Sound Quality Vs. Measurements

Status
Not open for further replies.
RXvyHfR.jpg


Well I made a start . It works and for a 15 inch bass it ticks the excellent box already . At 10 kHz it is OK and 22 kHz it is not .

It uses every bad idea I could muster and alas lives up to expectations .

It is stable . MPSA 42 and 68 pF . A very cautious start so as not to have smoke . 5 transistors total . 80 watts .

LTP current is a modest 1 mA and is using a tail resistor . Although calculated balance it must be about right as it has very little second harmonic . 9 mV offset . Gain is 48 . Inputs are 2SA970 .

I will work on to get 22 kHz right . Other than that a qualified success . Hum is rubbish . Looks a mess .

Correction . Just redid hum - 90 dB ref 1 watt 150 Hz ( worst ) . Not bad for a start and remembering 33.6 db gain . The gain will come down as step one to lower distortion .
 
Last edited:
VLBaXWG.jpg


As said this is not intended to be a full frequency amplifier ( the Mad Max as it will now be called ) . The 10 kHz is now good enough to need to use the Audio Precision . The 22 kHz is 0.3 % second harmonic is not too bad . Looking at it up and down the power range nothing bad happens which says it is just VAS as far as I can see . 2 pole compensation and going down to 15 pF was not the answer .

This amp goes against the grain with me . The VAS to LTP driven by 1K6 is pure trans-conductance ( u = 100 ) . I have a 2SC2910 of 3 times the gain . That should take it inside my rules . BTW . At LTP 1mA re = 50 . That is a minor reason to stick with it . If someone would prefer to use MPSA 42 I would raise LTP to 3 mA . That makes the collector resistor 510 R . I would add 39 R emitter resistors to the LTP . Move VAS cap to 68 pF at the start .

Everything was done to be " wrong " here . Wrong with OK maths . If I can get the 22 kHz distortion down I think I might use as a main amp for a while . I have a hunch it will be very detailed and sweet . As far as I can tell no obvious crossover distortion problems at any power level into noise . 220R was the resistor nearest to the bias point . Just think, one resistor . Bias failure can not happen .
 
Founder of XSA-Labs
Joined 2012
Paid Member
RXvyHfR.jpg


Well I made a start . It works and for a 15 inch bass it ticks the excellent box already . At 10 kHz it is OK and 22 kHz it is not .

It uses every bad idea I could muster and alas lives up to expectations .

Nice work Nigel! I love point-to-point aka dead-bug style circuits. Really, they are the epitome of efficient network wiring and the same way neurons in our brains are set up. They may look fugly, but they can work really well.

Here is something I did with a TPA3118D2 class-D chip and very small SMD dead-bug style on foam core board: http://www.diyaudio.com/forums/class-d/219730-tpa3118d2-6.html#post3413886

It's ugly but sounds great.

336455d1363460094-tpa3118d2-foam-core-amp4.jpg


With tiny 0.5 in copper tube ring heatsink (good for 30 w/ch)

340776d1365224557-tpa3118d2-image.jpg


I have since made a proper PCB version using a TPA3116D2 (good for 50 w/ch) doesn't sound better just capable of higher power.
 
Y6x7v7Q.jpg


10 dB improvement since last night .About at the end the road for a improvements . This one came at the expense of DC offset . Unity gain sable I suspect under 100 pF . To stress . This amp will live it's life between 5 and 1000 Hz . The 56 Hz filter works great . When unity retain the 1 K . Big surprise was a constant current source outperformed by a bootstrap . No high grade parts used . Bipolar caps sometimes outperform high grade .
 
NTFoddR.jpg


Here at last the comparison I wanted to make . The Hitachi and my 5 transistor version . 2 dB better distortion on Hitachi . I know the Hitachi is far better than that so checked the oscillator . It needs a tweak . No worries the distortion helped find the little there is in the 5 transistor . The 5 transistor will have miniscule distortion if I can get it checked on an Audio Precision set up .

Now the big deal . The Hitachi I show has added CCS for the LTP . 5 transistor none and higher gain ( 49 verses 34 ) . Hum on 5 transistor is 5 db better . Identical bias current used and same outputs . In fact the Hitachi PCB is in the one above .

The circuit has additions for bass speaker filter . It is unity gain stable .

Bootstrapped input works if wanted . That would help a passive preamp . Values need tweaking as they were the parts to hand . Big surprise bootstrap better then CCS for VAS .

5 transistor is marginally more powerful .
 
250 V FET's and MPSA 42's . The other transformers were to get the voltage up . It wanted one small digit change give it's real name ( 0.4 to 6 ) . It only liked resistive loads . It worked in the end .

xrk971 . Thanks for posting the dead bug class D . Wow . Saw it the other day . My Boss is just back from Switzerland . Good timing as I will have to get busy . I told him I have been training on a bicycle this week rather than the running track ( mentally ) . He sort of accepted it . I measured a few chokes for work which to be honest isn't that difficult .

I had so many my dead bug projects . Now where are they ? Some where even neat .
 
Last edited:
kz133Wu.jpg


I did some maths and worked out this just might work . The noise from 68 K will be high with a bipolar op amp . It is below the distortion so will do nicely . If a lower value it loads the 1 st order filter . I had gambled on some cancellation between the inverting and non inverting stages . Looks like I got that as measurement is below theoretical level ( about 5.4 db more like 10 dB second seen ) .

I only had 5532 . It's high current output is useful . The previous oscillator with TLO71 was giving 4 % at 46 kHz .

RA 53 was a self heating negative temp co thermistor ( 5 K ) . A 3 pole 4 way switch gives 1kHz 10 kHz 22 kHz 46 kHz . The red resistors get changed . The for example 500 Hz filter for 1 kHz ensures the slope is steeper .
 
5heiG3H.jpg


If wanting a simple workshop oscillator this works well . The 22K with 2 x 1N4148 are arranged to slightly reduce the squareness of the 1N4148 used for limiting . It might be reduced further ( 47 K seems OK ) . 10 K and 91K form a null for the 3 rd harmonic . This raises the 5 th a little . However the 5 th is well below the 3 rd . > - 62 dB ( 0.08% ) . With simple filtering maybe - 70 dB ( 0.03 % ) ? From the time when I put blobs on the joints !

It is probable that increasing the gain ( 33 K 1 K ) could get a little more out of the design .

34 K and 4n7 are the components setting the frequency . 3K4 4n7 = 10 kHz ( approximately ) .

The last op amp could be with passive filter before it . If so output will vary a little , probably wort it . If wanting 8 Vrms +/- 15 V supplies required . Some op amps need capacitors in the gain setting arms to work properly . The NE5532 was most obliging considering I asked it to do things it is known to be fussy about .

This is a variation of the Rosen circuit . Vastly simplified and lets face it not as good . Still 30 dB better than typical bench oscillators costing $200 used in schools . I have one because it has an excellent frequency counter . Also it is rock solid in voltage output to 5 MHz . Still annoying it has 1% THD .

If anyone has a circuit giving - 90 dB up to 200 kHz which is simple that would be interesting . There is a variation of Rosen using a light dependent resistor ( like in old science kits from Radio Shack ) and LED . That looks favorite .

From this design there is also a Cosine output . It has more distortion . If filtered it will be low enough . Ideal for driving a synchronous motor . A 12 V 7.5 stepper works nicely at 4 V rms ( sine and cosine required ) . A 7.5 degrees type is almost identical to a synchronous and much cheaper ( $10 ) . I suspect the first steppers were just synchronous sold with a different label ?
 
You could use a tracking filter after the output. It worked pretty well the one time I did it. It used several stages of simple identical first-order lowpass filters, with op amp buffer between each. The R of the RC was a VTL5c2 Vactrol, which is just a CdS light-dependent resistor encapsulated with an LED. But the R vs LED current is non-linear. So the trick was to wrap a feedback loop around the whole set of filters, with basically an averaging envelope follower in the feedback path, to detect the output amplitude, which then set the current to the LEDs. The way it worked was to crank down the cutoff fequencies until a preset amplitude was reached at the output. It was set up so that always resulted in the cutoff frequency being below the fundamental, somewhat.
 
Last edited:
OP2uUM7.jpg


Thank you . The Rosen comes up again Jesper . I am rather surprised by my 5532 version as 8V rms is at the top of the op amps range ( 80% ) . I now have the 47 kHz - 73 dB which is not too bad . There must be cancellation as the drop from - 55 dB to - 73 is way beyond what a first order filter will do . Thinking of that my SVF might do better with a non inverting final stage . The more I think about it the more I think what I did with the 5532 RA53 might work with the SVF . It was better than the 5532 pre fiter . Look how well the worlds most evil op amp the LM358 works in place of 5532 !

Here is the 1979 Hitachi design measured ( Exicon 10 N / P 16 2SA872A 2SD756 2SB716 ). Only 7 active devices . I feel I need - 90 dB before results are reliable . The Hitachi is like this at 90 watts also and down the scale ( 10 mW ) . I choose 5 watts as the oscillator can be set on the same scale of the analyzer 6.35 V rms . The output is via 5K 10 turn from my scrap box ( Spectol 2 watts wire wound , inductance seems OK ) ,very useful .
 
Status
Not open for further replies.