You should really turn it off! See next.
Howland.
Who was the designer of Zeck amps? certainly not Richard Beck.
Calcs for op amp loading vs THD?
- Thread starter atavacron
- Start date
Help Support GroupDIY Audio Forum:
thor.zmt
Well-known member
Indeed.
No idea. I'd argue (again) that the circuit is "obvious", just a Class AB Hitachi App Note circuit with "current dumpers" after the Mosfet's.
Zeck (like Behringer) was a German Brand that offered budget oriented products. Some likely outright copies, others best described as "inspired by" something.
Their highly popular 15/3 Speaker was basically inspired by the EV S15-3 Speaker, using the same box and woofer, but much less expensive midrange (essentially Audax PR170M0 instead of EV VMR) and tweeter (Beyma CP-16 instead of EV-ST-350).
In turn, in then East Germany I made very excellent copies of the Zeck 15/3 as side business, they were true copies, including Zeck badge, made maybe 100pcs. They used original drivers imported illegally via interesting channels, everything else was made local and thus much less expensive. I sold them at 60% of the originals and made a tidy profit.
The one shown above is original, mine actually used the "upper model" lineup with Beyma CP-21 Tweeter and EVM-15L Pro Line woofers, for a "rated" 400W Speaker, instead of the only "200W" of the standard model. We also has Subwoofers in the same format with EVM-1*B PL
Later on Zeck started producing some genuinely interesting unique stuff, but failed to retain their commercial success and eventually went under.
Thor
Last edited:
unfamiliar jargon, thanks for explaining , now I forget the context.
yes it is...
yes it is
OTAs are current output devices. The input LTP (long tail pair AKA input differential pair) converts a small input voltage to current, the OTA bias current scales or multiplies that output current proportionately. This relationship has been exploited to make current controlled variable gain stages.
It is not that trivial... The junior semiconductor engineer I was working with also thought it was trivial
. Since OTAs are only linear over a low mV input voltage range, so the input LTP needs to be very low noise. Further to prevent control feed through the input LTP must be very well matched. At Peavey we used a house number OTA that was selected for guaranteed low control feedthrough (important when used in a power amp input stage). When I tried to explain to the wet behind the ears junior semiconductor engineer that the OTA input LPT needed to dedicate enough silicon area to be low noise, "and" insure matching to prevent control feedthrough, I suggested perhaps paralleling multiple devices but he said not neccesary, he'd just use degeneration resistors to reduce the differential's sensitivity to DC matching. Clearly he did not understand how OTAs work, resistors in series with the input differential emitters would interfere with proper OTA operation. I decided it was not my job to teach the junior semiconductor engineer how to design ICs.
Now you want to instruct me about how DDT works? Since Peavey has incorporated DDT into almost every amplifier they made, there are millions of these out in the wild. Probably several different implementations but the typical one that comes to mind uses an input resistor feeding a high impedance non-inverting input. The OTA is configured to suck an opposite polarity current from that resistor at the + input node. Unlike a VCA this OTA is completely out of the circuit until active and reducing gain, so clean and quiet.
DDT compares the amplifier input to output and detects when the output is not agreeing with the input. DDT is triggered to reduce gain during output stage saturation (clipping), during current limiting, or any deviation from correct linear output. The OTA is driven with a control current using a fast attack, fast release side chain.
No... While I wasn't around when the very first CS800 was released back several decades ago, the ones I was familiar with (since the mid 80s) were all class AB, and used full complementary (motorola) power output transistors. Peavey established the class A current with special diodes in a bias string that were engineered to track the temperature coefficient of the output devices.These diodes were thermally coupled to the heat sinks.
Perhaps you encountered some counterfeit CS800s? I've heard of and seen pictures from China, and South America (I remember one that spelled Peavey wrong). The CS800 was iconic.
This stopped being interesting (to me) a while ago...
JR
thor.zmt
Well-known member
Correct. That is precisely what I wrote colloqially.
The Bias control is not an essential feature of an OTA, equally if a bias control is applied to a classic voltage output Op-Amp (O (no T) A) we also have a VCA.
This is not an inherent function of "OTA" but strictly of some specific OTA's.
It is not that trivial... The junior semiconductor engineer I was working with also thought it was trivial
Yes. TBH, pressing the CA3094 into VCA service seems suboptimal.
I can think of a range of possible limiter options (including real VCA's) that have better performance.
You used Intersil CA3094. Selected, who knows. Normally single chip LTP's are well matched and noise is just a question of chip area. I'm not really a Chip Designer and University is ages ago, so I am not that up to speed.
When I tried to explain to the wet behind the ears junior semiconductor engineer that the OTA input LPT needed to dedicate enough silicon area to be low noise, "and" insure matching to prevent control feedthrough, I suggested perhaps paralleling multiple devices but he said not neccesary, he'd just use degeneration resistors to reduce the differential's sensitivity to DC matching. Clearly he did not understand how OTAs work, resistors in series with the input differential emitters would interfere with proper OTA operation. I decided it was not my job to teach the junior semiconductor engineer how to design ICs.
I decided that when I had a department of "singing pigs" (* see below) I sadly had to tech them how to sing at least a single note, to give me the result I required.
Yes, I had a look before I wrote what I wrote. Quite obvious, if you what hides behind the relabelled IC.
A VCA can do the same. Set the VCA to "maximum attenuation" as default, take the voltage output to be opposite polarity and sum it into the signal. If inactive we are pretty much looking at a static resistor.
So the OTA is still a form of VCA. Input signal goes via an attenuator. From a buffer amp. There is a series resistor to mix OTA current output and the input voltage. One could also use a Voltage output VCA and a resistor to do the mixing. Same result.
It is inverted and variable with the bias input. No-Clipping it is set to maximum attenuation, which incidentally is NOT infinity, so the OTA/VCA is ALWAYS IN CIRUIT.
With clipping the gain is turned up and the output current rises and creates a voltage across the input resistor that reduced the amplifier input.
There are many other gain control elements that can be used in a functionally equivalent circuit (and may perform better - how much this matters with the amp near clipping I think is academic).
Not possible. There are (CS-1200X) six diodes and no additional resistors. And six BJT Junctions and emitter resistors. The result is pretty much zero bias in the output transistors. That is how electronics work. While all transistors biased by this chain are on the heatsink, there is still a temperature gradient.
Running at essentially zero Bias is class B, not AB.
Yes. And the ones I had were quasi complementary.
And essentially Class B. The schematic states 0V across the Output Transistor Emitter Resistors and 0.5V base voltage
QED.
Later versions may differ.
Same here. However you insist making false claims, that I do not feel I can leave standing as you make them.
For stuff that happend in the 70's and 80's I usually append "IIRC - If I Remember Correctly" as my memory ain't what it used to be.
Thor
again you are losing me VCA= voltage controlled amplifier.... Did I mention that one of my patents was for a novel VCA?
US04818951 Roberts 04/04/1989 Gain control or multiplier circuits.
?
it was not used as a VCA but as a peak limiter... Since it was only active while limiting very loud signals the fidelity was more than adequate.
Back in the 80s before I worked at Peavey I designed a quad noise gate limiter using OTAs in the feedback path of TL07x op amps. The path performance while not limiting or gating was like a unity gain op amp... very clean. No VCAs back then or now are that clean.
I do
exactly... My one class in semiconductor physics was in the early 70s. Indeed LTP noise is a function of device Rbb and the obvious way to lower Rbb is to allocate more silicon area (like I said).
?
that old schematic is a decade before I was hired by Peavey. Indeed it was designed by Jack Sondermeyer RIP, I recognize his signature.
You are correct that amp is quasi complementary, early PNP power devices were not very robust.
BUT that old CS800 is clearly class AB, NOT class B... Why in the world would the bias string diodes be thermally coupled to the heatsink if the power devices were being completely cut off? As I recall the diodes in that bias string were not typical rectifier diodes but some zeners operated backwards in diode mode, to match the tempco of the power devices. This stuff was all worked out years before I worked there so I don't know details.
Peavey was very opposed to using factory production trims in product designs. Most contemporary power amps use Vbe multipliers for setting class A bias trim. Production line trims a) require skilled production workers, b) involve extra cost and time, c) are an opportunity for mistakes and field failures.
Jack put a lot of effort into his "trimless" class AB biasing scheme. In my experience they worked adequately and were suitably reliable.
Not my favorite approach*** but I used them in a handful of amps I did while inside Peavey (15 years).
While I was product manager for all power amps I oversaw the design of a CS800S with a switching power supply... there were multiple versions of that iconic amplifier.
As the saying goes if you claim to remember the 70s you weren't really there.
JR
*** as I already shared the trimless class AB bias could be a little dicey when used with plastic darlington devices. Jack's design depended on the bias string diodes being suitable for the metal TO3 power devices in use. The plastic darlington amp was a low power guitar amp and there were field complaints from music store salesmen about how ratty it could sound on cold mornings before it was warmed up. I designed a bias scheme that sensed across the emitter resistors but I couldn't get Jack interested.
thor.zmt
Well-known member
Ok, strictly speaking ICA (Current Controlled Amplifier), if we want to be strict.
A VCA is just another building block.
You can put in the forward part of the signal path.
You can place it into the feedback path of of the signal path.
You can control it with a control current by adding a resistor to perform current to voltage conversion and so on.
If used the same way, you might find them cleaner and capable of operating at higher levels.
Personally I used VACTROLS, they are also "out of circuit when not used" and the build in timing is almost ideal for limiting, so just connect the LED so it lights when the Amp is clipping.
Many parts less than DDT and same result, give or take.
"Never attempt to teach a pig to sing; it wastes your time and annoys the pig."
Yet if all you have are "non-singing singing pig's" (meaning a bunch of ejits who cannot do their job's properly) and do not want to give up, you need to find a way to get the pig to at least sing one note (do one part of the job), then all of them will make a melody.
As the schematic states 0V across the Emitter resistors, the quiescent current through the output transistors must be also 0mA.
And 0mA in the Output transistor = Class B.
The output transistors are biased just shy of starting to conduct (0.5V BE).
As this is very close to conduction threshold bias to the drivers and outputs must track heatsink temperature, or the output stage will start conducting and crate a race condition that eventually triggers SOAR protection or destroys the output transistors.
Yes. And all the CS800/1200 Amp's I worked on had ~0mA output quiescent current, where (based on 0.33R) 78mA quiescent current per device would be needed to be "optimum bias Class AB".
Mind you, maybe what you encountered may have been generations later than what we used in the Eastern Europe behind the iron curtain.
Then why not use suitable bias circuit (there are much simpler alternatives to Class I) that creates sufficient quiescent current for low distortion operation using sensing from the emitter resistors? It's not very hard.
In Eastern Europe sadly we had no option but to remember. At least some of it.
Thor
Last edited:
I don't like repeating myself so I'll stop repeating stuff...
JR
JR
thor.zmt
Well-known member
If you have a screen between primary (L/N) and secondary (which supplies the "driven rail" output stage) that terminates into the power ground (s), what reaches the L/N lines will be what can pass that screen.
The majority of the current is (must be, according to Kirchhoff's law) returned into the loop from which it "escaped" without passing via L/N and PEN.
Even connecting the screen to PEN in the device and PEN to Ground will do the same thing (shorten the current loop to remain "on board").
Poor analogy.
It is like saying that an SMPS with a correctly wound shield in the transformer that is connected to the right ground will have very low conducted emissions even with minimal other EMI mitigation, because the "Emissions" are simply not emitted but returned to the source in a shortened loop.
It is not the shielded chassis, it is controlling the currents in both main and parasitic loops to prevent "escape". Any current that "escapes" has to return to the source somehow. If that return involves a long loop, then we have a problem.
Thor
First, not all power xfmrs have an electrostatic shield, because it reduces efficiency.
Second, even if there's a screen, it's not 100% perfect.
The majority, yes, but there is still some leakage that is significant.
A specific isolation xfmr has 120pF leakage capacitance. It takes special attention to reach this level.
The xfmrs that are usually used in power amps are not isolation xfmrs, because the latter have lower efficiency, and efficiency (power/weight) is a very important parameter in a PA amp.
Did you not read correctly what I wrote? I said exactly the contrary, that completely shielding an smps is not enough to stop conducted emissions.
It is exactly what conducted emissions are. If they didn't exist, no one would care about them, up to the point they are a major concern in certifications.
Remember the MT/MA line of amplifiers was designed much before anyone cared about EMI/RFI, CE certification and emissions.
On a more generic POV, you may be a brilliant designer, but it does not give you the right to be so condescending. Some of us are not complete idiots.
thor.zmt
Well-known member
It reduces WINDING efficiency, but not a lot TBH, not electric efficiency. I designed and tested enough transformers with a screen. Including SMPS.
No, but 30dB is not hard.
If the leakage is significant will depend on screen efficiency and ultimately system impedances.
I routinely get less coupling capacitance with fully code and em regulation compliant SMPS.
Commonly Main and PEN are expected to be < 1Ohm with modest inductance.
In this case the 120pF & 1Ohm yield a -3dB point of 1.3GHz and 70V RMS at 5kHz will see 108dB attenuation to 280uV.
Honestly, the case you cited, while possible and plausible suggests a PEN way outside electrical code spec and transformers with very poor winding arrangements, way past what is normal with mains transformers, including Toroidal. Seems it would need bifilar wound primaries and secondaries to get such levels of coupling.
I never had problems of noise from driven rail amps, but many from dimmer packs and other lighting electronics.
Did you read what I wrote? I wrote that adding the correct shield in the SMPS transformer connected to the correct circuit node stops (well, dramatically reduces) conducted emissions?
It does so incidentally without the need for Y capacitors or overall shielding.
The concept is to mitigate issues at the source and return parasitic "escaping" currents to their source locally. Basically a practical application of six sigma.
I put several commercial SMPS I designed though certification, thank you very much. And yes, they were of the type I described, no or minimal y-cap, no shielded case, no earth.
The problem is that most power supplies (switching or not) are designed by "engineers" who have not understood Kirchhoff. And that is EE101.
I'm not a brilliant designer. But I can read. And I understand Kirchhoff's law and other basic electronics that many today seem to consider "optional".
And I'm a stickler for facts. And I do not consider someone's past background and achievements sufficient to let them play fast and loose with facts.
And I'm German. So I am direct and to the point.
I am NOT condescending.
Thor
Last edited:
In the particular case I mentioned, the inductance of the PE was quite high. In mobile PA set-ups, most cables are wound on cable reels, which increases considerably the inductance when the cable is not fully uncoiled.
Everything you just wrote shows you are.
Last edited:
thor.zmt
Well-known member
thor.zmt
Well-known member
Wearing my "Heath & Safety" Hat - I have doubt that setup would have passed electrical safety requirements.
Anyway, we both agree that using "Driven Rail" has drawbacks and benefits and they are different to those of both more conventional linear amplifiers and of switching amplifiers.
We disagree how easily the problem can be mitigated to become immaterial (you -> impossible, me -> trivial).
Sadly it is simply factual, based on what I encounter.
Thor
If you need to start your bbq without a lighter or matches this is the way to go !
It's just how it is in mobile PA applications. One has to mitigate the risks between overheating the cable drum and that of someone tripping on a loop of cable.
Did I say it was impossible? I just said there is a problem, and it must be solved. For solving a problem, someone must be aware of it. The Crown engineers, who were no slouch, solved the problem when they knew about it. One can't blame them for not anticipating this issue.
I routinely compliment Abbey for being the nice moderator, I will add that he is also very patient.
JR
