers,
Hugh
Jeremy brought it around yesterday and we examined it in detail.
Offset at +/-20V on each rail was around -1.4V, with large currents drawn indicating output device failure. There was no sign of malfunction at the CCSs, or the VAS, or the LTP.
I found a flaw in the biasing system, already identified by John Clarke in the Errata columns of the mag, IIRC.
There are two pairs of 3281D/1302D outputs, the TO3-P devices with five leads incorporating a control diode. All four of these diodes are connected in series as the bias generator. No bias trimmer is provided for on the board, and Jeremy had added a 200R pot in series with the diode string.
The four emitter resistors are small, only 0.1 ohm. This makes bias adjustment tetchy, and increases the risk of thermal runaway. Anything under 0.22R is, in my experience, potentially troublesome.
A pnp and npn output had blown in previous attempts to get it running. These were removed, though we didn't have replacements.
Jeremy spotted before I did that removing the faulty outputs also removed the diodes, breaking the bias string and leading to rail to rail bias. I'm so used to a single BD139 Vbe multiplier! We soldered a couple of 4148s across the broken string, but quiescent current was then excessive (we used a variac for safety).
Bridged one diode, and tried again, all systems go, set it to 27mV on the trimmer, which corresponds to 270mA of quiescent, about three times that recommended. We adjourned for lunch. Bad decision!
On return an hour later, both remaining outputs had fried, taking out fuses and base stoppers, thermal runaway. Combined dissipation would have started at 30 watts on a 8" x 3" heatsink, too much. Two of the internal diodes in the outputs are not enough to pull back runaway, all four are required.
This points up some interesting lessons. These transistors have tightly controlled tempcos, and the stage as designed needs four diodes in the string to control the output devices, even just two. Furthermore, it's folly to expect accurate bias control by simply using a bias string without a trimpot. On top of that, 0.1R emitter resistors give very thin margins should the devices be subjected to thermal strain; currents shoot up very quickly.
Since the value of the emitter resistor is ameliorated by global feedback, I can see no benefit in going below 0.22R. I'm told there are sonic benefits, but in this design they come at a cost. Damping factor on the ULD 200W is probably better than 200 anyway. Certainly there are power losses with higher emitter resistors, but this can be offset by using three pairs rather than two for 56V rails.
Conclusion: these new transistors are a new technology and their use in audio amps can only improve as more people use them and adjust to their quirks. I believe they should be used with no less than 0.22R emitter resistors, though I've yet to prove this empirically. I will suggest to Jeremy on Thursday when he calls that we do just this......
Cheers,
Hugh
Nice to see an objective opinion, from a different design camp. I am not sure why you ran it at 270ma as my amps running anything above 11ma start to smell the big brown smoke when playing.
Nice work though, a pity you lost some "players", maybe the old amp Jeremy has, has some working components, I hear Hugh is right onto the green revolution
Keep on DIY ing
Alfred
Nice work though, a pity you lost some "players", maybe the old amp Jeremy has, has some working components, I hear Hugh is right onto the green revolution
Keep on DIY ing
Alfred
Re: SC ULD 200W AB Amp
Hello Hugh,
Your investigation is on the right track, but:
There is nothing wrong with 0.1 emitter resistors as long as five NJL diodes are used in the serial biasing string.
In this amplifier, wrongly designed after ONsemi ThermalTrack application note , only four NJL diodes are used, which is not enough to prevent OPS thermal runaway, especially when low value Res are used. Why? Because the Thermal Track embedded diode has 1.6mv/K thermal coefficient but the base-emitter junction has 2.1mV/K TC.
The only working solution, other than classical Vbe multiplier (which nullify the TT advantage), is to add another NJL like diode to the bias string. You can use MUR120, which is the same diode as the one used insight the TT power devices case, thermally bounded with the heat sink.
RMI-FC100 has 0R1 emitter resistors and five NJL diodes and is stable. It was tested in very harsh environments and was proved to be thermally indestructible. Also, the bias recovery ofter running the amp at high power level is very very quick.
Cheers,
Mihai
AKSA said:
Hello Hugh,
Your investigation is on the right track, but:
There is nothing wrong with 0.1 emitter resistors as long as five NJL diodes are used in the serial biasing string.
In this amplifier, wrongly designed after ONsemi ThermalTrack application note
, only four NJL diodes are used, which is not enough to prevent OPS thermal runaway, especially when low value Res are used. Why? Because the Thermal Track embedded diode has 1.6mv/K thermal coefficient but the base-emitter junction has 2.1mV/K TC.The only working solution, other than classical Vbe multiplier (which nullify the TT advantage), is to add another NJL like diode to the bias string. You can use MUR120, which is the same diode as the one used insight the TT power devices case, thermally bounded with the heat sink.
RMI-FC100 has 0R1 emitter resistors and five NJL diodes and is stable. It was tested in very harsh environments and was proved to be thermally indestructible. Also, the bias recovery ofter running the amp at high power level is very very quick.
Cheers,
Mihai
Thanks Mihai,
Useful information; a real shame the internal diode does not share the exact same tempco as the power tranistor.
I notice you use five internal diodes in your amp to give the 2.2V bias, but then, you are using six output devices, and this is easy to arrange. With four devices, you must add the MUR120, and frankly it would be easier to simply use a well engineered Vbe multiplier with the cheaper transistors.
Cheers,
Hugh
Useful information; a real shame the internal diode does not share the exact same tempco as the power tranistor.
I notice you use five internal diodes in your amp to give the 2.2V bias, but then, you are using six output devices, and this is easy to arrange. With four devices, you must add the MUR120, and frankly it would be easier to simply use a well engineered Vbe multiplier with the cheaper transistors.
Cheers,
Hugh
SC ULD2
Hugh and Mihai
The fact remains that many hundreds (thousands ?) of these amplifier modules would have been constructed to date from kits, and provided that the amended instructions, as included with the kits are closely adhered to, the amplifier modules will perform quite satisfactorily.
Jeremy's present problem was due to an unfortunate accident, and there is every reason to believe that with due care,his next module will work correctly.
That is not saying that design improvements can not be made, but then the cost of the kits may then become prohibitive for many people.
SandyK
Hugh and Mihai
The fact remains that many hundreds (thousands ?) of these amplifier modules would have been constructed to date from kits, and provided that the amended instructions, as included with the kits are closely adhered to, the amplifier modules will perform quite satisfactorily.
Jeremy's present problem was due to an unfortunate accident, and there is every reason to believe that with due care,his next module will work correctly.
That is not saying that design improvements can not be made, but then the cost of the kits may then become prohibitive for many people.
SandyK
Hi,
as soon as I saw the bias scheme for this amp, I raised the red flag (agreeing with Lineup).
Now, we have two experts telling us that the bias scheme is wrong.
When do the rest of you start believing this sad design is simply a bad implementation of potentially a great set of devices in a great but flawed schematic.
as soon as I saw the bias scheme for this amp, I raised the red flag (agreeing with Lineup).
Now, we have two experts telling us that the bias scheme is wrong.
When do the rest of you start believing this sad design is simply a bad implementation of potentially a great set of devices in a great but flawed schematic.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.