Revisiting some "old" ideas from 1970's - IPS, OPS

Yes, as Nigel suggests you need a safety margin for the caps. Most areas mains voltage is +/- 10% so you need to calculate that into your supply voltage. 40VAC is about the limit for 63V supply caps.

I don't think we ever did give a suggested supply voltage in our guides anywhere. The three Sanken output pairs can easily output 300W though. I think I have 40VAC transformers in the amp I have in my system right now (can't remember for sure). It's in a 4U chassis and doesn't warm up properly. It does play fine and sound good, but it's on 102dB/W speakers.

I haven't gotten anywhere with the tube preamp. We've been busy with a few other projects that have been taking a lot of time to complete.
 
  • Like
Reactions: 1 user
I usually use a 40VAC transformer with 63v capacitors.
Off load sometimes the transformer output can be a little high.
Also mains voltage fluctuations need to be taken into account.

I took the fluctuations in account and due to protection the few second (when powered up) the transformer (35VAC) rises slowly till it's connected giving up 47VDC. So even though I have at most -/+ 3% variation at the mains I calculated 57VDC at main capacitor bank with 42VAC transformer so there is still -/+10% headroom. I expected the new transformers to sag even lower because they will be 500-600VA while the existing ones are 725VA.

Still I consider what you and Jeff proposed, you are right, safety first.

I haven't gotten anywhere with the tube preamp. We've been busy with a few other projects that have been taking a lot of time to complete.

No need to hurry. I see your Simpelstark is going very well. Ride on
 
I have tried NSOPS and on the positive note I found that its way too snappier in bass, mids and highs are great but isn`t bass way too quick or is that way it sounds may be its more like super snappier? or is there anything to be considered while using NSOPS or is it the way it sounds. Overall sound is excellent though.
 
Hi Sandy,

Just to clarify:

- Is quiescent current of each output device set to 65 mA? (with signal off)
- Is bias clamping spreader tuned-up the right way? (with signal on)
- What front-end is used in the test?

Soundwise, good VFA front-ends demonstrate richer, more "purple", and at the same time punchy bass, then CFA ones. Even better result - no global feedback setup with some highly linear front-end.

Cheers,
Valery
 
Hi Valery thank you very much for the prompt reply. I have setup at 20ma as having 3 pairs at 65V is considerable amount of heat when we have class A driver stage along with it. So I guess I have to move to 65ma to get decent output? I felt one thing that using a low bias is fine as anyway the switching is not happening. As you know most of the amplifiers have higher bias especially to deal with crossover distortion but when there is no switching at all any bias which doesnt hit to zero should be good isnt it?

Now consider about 2W for each rail of class A driver stage which itself dissipates about 4 W and for both rails 8W of continuous heat and with 65ma of bias for each output device at 62V will be about 4W x consider 6 of them which is 24W and in total 8W+24W = 32W of continuous heat and it ends up on a massive heatsink as well. I`m using emitter resistor of 0.33ohm for each transistor. I haven`t tried at 65ma bias per each transistor. Will update you with that number how it sounds.

But my doubt is that when there is no switching happening what is the point of having a higher bias?
 
Please set it up

OK... that's what I was thinking. The point is - it's not "automatically" non-switching.
You need to set it up to be non-switching. Also - there are certain "sweet spot" settings for the best performance, even when it's already running as "non-switching".

The other point - the non-switching mode of operation does not lead to saving on the heatsinks. Not at all. At 50W output power, each output transistor will dissipate 16W of heat (6 output devices will dissipate 96W on average). Regardless if it runs as non-switching, or as a conventional emitter follower.

There are 2 spreaders in this OPS - let's call them bias spreader and clamping spreader.

The setup procedure consists of 2 steps. Both spreaders are set to the minimum bias initially. I recommend multiturn pots for more accurate settings.

1) Bias at idle. No load, no signal. Set 65-70mA per output pair. Let the OPS warm-up during 15-20 minutes and correct the setting.

2) Bias clamping (non-switching mode). Connect the load (resistor, powerful enough), send 1 KHz to the input so that you have 5-10V RMS at the output (some true RMS DMM will be useful here). As soon as you have 1KHz at the output, connect your DMM (voltmeter) to emitter resistors (I normally measure 2 emitter resistors in series for better accuracy). Note the value it shows - it will be much higher than the idle value as you have the load and the signal in place.
Start rotating the clamping pot carefully from its minimum position, watching the voltmeter. During some time you'll see no change. At some point, you will see the value increasing. Let it increase for some 5mV from the initial value.
That's it - leave it there. The non-switching mechanism is activated and set for the best performance.

Once again, important - first you set the idle bias with no load, no signal - then you set the bias clamping level with the load and with the signal in place.

After that, if you listen once again - that's what it is.
But the heatsinks have to be good enough. +/-62V DC rails will allow you up to around 200W @ 8 ohm before clipping, maximum heat dissipation of around 100W per channel will happen at around 90W output power. Even at 50W output power, heat dissipation is already close to 100W.

By the way, this OPS performs a very nice clean clipping.

Cheers,
Valery
 
>Also - there are certain "sweet spot" settings for the best performance, even when it's already running as "non-switching".

-> so how to achieve that sweet spot

> 1) Bias at idle. No load, no signal. Set 65-70mA per output pair.

-> I thought 65ma per transistor, per pair can be relatively managed for dissipation. No load? so you want to disconnect everything at the load no speaker no large dummy load resistor for just bias setup?

>(I normally measure 2 emitter resistors in series for better accuracy)

-> so the voltage in mV between the emitter resistors of both npn and pnp in the OPS?
 
>Also - there are certain "sweet spot" settings for the best performance, even when it's already running as "non-switching".

-> so how to achieve that sweet spot

> 1) Bias at idle. No load, no signal. Set 65-70mA per output pair.

-> I thought 65ma per transistor, per pair can be relatively managed for dissipation.

*** This sentence is not quite clear to me. By "pair" I mean npn + pnp output pair. As those npn and pnp are arranged in series, current flowing through them is the same (65-70mA). So, current through the pair, or current through each of them are the same. Heat dissipation is also the same.

No load? so you want to disconnect everything at the load no speaker no large dummy load resistor for just bias setup?

*** Yes - disconnecting everything from the output will ensure there's no current flow through the load (which is possible in case you have some DC offset by some reason).

>(I normally measure 2 emitter resistors in series for better accuracy)

-> so the voltage in mV between the emitter resistors of both npn and pnp in the OPS?

*** Yes - in this case, you have better measurement accuracy. But don't forget to divide the measured voltage by a double resistor value to find out the actual current value.

Attached picture clearly shows the output pair's collector currents - no zero crossing = no closing = no switching.
 

Attachments

  • NS no switching illustration vz.png
    NS no switching illustration vz.png
    250.9 KB · Views: 209
Last edited:
OK... that's what I was thinking. The point is - it's not "automatically" non-switching.
You need to set it up to be non-switching. Also - there are certain "sweet spot" settings for the best performance, even when it's already running as "non-switching".

The other point - the non-switching mode of operation does not lead to saving on the heatsinks. Not at all. At 50W output power, each output transistor will dissipate 16W of heat (6 output devices will dissipate 96W on average). Regardless if it runs as non-switching, or as a conventional emitter follower.

There are 2 spreaders in this OPS - let's call them bias spreader and clamping spreader.

The setup procedure consists of 2 steps. Both spreaders are set to the minimum bias initially. I recommend multiturn pots for more accurate settings.

1) Bias at idle. No load, no signal. Set 65-70mA per output pair. Let the OPS warm-up during 15-20 minutes and correct the setting.

2) Bias clamping (non-switching mode). Connect the load (resistor, powerful enough), send 1 KHz to the input so that you have 5-10V RMS at the output (some true RMS DMM will be useful here). As soon as you have 1KHz at the output, connect your DMM (voltmeter) to emitter resistors (I normally measure 2 emitter resistors in series for better accuracy). Note the value it shows - it will be much higher than the idle value as you have the load and the signal in place.
Start rotating the clamping pot carefully from its minimum position, watching the voltmeter. During some time you'll see no change. At some point, you will see the value increasing. Let it increase for some 5mV from the initial value.
That's it - leave it there. The non-switching mechanism is activated and set for the best performance.

Once again, important - first you set the idle bias with no load, no signal - then you set the bias clamping level with the load and with the signal in place.

After that, if you listen once again - that's what it is.
But the heatsinks have to be good enough. +/-62V DC rails will allow you up to around 200W @ 8 ohm before clipping, maximum heat dissipation of around 100W per channel will happen at around 90W output power. Even at 50W output power, heat dissipation is already close to 100W.

By the way, this OPS performs a very nice clean clipping.

Cheers,
Valery
Good explaint!;)
 
*** This sentence is not quite clear to me. By "pair" I mean npn + pnp output pair. As those npn and pnp are arranged in series, current flowing through them is the same (65-70mA). So, current through the pair, or current through each of them are the same. Heat dissipation is also the same.



*** Yes - disconnecting everything from the output will ensure there's no current flow through the load (which is possible in case you have some DC offset by some reason).



*** Yes - in this case, you have better measurement accuracy. But don't forget to divide the measured voltage by a double resistor value to find out the actual current value.

Attached picture clearly shows the output pair's collector currents - no zero crossing = no closing = no switching.

Very well explained thank you Valery will do the way you have mentioned.
 
I finished a build of Valery's Sauberkeit amp with VFET output.
All working fine - on the lab bench. Tomorrow, if time allows, will do music listening tests.
Very stable amp, real build worked exactly like sim predicted, no surprises.
Idle current set to approx 40mA per pair of outputs.
Output DC offset: 0.5mV on both channels.
Tried 2 different op-amps: LT1056 and TL071; no visible differences on the scope; both work the same.
Input + VAS are on the smaller daughter-board, sitting on top of aluminum chassis with OS.
External PSU. After listening to actual music (hopefully tomorrow), will post final photos.

sauber.8nov22.png
 

Attachments

  • sauber.0.asc
    16 KB · Views: 87
  • DSCN0776.jpg
    DSCN0776.jpg
    100.9 KB · Views: 172
  • DSCN0772.jpg
    DSCN0772.jpg
    133.6 KB · Views: 146
  • DSCN0765.jpg
    DSCN0765.jpg
    133.8 KB · Views: 145
  • DSCN0764.jpg
    DSCN0764.jpg
    130.9 KB · Views: 144
  • DSCN0763.jpg
    DSCN0763.jpg
    132 KB · Views: 137
  • DSCN0761.jpg
    DSCN0761.jpg
    133.8 KB · Views: 122
  • DSCN0760.jpg
    DSCN0760.jpg
    132.8 KB · Views: 126
  • DSCN0757.jpg
    DSCN0757.jpg
    121.6 KB · Views: 160
  • Like
Reactions: 1 user
Large parts with radiators are on the leads (Fig.1,2). Not the best mounting method for transportation and long-term use :(

Do you mean 2 transistors TO-126 with small aluminum heatsink ?
The heatsink is very small/light...
But you are right - the whole idea of 'open' platform is not safe, especially for transportation.
That's why this whole rack, now with over 10 'open' amps is not going anywhere :)

20200731_213739.jpg
 
Last edited:
Music tests finished.

Great sound - but as usual, with my tin ears I can't distinguish this amp from other 10+ 'good' amps that I use.
I guess that's the good thing. If it sounded different - most likely something wrong with it :)
Next step - measurements with QA401, maybe over the weekend..

These TO-126 transistors on the input/vas doughterboard are not running hot (approx 250mW each), so I guess
the heatsink is not really needed on them, but I'll leave it since it's already done...


DSCN0784.jpg
 
Last edited:
  • Like
Reactions: 1 user
Tribute-3000 spectrums

More precise spectrums. Previous quick measurements were taken on the edge of overload at the input of the audio analyzer. Correct attenuation results in more accurate measurements. Output swing is 20V RMS @ 8 ohm load.

A very fine amplifier.
RIP Valery....
Any body made this amp?
What is the recommended power supply for this amplifier.
Thanks,
Sumesh
 
Revisiting some old good amplifires.
Still under construction,two mono.
I have modified the toroidals adding separate windings,6.3v for filaments 15-0-15 for VAS,12v for protection
I'm listening the TubeSumo+NS now with my Asathor speakers.
Great sound experiences!
 

Attachments

  • IMG_20231201_195159.jpg
    IMG_20231201_195159.jpg
    450 KB · Views: 49
  • IMG_20231201_195743.jpg
    IMG_20231201_195743.jpg
    380.9 KB · Views: 49
  • IMG_20231201_195809.jpg
    IMG_20231201_195809.jpg
    420 KB · Views: 49
Last edited:
  • Like
Reactions: 1 user