New design 135W 8, 270W 4

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi. I've been designing this amplifier for a few months now, and I've gotten it to where I like it and want to go ahead building it. It will take a maximum of Vcc = |Vee| = 50V supply, and is designed around this.

It should do 135W in 8 ohms and 270W in 4 ohms.
Simulated THD is about 0.002% at 12V peak in 8 ohms.
Input impedance should be greater than 1Mohm.
Gain is 32dB.

What do you guys think?

Symm135.png
 
Hi Guys

There are a lot of things that seem strange - an odd mix of antiquated parts and modern circuitry.

First, why TIP outputs and drivers? I assume it is because you already have them, but these are slothy devices. It is no wonder THD is so high in the sim.

Second, I would lose the networks tied to the diff amp emitters. These look like they effectively bypass the current source and defeat part of its function. They also contribute potentially to poor DC balance and high output offset. These networks make the input impedance high in real resistance, not just as potential bootstrapped loading, and it does not match the feedback impedance.

Third, the feedback network values are odd. R14 is pretty high, although typical for a 1970s hobby circuit, and will contribute to the DC offset and to output noise. Modern amps use a much lower series value with a shunt element sized for the gain needed. The shunt cap (47u) is ridiculously low and will contribute to the THD. DO NOT use this cap position to shape the frequency response; just use it to roll DC gain off to unity.

Fourth, there should be a cap across the bias regulator.

Fifth, there should be decoupling between the output stage and the front-end.

I've never been a fan of the TO-126 MJE devices as there is little or no information regarding their Ft. people have made measurements of samples to develop models for simulation, but they are no where near as fast as the 2SA1381-2SC3503 pair or similar Japanese drivers. I also would not use the 2N5401-2N5550 for the diff amps although they are quite linear. usually you want input devices with higher beta to draw lower currents from the input base-leak and from the feedback path.

I also place a fixed R from the output to ground to help settling and to provide referencing for the input stage and output stage. Every direct-coupled opamp app and PA should have this.

Take of any of this what you will.

Have fun
 
Thank you all for the useful thoughts. Struth, I agree with most of what you say, I will investigate the other things. The circuitry on the differential stage is a boot strap, and simply raises the input impedance. But i don't make use of this by lowering the feedback resistor. And yes, the 70k DC distance isn't balanced with the feedback resistor - I shouldn't have missed that one.

In the past I used BC546/556 for small signal, but i found the 5551/5401 to be more versatile. I will investigate other options. I don't like the MJEs at all, but I've found them better than BD140 and other BDs. I found some nice drivers recently, which i use rather than the MJEs, I just don't have their model on the computer I stimulated on. I will do better simulations.

I really like the TIP41/2s. I've had the best results with them so far. I don't have much choice in south Africa - I tried some new output transistors, and they all turned out to be fake. The TIPs are a sure bet. I went through the trouble of getting some 2sa1943 and 2sc5200, and they cost me an arm and leg, but i was really not impressed. The same amplifier with TIP35/6s in did much better.

I will make some changes and do some simulations. Thank you for the insight!
 
Hi. I've been designing this amplifier for a few months now, and I've gotten it to where I like it and want to go ahead building it. It will take a maximum of Vcc = |Vee| = 50V supply, and is designed around this.

It should do 135W in 8 ohms and 270W in 4 ohms.
Simulated THD is about 0.002% at 12V peak in 8 ohms.
Input impedance should be greater than 1Mohm.
Gain is 32dB.

What do you guys think?

View attachment 554100
I think you mean a dual polarity supply with ±50Vdc.
That will not give you 135W into 8ohms and certainly not 270W into 4ohms.

If you build it to deliver all the current that a low impedance could demand then you should get around 110W into 8ohms and 200W into 4ohms. 2pair of 230W devices will do this. But if you want good reliablity driving 4ohms hard, when the temperatures are summer time, maybe consider 3pairs.
but more likely you end up with ~100W into 8ohms and ~170W into 4ohms.2pair of 200W devices should be reliable for this.
 
Last edited:
Thanks for posting a 1 mohm input impedance amp. 47k input impedance will not run from a 12AX7 output preamp, and my PAS2 is still the low noise king around here. Bootstrap input huh? I'll have to read up on that.
I Like the use of all the commonly available parts. 2n5401 & 2n5551 are long term production parts usually available. TIP42c is about $.33 and those aren't rejects. I blew $50 at newark buying 25 (obviously rejected) NJW21194G OTs which were rated 200 mw and that experienced proved was not a misprint. They blew up immediately after putting 85 v on them.
If I can beat 0.2% THD and 60 W/ch 8 ohm it would improve on my dynaco ST120 circuit. .002% THD is rediculously low IMHO for driving speakers, no problem.
 
Last edited:
Hi indianajo. I'm glad you like it. I need to make a few mods still, and i wouldn't recommend building this one yet. I want to build a prototype sometime to be confident it will run smoothly. I want to still investigate the effect of changing the 10-100 ohm resistors, as these could possibly affect thermal stability and output bias vs driver current. But i definitely want to build one soon. Also, I would recommend TIP35/6 if you are thinking of driving 6 ohms nominal or less.

I will post a short explanation of the input bootstrap soon. I love the concept.
 
Using simulations only (simply to show the concept).

Consider:
An externally hosted image should be here but it was not working when we last tested it.


Simulations show an input impedance of around 3.9k. Adding a bootstrap:

An externally hosted image should be here but it was not working when we last tested it.


Raises the input impedance (simulated) to about 26k.

The reason is that the AC voltage at the junction of R3 and R2 is the same as the AC at the emitter, so V=IR, where V would be very low, so current is reduced dramatically, and thus R increases. Increasing R3 can further increase input impedance, but I'm not sure what the effect of this is overall. The reason it increases input impedance is that it increases the resistance path from input through R3, through C1 and through R4 to ground. With R3=10k, input impedance is 190k.

This concept is applied to the differential amplifiers.

And that's about the extent of detail I'm willing to go into. There's plenty to find online. :)

FWIW, distortion of this circuit is very low (with R3=10k), considering it's a basic CE amplifier. With 1V peak output, distortion simulates to a 2nd harmonic at -53dB and a third at -75dB.
 
Last edited:
I thought in my ignorance the high input impedance came maybe from the Q9-Q10 dual input transistor configuration. Or maybe from the Q11 Q12 boost transistors before the VAS Q1 & Q4 .
In my AX6 copy for the ST120, I'm trying to fit in a 1n7000 nfet ahead of the input transistor as an impedance increaser. 100 kohm drain to +30v, 100 kohm source to ground, gate to capacitor to the outside (preamp 12AX7 plate through cap) drain to the input base of the AX6 normal circuit. Same sort of follower function that the 12AX7 has in the preamp, I've never seen anything like that, but nobody seems to need a 1 megohm input impedance amp board but me. In South Africa you could use a salvage nfet from a PCAT power supply blown up, to get some genuine part. A 1n7001 has much lower gate capacitance to avoid loading down a 12AX7 which is loaded enough by the 6' of coax RCA cable. No, I don't have +-15v PS to install an op amp in my ST120. The +30v is derived from the +72 of the St120 with 2 30 v zeners as regulators for the +60 for the input transistor for the AX6. Besides I've got 1 square inch of perf board to squeeze this impedance booster into. The whole AX6-ST120 driver card is 3.5"x5" with the output transistors remote on the heat sink.
 
Last edited:
Ok, cool! I was going to ask why you don't use an op amp buffer.

If you do want to use an op amp buffer, you could do like this:

An externally hosted image should be here but it was not working when we last tested it.


R3 and R4 must be 0.5W at least.
Transistors must be on a small heatsink each, and I would recommend the same for the regulators.
Zeners should ideally be 0.5W.

It must be noted that running a preamp off the same power supply as the power amp can cause some grounding issues. You're better off with a small, separate power supply if you want to use the op amp buffers.

I will be doing a prototype of the amplifier sometime soon. I'll post when it's tested.
 
Last edited:
Back to the amplifier, I changed R12, R14 and R30 to 15 kOhm, and I changed R15 to 390 Ohm. I also changed C2 to 470 uF. Performance is similar, but DC offset is decreased from 80 mV to 10 mV in the simulations. In simulations I tried BC547/557, also with no effect. I will think about everything before I build. I like the BCs anyway, and I have. I think I will use TIP41/2C on the output and drive only 8 ohm for now. They're much cheaper than TIP35/6C, and I'm on a tight budget with a baby due to arrive very soon. :)

I will simulate replacements for the MJEs sometime. I will use 2SC4793/2SA1837 rather (I looked up the model number).
 
With this style of circuit you will end up with stability problems .... Using slow output devices might get you out of there , but there is a trade off between sound quality and stability ..

Working like that you are pushing a simulator to situations that cannot predict .
A EF3 circuit will have stability problems on its own but still be very forgiving and tolerant to many mistakes CFP like you do it i think not ...

benefits of CFP ( i think ) cannot be reproduced as its done in EFP ...cook there is very simple IE increase rails add output transistors CFP doesn't work like that ...
Kind regards
Sakis
 
With this style of circuit you will end up with stability problems .... Using slow output devices might get you out of there , but there is a trade off between sound quality and stability ..

Working like that you are pushing a simulator to situations that cannot predict .
A EF3 circuit will have stability problems on its own but still be very forgiving and tolerant to many mistakes CFP like you do it i think not ...

benefits of CFP ( i think ) cannot be reproduced as its done in EFP ...cook there is very simple IE increase rails add output transistors CFP doesn't work like that ...
Kind regards
Sakis

Yes, I've investigated a couple of things, and this one isn't stable at all. I'm scrapping this design. The concept I started with worked better than I expected, but the changes I made don't seem to do well on closer look.

Thanks!
 
Well, the positive feedback input could add to instability, and I don't quite see the need of a tip42c gain booster ahead of triple tip42c outputs, when you've got mje340/350 as drivers.
Still building an amp out of 2n5401/5551 + Mje340/350 + TIP42c should be doable. These are things I can actually buy instead of all these exotic 2sc/2sa stuff which would mostly be counterfeit in the US market.
SOA on real mje340/350 looks like 300 ma @ 50v and 250 ma @70v. SOA on TIP41c looks about .7 A @ 50v with a gain minimum of 30 at .3 amp. So each output pair draws .02 A at maximum output soa current, so you might be able to drive 10 parallel pairs output just with a MJE340/350 driver. That would give you 7 amps out @ +-50 v which would be some useful wattage. More wattage short term ("music power") but not more than 7 amps for house music or low peak music.
Look at this thread http://www.diyaudio.com/forums/soli...across-emitter-quasi-complimentary-stage.html
which is a **** simple amp worked out in both single supply and split supply versions . I'd say just parallel the outputs, and forget all those exotic BC*** transistors which Europeans are so fond of.
I've got a schematic on a 200W amp from the 80's which is just sort of this arrangment, with 2 pairs output 2n3772 instead of one.
Note to jumpers in, mrclock lives in South Africa where he can buy counterfeit transistors of any marking he wants, but hardly anything real. I'm familiar with trying to push things into Mexico, a lot of stock gets changed at the border unless both parties are somebody important to the customs guys.
BTW I have a single 80 v supply in the amp I'm upgrading, and no room for the circuitry in post 15. Thanks anyway. I used up the regulator board space to put in 5 parallel TIP142's to regulate down to 72 v at 7 amps continuous.
 
Last edited:
Thanks for the reply and the link indianajo. To me, if the cheap, available and guaranteed transistors work, then why not? I try to make my designs transistor independent, obviously within limits.

The one major place using a BC547 over a 2N5550/1 has huge advantages is in the bias control. The BC has a much more desirable transfer characteristic for this purpose - the Vbe vs Ic curve is less steep, and thus adjustment is much easier, and the bias won't jump as much.

Well, the positive feedback input could add to instability, and I don't quite see the need of a tip42c gain booster ahead of triple tip42c outputs, when you've got mje340/350 as drivers.

The positive feedback for the input bootstrap has no effect on stability, you just have to make sure it operates well below the spectrum of the amplifier - so the input capacitor and bootstrap capacitor need to be quite large. But the possible response peak isn't high, and is at 50Hz or so. No issues with the bootstrap. I agree that no extra gain is needed. I was experimenting with the output stage.

Still building an amp out of 2n5401/5551 + Mje340/350 + TIP42c should be doable.

It most certainly is more than doable - I've done it with excellent results. But I must say I don't particularly like the MJE's, however suitable they are. I rather use 2SC4793/2SA1837 because they are easy to mount and require no insulation, and they work well.

I have changed the design significantly, and I will post the schematic soon, I just don't have it with me at the moment. But it's much simpler, should be perfectly stable, has excellent rail-to-rail performance, and has a perfectly acceptable THD figure (best at 0.001%). And I've built a low power version which blew me away (using MJE's :) ).
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.