100W HIFI Amp

[Aniket]

My next project, it's a low power version of my 250W amp that I recently made. 250W is a bit of overkill for home use and for normal casual listening hardly 100W max is enough.
So, I removed 2 pairs of output transistors, changed the driver and predriver stage. Lower the supply rails and now we have a decent 100W into 8 ohms amplifier.

I simulated the circuit in Multisim and the results are as follows:

Supply voltage- +/-50V DC
Power Output: 140W into 8 ohms load at clipping, 260W into 4 ohms load at clipping.
Distortion Figures:
THD1- 0.001% THD @ 1kHz 120W into 8 ohms.
THD20- 0.005% THD @ 20kHz 120W into 8 ohms.
Input Sensitivity:
1.6V for 120W into 8 ohms.

Although Multisim is a very conservative simulation software, it's simulation results are very much close to the real world. I have simulated every project that I made and they performed and measured very close to the actual results.

This amplifier would easily do 120W into 8 ohms with +/-50V supply and 160W into 4 ohms with +/-45V supply with the specified output transistors.

+/-60V is really the max supply voltage if you are using an 8 ohm speaker and at this supply voltage a 4 ohm speaker should never be connected, the output transistors would fail driving a 4 ohm load at that supply.

For a 4 ohm load +/-48V is the max supply voltage and of course you would need big heatsinks and fan cooling is advisable.

Designed the PCB in sprint layout and I have integrated the Power supply on the Amplifier board itself. So, in a stereo amp, each amp module would have it's separate Bridge rectifier and filter capacitors for minimum crosstalk between the channels. Of course, independent power transformer is also an option.

Here's the schematic:

PCB:

Regards & Cheers !!!!
Aniket

 

[SoundCraft]

Aniket thanks for the same,Am very much interested in building the same ! But have a request,can you please add a protection circuit to the same ? its much safer to have it.

Also please post the BOM and approx cost of the whole thing,heatsink needs etc.

Are the output devices easily available ? if not what substitutes can we use ?

How do you describe the sound as ? I keen on this too.

Thanks once again for posting it,Once I get the info required I would request you to kindly mail me the gerber files for me to go ahead with the project.

 

[Aniket]

Aniket thanks for the same,Am very much interested in building the same ! But have a request,can you please add a protection circuit to the same ? its much safer to have it.

Also please post the BOM and approx cost of the whole thing,heatsink needs etc.

Are the output devices easily available ? if not what substitutes can we use ?

How do you describe the sound as ? I keen on this too.

Thanks once again for posting it,Once I get the info required I would request you to kindly mail me the gerber files for me to go ahead with the project.

Hi SoundCraft,

Protection circuit could be made on a small separate PCB with it's own small transformer, which could be used to power a fan if you want to have.

Check this link. http://www.diyaudio.com/media/build-guides/diyaudio-softstart-build-guide-v1.0.pdf

I am using this speaker protector on my amps, it works good with delayed switch on. I am using a small 12V transformer to power it. Also u could use any general purpose NPN transistor, eg BC548/547/546,2N5551 and BD139 to drive the relay.

I would upload the BOM soon. after I procure the parts.

The output devices I have used are ON semiconductor NJW0281/0302. These are 150W devices and are available in Delhi for 300INR a pair.
A number of other transistors could be used as outputs, like
2SC5200/A1943 or TTC5200/A1943, 2SC5949/A2121 from Toshiba,
MJL3281/1302, MJL4281/4302, MJL21193/21194, NJW3281/1302, MJW3281/1302 from On semiconductors,
2SC3263, 2SC3519, 2SC4468, 2SC6011, 2SC6145 from Sanken,
2STC5200/A1943, 2STC5949/A2121 from ST semiconductor,
Also, 2SC4793/A1837 and 2SC5171/A1930 from Toshiba could be used as drivers.
For VAS duty, i.e Q4, Q7 KSC3503/A1381 could be used for even better performance.

Regarding the sound quality, I haven't made it yet, but surely it would have inherent sound quality of it's higher power version 250W Amp which I made recently.
I won't claim that it's the best sounding amplifier in the world, as it's purely subjective and depends on individual tastes, but surely, technically it's a good amplifier with really low distortion.
For me here's the description of what it sounded like:
1. The highs were very crisp and detailed, not harsh but as sharp as pin.
2. Mids are very smooth, open and very detailed.
3. Bass is really where this amp shines, a lot of authority and control over the cone movement. With my main speakers, the bass punches in the chest like a hammer.

Above all, the final sound quality depends on how solid your power supply is. It takes power to create power.

Would update soon.

Regards,
Aniket

 

[linuxguru]

So, I removed 2 pairs of output transistors, changed the driver and predriver stage. Lower the supply rails and now we have a decent 100W into 8 ohms amplifier.

Thanks for the excellent design - this is a very practical triple-darlington version of the Blameless, with a Type-II EF for the driver.

If you haven't yet finalized the PCB, a few minor (and major) improvements are in order:

1) Separate the grounds into Signal GND (SGND) and Power GND (PGND). The two are connected at a single point with a 1 to 10 ohm resistor. SGND has all the ground returns related to the input LTP and feedback network connected to it, and nothing else. PGND has all the other ground returns, including speaker, zobel, bypass caps, etc. The trafo 0V (centre tap) is connected to PGND. The input signal ground from the RCA goes to SGND.

1.1) Isolate Chassis ground (Earth) from PGND (i.e. Trafo 0V).

2) Shift R27 between R2 and C2.

3) Omit R13 (the 2N5401 can easily handle 2 mA at 50V Vcb).

4) (Requires simulation and optimization) Use Cherry-style compensation for the VAS, with a single ~100pF capacitor between the output and the base of Q5.

5) Consider using Leach-style RC lead compensation from the middle of R15 (split into two resistors) to the base of Q2. This requires some other changes to the feedback network that can only be optimized after simulation. A starting value for the RC is 22k in series with 47 pF.

6) Add a jumper between R11 and the base of the Vbe multiplier Q6. It can be shorted (to retain the present schematic) or have a 3..4 diode chain connected in series to obtain a discrete Thermaltrak-type compensation network. The diode chain is kept in close thermal contact with the heatsink.

7) Change the VAS transistors Q7 and Q4 to the BF469/470 or BF871/870 pair (the pinouts are probably different). Alternatively, allow both pinout options to be used by using two footprints with the appropriate connections.

8) Optimize the ground return of the Zobel cap C9 to be physically close to the speaker return (the speaker return can be moved close to the Zobel). Also allow for wider (~8mm or wider) capacitors at C9 - the best zobel caps are film/foil caps like the ERO MKP1837, which are physically wide.

 

[Aniket]

Thanks linuxguru,

While designing the PCB and parts selection the foremost thing was to make it DIY friendly.
Earlier, i was considering lifted ground for the signal and feedback path. Thanks for reminding me. updated in the PCB.
Although R13 isn't absolutely necessary, but it limits the current through Q5 thereby preventing over driving VAS(Q7) during clipping. So, I am leaving that in place. No harm in having it.
I am skipping points #4, 5, 6 mentioned by you. It's a TEF output stage, if it oscillates it's hard to stabilize, so I went away with standard miller compensation scheme. Also, the amp is stable upto 47pF for C3, below that it oscillates. Surely, would try other compensation schemes in simulation and post results.
For VAS Q4,Q7 MJE340/350 are used only because part's availability is a issue in local market. However, there are a number of devices in TO126 package that serve excellent VAS devices. Including BF469, you mentioned.
KSC3503/A1381, 2SC3790/A1480, 2SC3902/A1507, 2SC3601/A1407, 2SC2912/A1210 are great VAS devices. but again part sourcing is an issue.
Even BD139/140 could be used for rails less than 42V.

Would post the updated PCB soon.

Regards,
Aniket

 

[pranavkumar]

Looks like a great amp !
waiting for the listening impressions

 

[linuxguru]

I am skipping points #4, 5, 6 mentioned by you.
....
For VAS Q4,Q7 MJE340/350 are used only because part's availability is a issue in local market.

Please implement point #6. A jumper can easily be used to omit the thermaltrak Vbe multiplier compensation, or use it if the builder so desires.

BF469/470/471/870/871 and related parts are still available in the market, because there's surplus left over from TV/CRT manufacturers after the latter became obsolete. It won't be available forever, but at the moment they're easier to find in Chennai/B'lore than MJE340/350, and also competitively priced.

They're also better VAS devices than almost anything else because they have moderately decent beta (100..150) and extremely low Cre at ~2 pF.

For low rails in the region of +/-24..36V, BF422/423 may also be usable for the VAS and pre-drivers.

 

[shibashis]

I am not a DIY guy primarily because I do not have any knowledge of Electronics, I come from a very different background, but would be very very interested in a tentative price.
Thanks to all involved, I will keep an eye on this.:thumbsup:

 

[captrajesh]

Wrong post in the thread. Deleted

 

[Aniket]

Attached is the proposed component list and Bill of Materials for a stereo amp(2 channels).
Prices are an estimate and may vary for different markets. These are for Delhi.

Cheers!!!!
Aniket

 

[Aniket]

Updated and possibly final PCB files

[/URL][/IMG]

[/URL][/IMG]

[/URL][/IMG]

 

[Aniket]

In case 2n5551/5401 aren't available,
BC546 could be used for Q1, Q2, Q3 and BC556 for Q24, Q25, Q5 as substitutes.
There are a lot of other devices also. but, we could get the BC's easily anywhere.
For rails <=40 volts, low noise BC550 and BC560 could also be used for Q1,Q2,Q3 and Q24,Q25,Q5 respectively,
and BD139 for Q4,Q8 BD140 for Q7,Q15.

Regards,
Aniket

 

[linuxguru]

Looks workable, though I haven't checked all the small-signal routing and connectivity. I noticed that a lifted small-signal ground has been added, which looks OK. It won't hurt to make that a rectangular or polygonal ground plane/fill below the 3.3uF cap, 27k resistor and similar input side components (this is easy to do in Eagle, I don't know about Sprint).

However, you can optimize the star ground a bit. The track that connects the ground-lift resistor to the power ground (just to the bottom-right of the -ve rail capacitor) can go much further and below the 0v connector of the trafo, without much impact on the high-current flows. The closer it gets to the 0v potential without being disrupted by high-currents from the PSU filter caps and other bypasses sitting on PGND, the better - this trace ultimately defines the lifted ground potential, so it should be kept as quiet as possible. I would move the trafo 0v connector below/between the PSU filter caps, and bring the above trace as close as possible to it.

I also noticed you've made the zobel cap longer, i.e. increased the pin pitch to 7.5 mm (?). That's OK, but my earlier suggestion was to make it wider, i.e. fatter - at least 8mm. If you can, try to provide multiple pin pitches - 5 mm, 7.5 and 10 mm. There are a lot of MKP-X2 caps in 10mm pitch, as well as a few high-quality caps like the MKP1837 in 5 mm pitch.

The rest of it looks OK - the BF470/471/870/871 ECB pinout will probably work interchangeably with the MJE340/350 pinout (may need a 180-degree rotation during assembly). The pinout issue will also affect swapping of small-signal transistors with BCE pinouts with those having CBE pinouts (most of the BC5xx series).

The 2N5551/5401 usually have far lower DC beta (~100) compared to the BC5xx (~300..400). The latter would be preferable at high gain locations like the LTP and VAS buffer, but the pinout issue will make assembly messy.

 

[kanwar]

The 2N5551/5401 usually have far lower DC beta (~100) compared to the BC5xx (~300..400). The latter would be preferable at high gain locations like the LTP and VAS buffer

Not always...... When operating at high rails[along with cascoding], i prefer low beta ones but tightly matched, less prone to oscillations and not needing much heavy emitter degeneration as well. Equally i keep the input bases both at low impedance to keep the noise floor low, last time i implemented this got 120+dB SNR. . Another added advantage is lesser inter-modulation products.

 

[Aniket]

Looks workable, though I haven't checked all the small-signal routing and connectivity. I noticed that a lifted small-signal ground has been added, which looks OK. It won't hurt to make that a rectangular or polygonal ground plane/fill below the 3.3uF cap, 27k resistor and similar input side components (this is easy to do in Eagle, I don't know about Sprint).

However, you can optimize the star ground a bit. The track that connects the ground-lift resistor to the power ground (just to the bottom-right of the -ve rail capacitor) can go much further and below the 0v connector of the trafo, without much impact on the high-current flows. The closer it gets to the 0v potential without being disrupted by high-currents from the PSU filter caps and other bypasses sitting on PGND, the better - this trace ultimately defines the lifted ground potential, so it should be kept as quiet as possible. I would move the trafo 0v connector below/between the PSU filter caps, and bring the above trace as close as possible to it.

I also noticed you've made the zobel cap longer, i.e. increased the pin pitch to 7.5 mm (?). That's OK, but my earlier suggestion was to make it wider, i.e. fatter - at least 8mm. If you can, try to provide multiple pin pitches - 5 mm, 7.5 and 10 mm. There are a lot of MKP-X2 caps in 10mm pitch, as well as a few high-quality caps like the MKP1837 in 5 mm pitch.

The rest of it looks OK - the BF470/471/870/871 ECB pinout will probably work interchangeably with the MJE340/350 pinout (may need a 180-degree rotation during assembly). The pinout issue will also affect swapping of small-signal transistors with BCE pinouts with those having CBE pinouts (most of the BC5xx series).

The 2N5551/5401 usually have far lower DC beta (~100) compared to the BC5xx (~300..400). The latter would be preferable at high gain locations like the LTP and VAS buffer, but the pinout issue will make assembly messy.

Thanks a lot linuxguru,

Your name should be hifiguru, anyways enough of teachings now. The zobel cap pitch is 10mm.
If anyone is sourcing parts online through Digikey, Mouser, Arrow or any other online electronics store then, a lot better components could be used in LTP, CCS and VAS stages.
But what I always embark upon is the easy availability of the parts in local market. I know the specs of 2N5551/5401, they work good and are widely used in lots of hifi and commercial amps.
I always follow the Idea- "Keep It Simple".
I wanted to keep the layout DIY friendly, a lot more could be improved in the schematic and layouts including a number of artifacts, but the marginal performance gain we would get is not worth the added complexity.

Regards,
Aniket

 

[Aniket]

Hi,

It's been a while since I posted. With some suggestions from linuxguru and in order to increase the performance of the amp, i was working on the amp and upgraded a few things.
Improved the CCS for the LTP and VAS. added Q12 and Q13.
Added speed switch off cap, C20 across R16 and a little more changes. What I was missing till now, was to analyze the clipping behaviour. Before the above modifications there was rail sticking on the -ve rail during clipping and a little oscillation, because of the saturation of the VAS transistor, Q7.
Now, with the above mentioned changes the clipping behavior is very good, no rail sticking and oscillations. THD performance has also improved. 0.005%THD @20kHz 135W into 8 ohms load.

One more thing i was thinking about was to separate the Power supply from the amplifier PCB. With this option I think we could design the power supply as we want, no constraint on the capacitor choices, dedicated 35A bridge rectifier and 3~4 caps in parallel to get as much capacitance as the builder wants. I am working on new separate PCB's for the amp and the power supply. would update soon.

Updated schematic

[/URL][/IMG]

Clipping at 20kHz

[/URL][/IMG]

Regards,
Aniket

 

[SoundCraft]

Hey good to see the changes implemented !! Linuxguru knows his stuff,been reading bout his stuff over the DIYA,so for sure this now will benefit from the changes which already evident !

I had already requested and will do so now,since you have removed PSU can you please,please add the protection circuit to the main board ??

Would be super great if that is included,Will start as soon as that is updated.

 

[Aniket]

I had already requested and will do so now,since you have removed PSU can you please,please add the protection circuit to the main board ??

Would be super great if that is included,Will start as soon as that is updated.

Although protection circuit could be added on the amp PCB, there are two ways to power the protection circuit. 1st is to take the power from the amp rails itself and using a regulator to get 12~15V DC to power it, 2nd is to use an auxiliary power supply with a small 12V transformer which could also be used for soft start and fans.

Another option is there which I believe is the most relevant is to integrate the Protection and soft start circuit on a separate board with its dedicated power supply and keep these modules far off from the amp and/or preamp circuits.
By this way you would have an overall modular construction of your amp, which would be easy to assemble and troubleshoot.
If you want I could design a PCB for the integrated soft start and protection circuits.

Regards,
Aniket

 

[pranavkumar]

I think, going the modular way is more logical.

 

[SoundCraft]

Another option is there which I believe is the most relevant is to integrate the Protection and soft start circuit on a separate board with its dedicated power supply and keep these modules far off from the amp and/or preamp circuits.

Yes this is most used (as I've seen) but my request was purely to keep the protection circuit close to outputs as the trace is not long,if we have a separate PCB then the wires from output to that and from there to terminal,to me this is bit lengthy process (too much wires degrading the sound)

But if you think that is the logical way then please do that itself and share it.