Zero Feedback Impedance Amplifiers

Frequency characteristic is in this case +/- 6 dB, but curve have big " holes " in near frequencys, which is not good for correct reproduction. But I will stop my talking here, 'cos enthusiasm of some individuals is too big and logical arguments havn't here target. Good luck, a few devices is realy heavy advantage. ;)
 
Hi darkmoebius,

I seldom need to use it, but I went over to the German abbreviation 'trafos' so that there could be no confusion between transistors and transformers, also, and without wanting to be disrespectful, I'm pretty sure that nothing electronic would come back if I entered 'trannis' into Google these days.


Hi Susan,

Glad I haven't scared you away.

Dave has kindly provided us with a graphical representation of your figures, and there appears to be a suckout about 400kHz with a possible peak about 200kHz.
Anything higher than the suckout could be dealt with using a Zobel in exactly the same way we are currently obliged to with feedback amplifiers.
A 200kHz turnover is goung to have phase impact circa 60kHz, though most commercially produced 20th century audio amps begin to phase change before 20kHz anyway.
Don't forget that your sinewave 50 ohm input/output loading is quite different to music driving with a reactive (non resistive) loudspeaker load.

I should have been more interested in 10Hz to 100kHz, also for the entire amplifier and not just for one transfomer in isolation. The overall phase response could additionally be determined from the time delay in zero voltage crossover between input and output on a dual beam scope.

Did you try a 1kHz or 10kHz squarewave. Digital photo ?

In that the input trafo is likely to be the weakest link I would suggest that there is merit in Hugh's earlier suggestion for a non-feedback thermionic input stage which will not introduce any audibly consequential passband phase or amplitude distortion.

Susan, the rest of us here are flummoxed, for we cannot adequately simulate your design on our computers. So don't feel that you are necessarily obliged to reveal the results for any tests that might affect possibile commerciality of your design. There is no point in drawing further derogatory comment or suggestion unless you are trying to prove your design.

Cheers ........ Graham.
 
Comments

Susan,

I just want to support Graham's comments.

I cannot speak for others, but I am for one grateful for your generosity of sharing your experience and your ideas with us. There will always be people in this world who think they are smarter than the rest. I don't think they have a right to criticise anyone else's work unless they have experimented with it and have solid proof to demonstrate their arguments. And even then, audio is a subjective thing, which is why it is such a wonderful spare time occupation for many of us. I am sure there are more people on the forum who like your work than those who don't.

I am not saying that I am going to stop my power amp project and build one of yours tomorrow. But I have a lot of respect for your lateral thinking, not accepting the nom and be prepared to design from first principles. I am sure the world is a richer place with such different ideas.

I just felt that I should air my support (not meaning that you would need any).


Patrick
 
  • Like
Reactions: 1 user
More Statistics

Hi Everyone,

Thank you for your comments - I am afraid that I am very behind in replying to the points raised.

Here is a similar test but using a MOSFET drive but still into a 50 ohm load. I am not sure what the speaker impedances will be at these frequencies but I suspect they will be higher in many places.

Graham asked about the phase response, which I have measured as best I can with by scope (Tek 465). It becomes somewhat moot above 1.0 MHz as the signal is becoming distorted but I included some extra figures to show the trend.

The point of going from 100 kHz is that the amplifier bandwidth is higher than this but my HP 8903B Audio Analyser stops at 100 kHz.

The Audio Analyser also requires a pre-amp to drive the imput transformer and stuff and as I have not done much audio work on the amplifiers for quite a few years so my memory is a bit fuzzy around the edges. I am having to build and set up stuff to be able to do these tests, and then do them, hence the time taken to answer questions. I will be able to give you some distortion figures soon.

It is important to see what happens above the normal audio bandwidth as surprises can happen (as another thread which is asking about GSM mobile interferance illustrates).

It is also my understanding that high frequencies can feed back as sub-harmonics into effects that can be heared - which is a problem sometimes notived in over sampling digital filters.

Anyone who has put stuff in for EMC/RFI CE or FCC compliance testing - so the product can be sold commercially - knows the horendious conducted and radiated emmissions tests that these poor circuits are flayed with.

In this test the IRFP140 mosfets start pegging out at 1 MHz, with the waveform becoming horribly distorted.

However above 4 MHz they stop trying and just act as capacitors and the output becomes a nice but low level sine wave. This illustrates the direct input to output path bit of my design.

In practise with the regular input transformer with its high impedance drive to the mosfet gates this won't happen as it is a LC filter. It is important to know the characteristics of each stage however and that the output has higher bandwidth then the imput is in my book "a good thing".

I am doing these tests at 50 ohms because this is what I can measure with the equipment I have and although not the same as audio you should be able to build it in Spice and see results...

... except I have been singularly unsucessful in that endevour.

N.B. In the current issue of Electronics World (Nov 2004) there is a artical about the problems of modling mosfets.

Simulation works up to a point, and usually that is quite close. However transformers and mosfets seem to be the exception. I am also not good at maths (being dyslexic) which doesn't help.

Others who are suitably skilled in the art will hopefully have better sucess.

And I may be wrong anyway.

====

Tx + IRFP140 drive - In tx 1:2 iron powder ring core (gives double the volts output drive).

kHz Volts Phase
100 1.476 10
120 1.495 12
150 1.5 15
200 1.508 20
250 1.474 22
300 1.443 25
350 1.4 27
400 1.358 30
450 1.317 32
500 1.278 35
550 1.238 38
600 1.198 40
650 1.157 43
700 1.117 45
750 1.077 45
800 1.048 50
850 1.008 50
900 0.9631 50
950 0.9226 55
1000 0.8818 60
1050 0.8413 62
1100 0.801 65
1150 0.759 70
1200 0.7145 75
1250 0.666 80
1300 0.6112 80
1350 0.5501
1400 0.4912
1450 0.4468
1500 0.4195
1550 0.4028
1600 0.3912
1650 0.3796
1700 0.3665
1750 0.3564
1800 0.3453
1850 0.3403
1900 0.339
1950 0.3368
2000 0.3334
3000 0.022
4000 0.2206
5000 0.274
6000 0.2142
7000 0.1631
8000 0.0786
9000 0.02044
10000 0.0336
12000 0.1067
15000 0.04265
20000 0.0487
25000 0.04855
30000 0.00668
35000 0.0139
40000 0.0263
45000 0.0139
50000 0.02237
- up to - Similar levels - variations
1000000 0.01107

URV5 background 0.002 V
OP harmonic distorted 1.0 to 3.5 Mhz
OP sine 4 Mhz and above

Source HP 8657A
Measure Rohde & Schwarz URV5 + UTY-Z4 head
Load Becibel DB4303B 50 ohm

+/- 3 dB puts the bandwidth at 1 MHz, which is the mosfet limit.
- 3dB is circa 600 kHz

N.B. Some of this variation is probably in the input transformer.

The output of the output transformer is connected in exactly the same manner as in the previous test. However there is no kick up or suck out which will be due to the lower impedance drive by the mosfets to the transformer.

----

Best wishes,
Susan.
 
Fame or Infamy?

Hi,

Bas Horneman said:


Susan Parker = Audiophonics..

Well Susan...that means that you are famous!!! ;)

Cheers,
Bas

Famous, or infamous?

Yes, I did the AudioPhonics designs.

The speaker in the Ad used a 52 mm cone driver from Denese? Jordan but they proved unreliable, so I redesigned for the 62 mm cone driver from Mr Jordan. However these don't seem to be available anymore so I am having to look at alternatives.

I set up with a now long since ex partner as there seem to be an opportunity in the higher end of the Audio market for a small business. However I underestimated the amount of capital that even such a small and specialist manufacturer would require.

In the end if one doesn't have the funds to pay heavily for advertising one is very unlikely to be successful. Marketing glitz is far more important than technical specifications.

Graham's comment in his Electronics World article prompted me to approach him on the DIYaudio forum, and he suggested I started this thread. I was uncertain about the reception I might receive but decided - for better or for worse - to stick my head above the parapet.

I added the other information on the speakers etc. as it gives a better explanation as to where I was coming form with this amplifier.

I have done my best to explain my reasonings as I have endeavored to base all my design work on technical merit rather than techno babble (which I deplore). For example my speaker's form is entirely dictated by function, NOT style.

If I am wrong or limited in understanding in any area then I have at least given you the base line information so you may see where I have been mistaken.

Much of Audio is personal preference and I am aware that hearing is a very personal thing. My own sensitivity to phase may well make me less aware of other distortions that others would not be happy with.

I am pleased and I would admit somewhat flattered that some of you have expressed an interest in building to my design. I would suggest in the first instance that if possible this is done by one individual (who has some test equipment?) who can then confirm or deny my design's merits.

Thank you all once more for your interest and constructive inquiries and comments.

BW,
Susan.
 
I set up with a now long since ex partner as there seem to be an opportunity in the higher end of the Audio market for a small business. However I underestimated the amount of capital that even such a small and specialist manufacturer would require.

Hi Susan,

These are very different times from 1993. And a half decent web presence is a very good marketing weapon. That said...the Audiophonics speakers are something that seems to me, relatively hard/expensive to manufacture.

Regards,
Bas
 
Re: More Statistics

Susan-Parker said:
I am doing these tests at 50 ohms because this is what I can measure with the equipment I have and although not the same as audio you should be able to build it in Spice and see results...

... except I have been singularly unsucessful in that endevour.

If some Spice guru could offer us the transformer models
based on the data Susan provided, we could start playing in the virtual world.
This would also be a nice return to Susan, assuming that her endeavor was unsuccessful in this area.

/Hugo
 
Hi Susan,

If your AF genny has 600 ohm output, surely you could feed the 600 ohms input transformer directly, or you could use approx 470 ohms *live* and 220 ohms *ground* in series, and feed your 150 ohm input impedance amplifier from the 220 ohm resistor.

I feel that response above your 400kHz transformer suckout frequency is not greatly relevent in a non-feedback design; the core is little involved in mutual coupling at such high frequencies.

Cheers .......... Graham.
 
Hi Susan,

Basically good.

The 2MHz resonance is well damped, but even better might be possible by empirically trying different damping resistors at different points, also a Zobel.

Clearly the amp is affecting your signal generator output, though additional trafo damping might simultaneously improve this too.

Keep up the good work !

Cheers ........... Graham.
 
Re: Vbias circuit

Hi Tyimo,

Tyimo said:
Hello Susan!
Congratulate for your new and "mind refreshment" design!
At the same time I would like to please you to post here a whole, detailed schematic with the bias circuit and PSU, if it is possible. I am affraid that I didn't understand the "Vbias" 100%. I am an amateur...

Thanks and best regards:

Tyimo

Thank you.

Sorry for the delay in reply to your question.

Add a standard power supply (transformer, bridge rectifier, smoothing cap) noting that it should not be greater than 1/2 the voltage rating of the mosfets, and to allow for mains suply variations a little less. E.g. for IRF150 or IRF140 use circa 45 volts max.

The power supply and the V Bias circuit I used is on:

http://www.susan-parker.co.uk/zeus-schematics-1.htm

I will add more to this page in time.

Many thanks for your interest in my design.

Best wishes,
Susan.
 
Hi Graham, Mikek,

Graham Maynard said:
Hi Susan,

Basically good.

The 2MHz resonance is well damped, but even better might be possible by empirically trying different damping resistors at different points, also a Zobel.

Clearly the amp is affecting your signal generator output, though additional trafo damping might simultaneously improve this too.

Keep up the good work !

Cheers ........... Graham.

Thank you, that's pleasing.

I haven't been doing any audio stuff for a long time and I am still rummaging around trying to find what bits and documentation I still have.

The amp input is being buffered by a op-amp ballenced line driver board that I had handy. This works but needs something a bit better as it really doesn't like the load.

I really need to clean up this line driving as until I can feed in a clean square wave trying to adjust the output damping isn't going to be optimum.

I have some ideas about this but it will probably take a little while to try out.

Thanks.

Best wishes,
Susan.
 
Hi Dave,

Originally posted by planet10

Numbers converted to a picture

dave

Many thanks for the graph.

==========

Hi Everyone,

Here are some distortion figures for the 35 Watt transformer amp with IRFP140s at 1 kHz.

Please note that the distorsion is measured on a HP 8903B communications test equipment and is % distorsion, not THD + noise (I believe there is a small difference).


Supply 34.0
Vbias 3.60
Q1 bias 0.225
Q2 bias 0.233
Load 8.02

Watts RMS Dist%
40.4 18.00 0.465
36.1 17.02 0.479
32.2 16.07 0.425
28.1 15.01 0.331
24.6 14.06 0.304
21.3 13.07 0.335
18.2 12.07 0.374
15.1 11.00 0.420
12.7 10.08 0.463
10.2 9.04 0.518
8.10 8.06 0.574
6.12 7.01 0.636
4.52 6.02 0.688
3.19 5.06 0.714
2.05 4.06 0.655
1.19 3.09 0.445
0.79 2.51 0.268
0.50 2.00 0.149
0.39 1.76 0.113
0.28 1.51 0.076
0.19 1.25 0.054
0.13 1.01 0.042
0.10 0.90 0.038
0.080 0.80 0.030
0.061 0.70 0.031
0.045 0.60 0.023
0.031 0.50 0.028
0.020 0.40 0.039
0.011 0.30 0.037
0.008 0.25 0.027
0.005 0.20 0.013
0.003 0.15 0.039
0.001 0.10 0.092

===========

Best wishes,
Susan.
 
Hi,

mikeks said:
At what frequency were these readings obtained?

Did you use a bandwidth-limiting filter at the test point?

Does THD (at 1KHz and say 10KHz), change significantly when the nominal load is reduced to 4R, and again to 2R?


Test at 1 kHz.

No filter (test bandwidth is 500 kHz).

I haven't tried the other loads yet, and wouldn't do the lower without changing the windings to match the load - but the test transformer isn't wound with split secondaries.

Thus with 4 ohms the actual load seen would be higher.

I have to do the measuments manually, so it all takes time to do. Very frustrating as much of my setup actually has GP-IB ports, but I don't have the computer interface and software.

Something for tomorrow :)

Best wishes,
Susan.