300 Watt Amp

[raitraak]

Hey audioholics!

I am working on my first discrete project here and could use some help.

Goal of this project is to achieve 300W into 8ohm load. Why so much? Because it's interesting Main use will be subwoofer amplification, but it can also

be used as full range amp.

Power

Schematics

I plan to use 2 toroidal transformers (2*54V 500VA and 2*15V 50VA). Smaller one is for preamp, soft start and protection circuits. I have separated the

main supply into two: low current one for input and VAS stage and other for output stage. The goal is to minimize noise from class B currents. Secondary

supply is for preamp and is precisely adjustable.

Questions:
* Is the divided supply idea worth using?

Soft Start

Schematics

Soft start is using secondary transformer for power.

Questions:
*Is it a good idea to fuse second transormer's primary and secondary? When one of these fuses blows, soft start will be without a supply and main

transformer is powered through resistors, that will overheat very quickly.

Preamp

Schematics

Input is obtained from a balanced line through unified XLR and phone plug connector. Input stage is from Jensen transformers. It is followed by phase

inverter and phase controller (can be bypassed). Then comes adjustable low pass filter (can be bypassed). Last stage is voltage amplification to create

suitable voltage for main amp.

Questions:
*Are the values for input filtering suitable?
*Should i create separate ground for signal and power. This way the signal ground should unite with main ground in the main amp. Otherwise there will be

a ground loop (preamp - (signal ground) - main amp (ground) - power). Is it worth the concern?

Main amp

Schematics

Main amp is a compose of mirrored differential input and mirrors VAS and output triple. On the same board also resides speaker protection.

Questions:
* Is the concept ok?
* Can a MJE15032(MJE15033) in TO220 casing dissipate 1.5W without heatsink (used as current source for input)?
* Should I mount Vbe multiplier onto main heatsink or on a separate one with predrivers?

That's it for now
Thanks in advance!

Aq

 

[Leolabs]

500VA for one or both channel??Are you planning to drive anything below 8 ohms???

 

[woody]

I just looked the mje15032 up and the data on it says
2w max at 25c ambient without a sink 1.5w max 55c ambient
without a sink. I believe Nelson Pass said in the old A40 article
that designs with a 2 to 1 safety margin generaly blow up!
So I darn well wouldn't use it without a sink.

If you don't mount the vbe multiplier on the sink what are you
going to use to stabalize the bias to prevent thermal runaway.

I also see that in the preamp power supply cap c11 and c12
are tantalums. You probably don't want to use tantalums I
believe they sound realy harsh.

One other thing in the Hafler 200 thread there is mention of
a new drive board sutiable for an amp like this with realy outstanding specifications

 

[hienrich]

just to butt in:

I could see Randy Slone's name under the main amp page......

is this Randy Slone's new design?........from which edition?

thanks

 

[AndrewT]

I think C11 & C12 are in the wrong place. adj to common?
I don't like your choice of input HP filter and NFB HP filter.
How about adding the extra diode + R from protection base to ground for two slope characteristic?
Have you deliberately chosen transistors with a similar fT? why?
Output Re=0r1 seems a bit low. Optimum output bias for ClassAB is around 800mA to 1000mA giving around 150W of quiescent dissipation. Using Cordell's rule for heatsinking stability, these may need to be raised.

4pair off +-75Vdc supplies is asking a lot. 4ohm is not on.
6ohm might be OK. With 8ohm, the sink must be kept below 30degC, how with 150W of dissipation?

 

[raitraak]

Hey!

Thanks for your replies.

I have updated the schematics (see first post).

Leolabs

*There will be only one channel. 8 ohm load will be most important (my current subwoofer is based on Visaton's TIW300-8). But I'd like to drive 4 ohm loads also.

woody

*I will use separate small heatsinks for current sources.
*The question about Vbe multiplier was, that should it track the temperature of the predriver or output transistor. It's a
question of topology, predriver and driver form a CFP, which is confusing.
*I replaced tantalums in power and preamp circuits with 100u/50V electrolytics.
*Can you be more precise about this "new drive board", because the thread is quite large and it would take hours to go through it.

hienrich

*I have updated the refence on the main amp schematics. I have a copy of "High power audio amplifier construction manual. 50 to 500 Watts for the audio perfectionist" by R. Slone. Unfortunately i don't see an edition number.

AndrewT

* You are right. I'll correct that.
* Can you suggest something better?
* I will add two slope protection.
* I saw the combination of these transistors on this page http://sound.westhost.com/project68.htm. Already ordered samples from ONSemi, so I should stick with these. Is there a problem?
* 0R1 value is recommended by D. Self in his book. Also he states that Class AB is not recommended. I'm thinking of
quiesent current in range of 50 - 100 mA.
* Do you suggest adding more ouput pairs to support 4 ohm operation? Why must the sink be below 30degC?

Regards,
Aq

 

[AndrewT]

* I will add two slope protection.
* I saw the combination of these transistors on this page http://sound.westhost.com/project68.htm. Already ordered samples from ONSemi, so I should stick with these. Is there a problem?
* 0R1 value is recommended by D. Self in his book. Also he states that Class AB is not recommended. I'm thinking of
quiesent current in range of 50 - 100 mA.
* Do you suggest adding more ouput pairs to support 4 ohm operation? Why must the sink be below 30degC?

D.Self's ClassB is the same as optimised ClassAB that the rest of the industry uses as a definition.
Have another look and note he recommends by setting the voltage across the output Re, not by current.
This Forum has discussed this at length and there seems to be general agreement that for an EF output stage that 15mV to 25mV across each Re maintains optimum ClassAB (=Self's ClassB). Try to stay within this range for all operating conditions and temperatures.
You adopted those transistors because you saw them on ESPs site. good enough reason, but is it well informed?

I have done a few more calculations and 4pair can do 6ohm in domestic duty for wideband non clipping music use.
You must stay within the temperature de-rated SOAR. The higher the case temperature of the devices the lower the SOAR.

I don't think you'll hit your target with +-75Vdc rails and that tiny amount of smoothing capacitance. More likely you'll need +-81Vdc and that increases your quiescent dissipation even higher (160W).

If you want 4ohm capability AND want sub capability, I recommend at least 6pair for 300W into 8ohm and use an enormous heatsink or fanned sink with temp protection/cut-off.

 

[ashok]

Wouldn't it be better to go for a class D amp if the application is a sub woofer . 300 watt class D and maybe 150 watt class AB for the main speakers if a wideband amp is also required . Wouldn't that be simpler to do ? May cost a bit more I guess.

 

[woody]

Do a google for the LME49811 this driver chip has outstanding
specifations and sure would make building a 300w amp a lot easier and cheaper.

 

[raitraak]

Then it's a matter of defining I read books, because there is all nessecary information, to get through this forum's topics, it takes at least a full time job

At first I thought to build ESP design, but then started reading D. Self-s book and decided to create something more advanced. I already had ordered the devices, so I'm stuck with them.

I plan to use two heatsinks (1760cm2 50*191*100mm), but i'm planning to place them inside of the enclosure. If needed, I can add two thermally controlled fans. This creates strict limits on power dissipation and 150 watts for standby is clearly too much. Can you suggest "better" output topology with less standby current.

I have 5 pairs of MJL4302AG/MJL4281AG and 5 pairs of MJL21194G/MJL21193G. Can I swap driver transistors for MJL21194G/MJL21193G? This way I would have 5 output pairs.

I think I can forget 4 ohms and settle with less output power for 8 ohms. Caps are quite expensive around here, and I cannot afford more. Budget for this project is already quite big.

 

[wg_ski]

I have 5 pairs of MJL4302AG/MJL4281AG and 5 pairs of MJL21194G/MJL21193G. Can I swap driver transistors for MJL21194G/MJL21193G? This way I would have 5 output pairs.

I think I can forget 4 ohms and settle with less output power for 8 ohms. Caps are quite expensive around here, and I cannot afford more. Budget for this project is already quite big.

A better way would be to use the 21193/4 pairs as outputs with the 4281/4302's as drivers. If you have three stages of current gain you DO NOT need sustained-beta outputs. Using such devices as drivers and predrivers is always beneficial. The 21194's actually have higher gain and fT at 15-20 amps and better SOA at 80-100 volts. That means 5 pairs drives 2 ohms with some to spare.

Don't sweat the smallish caps. It won't blow up driving low Z in domestic use. It just won't have as good a low-frequency performance as it could have. Anyway, you'd be surprised to see what they get away with skimping on transistors, transformer, cap, and everything else in modern pro amps that are advertised as 2-ohm compatible.

 

[wg_ski]

One more thing.... not all or Slone's designs were tested prior to publishing. Make sure your front end is thermally stable before you go marrying it to an expensive set of outputs and drivers. One of the recommended practices is to dispense with the current mirrors when using complementary diff pairs. With a cascode darlington VAS you don't need the extra 6dB open loop gain on the front end anyway.

 

[nigelwright7557]

One more thing.... not all or Slone's designs were tested prior to publishing. Make sure your front end is thermally stable before you go marrying it to an expensive set of outputs and drivers. One of the recommended practices is to dispense with the current mirrors when using complementary diff pairs. With a cascode darlington VAS you don't need the extra 6dB open loop gain on the front end anyway.

I never power up my amps first time with the output transistors in the cct. I wire the driver output back into the LTP.
This proves the driver cct is working OK.

I always set the Vgs multiplier to give lowest voltage possible out.

I also always include an output relay married with a PIC microcontroller to monitor the output volts. If the o/p goes DC for more than 500ms the relay is switched off to save speakers in the event of failure. The same relay is also held off for 3 seconds to let the amp settle down.

 

[GK]

One more thing.... not all or Slone's designs were tested prior to publishing. Make sure your front end is thermally stable before you go marrying it to an expensive set of outputs and drivers.

This is the unworkable design. The voltage at the bases of the VAS transistor buffers and the quiescent current of the VAS is not defined.

Cheers,
Glen

 

[TonyTecson]

this is first time i see a sziklai connected predriver....would have been simpler to use triple darlington...

yes, this design is doubtfull........

 

[EWorkshop1708]

You mention you have a 55-0-55 transformer and want 300W.....

I've built amp really similar, so I can help.

Now your transformer voltage is a bit low and may sag, so you may not get 300W at 8 ohms. At 4 ohms, you can get that much no problem.

I currently have a DIY subwoofer amp that outputs over 350W+ RMS at 4 ohms. I'm also using a 55-0-55V transformer like you, and I get 75-0-75V DC +/- rails. I estimated this amp to be 200W/8 ohms or 350W 4 ohms. It uses 5 pairs of MJL4281/4302 for outputs, and MJE15034/35 drivers on the heatsink. 4 or 5 pairs (or more) should be plenty for your amp as well IMO. The 4281/4302 series are great transistors and will do fine in your amp too.

However, I completely burned up two subs with the 350W amp, so just to let you know, over 300W RMS is a lot of power, use it wisely!

IMO I'd recommend making an amp that can drive 4 ohms anyway, even if you use an 8 ohm sub, because you just never know what kind of speaker you may hook up to it in the future.

Have fun with your monster amp project!
EW

 

[TonyTecson]

The voltage at the bases of the VAS transistor buffers and the quiescent current of the VAS is not defined.

yes, although the current is limited to 46mA or so.....

 

[CBS240]

Hi

I think what Glen is getting at with his comment regarding the undefined VAS current is that with a mirror on both differentials, there is no reference to the VAS current to the input stage bias. If Q29 & Q7 happened to turn on a tiny bit more, maybe they warm up a little, then the resulting increase in VAS current is not referenced back to the input bias, therefore the dependent VAS current remains undefined. Undefined currents in solid stage typically cause saturation. Hence why you need the current sources to keep the VAS transistors from cross-conducting and releasing their magic smoke....and saturating (smoking) your outputs. Unfortunately, the CCS's make the output impedance of the VAS very high above 46mA. If the VAS current is limited to 46mA by the current sources that are now in series with VAS load, how can it supply AC current needed by the pre-drivers as is required? If you omit Q32 & Q37, you will have a cascode VAS which will work very well. But, you will have to balance the current via the input stage. For this you need a voltage reference related to the input transistor's current in order to properly drive the base of the VAS transistors w/respect to the VAS transistor's emitter. Replacing the current mirrors with ordinary resistors will accomplish this....and with less parts. But then, you will have a complementary balanced diff input with a cascode complementary VAS which is quite common....and is hard to beat when it comes to performance, IMHO.

 

[wg_ski]

this is first time i see a sziklai connected predriver....would have been simpler to use triple darlington...

yes, this design is doubtfull........

MJL21193 found out about this The Hard Way (t.m.) too. You normally don't want to go driving a huge output stage that way, because it usually takes several rounds of layout optimization to get it to quit singing.

 

[wg_ski]

Hi

I think what Glen is getting at with his comment regarding the undefined VAS current is that with a mirror on both differentials, there is no reference to the VAS current to the input stage bias.

It would either eventually thermally stabilize (if enough degeneration in the mirrors and VAS were used), or run away until it either smokes something or limits. The one thing you CAN'T count on to tell you which will happen is a spice simulation.

If you built it and fooled with it long enough, you could probably add enough thermal feedback to make it stabilize. But it would be trial and error at best, and you might not end up with the quiescent currents you planned on. It's far more cost effective to dispense with the current mirrors because with that much output transistor capacitance, something will have to give to make it stable anyway. The extra open loop gain you get with the current mirrors will end up being flushed down the john so why bother? Differential balance (which you always want) can be enforced by matching if the input stage devices, and trimming or careful matching of the degeneration resistors in their emitters.