EMRFD Message Archive 6977

Message Date From Subject
6977 2011-12-08 19:07:50 jasonb1963 Class D RF Amplifiers
I'm looking at various amplifier options for the portable software defined transceiver I'm in the process of designing. Originally I was looking at Class C amplifiers, but I was not satisfied with the 60% efficiency I was seeing (at best). I have a love/hate relationship with Class E amplifiers, having achieved some impressive efficiency numbers up through 20M using paralleled BS170s (IRF510s don't work well for me on 20M and above). However, as soon as I change the load angle by putting a low pass filter on the output, the efficiency drops into the Class C range and no amount of re-tuning will bring it back. I've experimented with some carefully designed low pass filters (5th order Chebyshev) that have exactly 360 degrees of phase delay at the frequency of interest, and this worked, but it exacerbated the already very limited tuning range of the amplifiers (though it's fine for single frequency operation)

So I was looking at a high efficiency alternative to the Class E design, and started looking into Class D and Class F designs. After a couple of false starts, I did manage to get a Class D amplifier working with around 80% efficiency on 15M with a pair of BS170s (2N7000) in a push-pull configuration. This is slightly better than the best Class E configuration I have been able to build on the same band using a pair of BS170s in parallel. Output powers were around +33dBm with the Class D amplifier and around +32dBm with the Class E amplifier.

Two things I found very pleasing about the Class D configuration are that it is much more broadband than the Class E designs, and there was very little effect on efficiency when I added a low pass filter to it. Also, since the amplifier is a push-pull configuration, the even order harmonics are suppressed and I found a 3rd order low pass filter was sufficient to reduce the 3rd harmonic to more than 40dB below the carrier.

My present experimental design uses two BS170s in a push-pull configuration. I am driving the inputs with logic gates, so it is a simple matter for me to run the carrier through an inverter to achieve a phase reversal, eliminating a pesky transformer. On the output side, however, I am stuck with a transformer. There is a bifilar winding which serves to connect the power supply to each drain of the two BS170s. It is important that the phases of this pair of windings are reversed relative to each other or the amplifier doesn't work. The secondary winding is wound at a 4:3 ratio and provides the impedance transformation I need to get +33dB into 50 ohms. I have not tried to squeeze more out of it, but the BS170s are cool to the touch so I suspect I can bump the power up to at least 3 watts or more (as has already been done by others on lower bands).

I'm curious to know if anyone else has played with this type of amplifier? Their insensitivity to load angle, broadband output capability, and ability to work at considerably higher frequencies than could be achieved with the same transistor in a Class E configuration seem to make them a superior choice for CW, FM, and some digital modes (RTTY/FSK/OQPSK for example) in the upper HF band and most likely into at least the lower VHF bands as well.

Jason
6978 2011-12-08 22:10:08 ehydra Re: Class D RF Amplifiers
Yes, works. Nothing new.
IXYS has papers on it.

- Henry


jasonb1963 schrieb:
> I'm looking at various amplifier options for the portable software defined transceiver I'm in the process of designing. Originally I was looking at Class C amplifiers, but I was not satisfied with the 60% efficiency I was seeing (at best). I have a love/hate relationship with Class E amplifiers, having achieved some impressive efficiency numbers up through 20M using paralleled BS170s (IRF510s don't work well for me on 20M and above). However, as soon as I change the load angle by putting a low pass filter on the output, the efficiency drops into the Class C range and no amount of re-tuning will bring it back. I've experimented with some carefully designed low pass filters (5th order Chebyshev) that have exactly 360 degrees of phase delay at the frequency of interest, and this worked, but it exacerbated the already very limited tuning range of the amplifiers (though it's fine for single frequency operation)
>
> So I was looking at a high efficiency alternative to the Class E design, and started looking into Class D and Class F designs. After a couple of false starts, I did manage to get a Class D amplifier working with around 80% efficiency on 15M with a pair of BS170s (2N7000) in a push-pull configuration. This is slightly better than the best Class E configuration I have been able to build on the same band using a pair of BS170s in parallel. Output powers were around +33dBm with the Class D amplifier and around +32dBm with the Class E amplifier.
>
> Two things I found very pleasing about the Class D configuration are that it is much more broadband than the Class E designs, and there was very little effect on efficiency when I added a low pass filter to it. Also, since the amplifier is a push-pull configuration, the even order harmonics are suppressed and I found a 3rd order low pass filter was sufficient to reduce the 3rd harmonic to more than 40dB below the carrier.
>
> My present experimental design uses two BS170s in a push-pull configuration. I am driving the inputs with logic gates, so it is a simple matter for me to run the carrier through an inverter to achieve a phase reversal, eliminating a pesky transformer. On the output side, however, I am stuck with a transformer. There is a bifilar winding which serves to connect the power supply to each drain of the two BS170s. It is important that the phases of this pair of windings are reversed relative to each other or the amplifier doesn't work. The secondary winding is wound at a 4:3 ratio and provides the impedance transformation I need to get +33dB into 50 ohms. I have not tried to squeeze more out of it, but the BS170s are cool to the touch so I suspect I can bump the power up to at least 3 watts or more (as has already been done by others on lower bands).
>
> I'm curious to know if anyone else has played with this type of amplifier? Their insensitivity to load angle, broadband output capability, and ability to work at considerably higher frequencies than could be achieved with the same transistor in a Class E configuration seem to make them a superior choice for CW, FM, and some digital modes (RTTY/FSK/OQPSK for example) in the upper HF band and most likely into at least the lower VHF bands as well.
>
> Jason

--
ehydra.dyndns.info
6979 2011-12-08 22:23:29 Tamás Fábián Re: Class D RF Amplifiers
Hello!

On 9 December 2011 04:07, jasonb1963 <jasonb1963@yahoo.com> wrote:

> **
>
>
> I'm looking at various amplifier options for the portable software defined
> transceiver I'm in the process of designing. Originally I was looking at
> Class C amplifiers, but I was not satisfied with the 60% efficiency I was
> seeing (at best). I have a love/hate relationship with Class E amplifiers,
> having achieved some impressive efficiency numbers up through 20M using
> paralleled BS170s (IRF510s don't work well for me on 20M and above).
> However, as soon as I change the load angle by putting a low pass filter on
> the output, the efficiency drops into the Class C range and no amount of
> re-tuning will bring it back. I've experimented with some carefully
> designed low pass filters (5th order Chebyshev) that have exactly 360
> degrees of phase delay at the frequency of interest, and this worked, but
> it exacerbated the already very limited tuning range of the amplifiers
> (though it's fine for single frequency operation)
>
Did you try diplexers, or just "simple" lowpass filters? EMRFD discusses
class E amps, and recommends using a diplexer instead of a filter, so the
class E network will see an 50 ohm load on a broad range of frequencies.


>
> So I was looking at a high efficiency alternative to the Class E design,
> and started looking into Class D and Class F designs. After a couple of
> false starts, I did manage to get a Class D amplifier working with around
> 80% efficiency on 15M with a pair of BS170s (2N7000) in a push-pull
> configuration. This is slightly better than the best Class E configuration
> I have been able to build on the same band using a pair of BS170s in
> parallel. Output powers were around +33dBm with the Class D amplifier and
> around +32dBm with the Class E amplifier.
>
> Two things I found very pleasing about the Class D configuration are that
> it is much more broadband than the Class E designs, and there was very
> little effect on efficiency when I added a low pass filter to it. Also,
> since the amplifier is a push-pull configuration, the even order harmonics
> are suppressed and I found a 3rd order low pass filter was sufficient to
> reduce the 3rd harmonic to more than 40dB below the carrier.
>
> My present experimental design uses two BS170s in a push-pull
> configuration. I am driving the inputs with logic gates, so it is a simple
> matter for me to run the carrier through an inverter to achieve a phase
> reversal, eliminating a pesky transformer. On the output side, however, I
> am stuck with a transformer. There is a bifilar winding which serves to
> connect the power supply to each drain of the two BS170s. It is important
> that the phases of this pair of windings are reversed relative to each
> other or the amplifier doesn't work. The secondary winding is wound at a
> 4:3 ratio and provides the impedance transformation I need to get +33dB
> into 50 ohms. I have not tried to squeeze more out of it, but the BS170s
> are cool to the touch so I suspect I can bump the power up to at least 3
> watts or more (as has already been done by others on lower bands).
>
> I'm curious to know if anyone else has played with this type of amplifier?
> Their insensitivity to load angle, broadband output capability, and ability
> to work at considerably higher frequencies than could be achieved with the
> same transistor in a Class E configuration seem to make them a superior
> choice for CW, FM, and some digital modes (RTTY/FSK/OQPSK for example) in
> the upper HF band and most likely into at least the lower VHF bands as well.
>
> Jason
>
Do you drive the transistors at a much greater frequency than 14 MHz? This
is a usual technique at AF, but I suspect you're driving your PA at the
output frequency. Am I right?

I'm experimenting with a simple transmitter and I need an efficient PA (I'd
like to SOTA a bit, but hate carrying large batteries ;) ), so good luck
with the project! :)


[Non-text portions of this message have been removed]
6981 2011-12-09 06:47:11 nm0s_qrp Re: Class D RF Amplifiers
6982 2011-12-09 07:00:15 ehydra Re: Class D RF Amplifiers
Tamás Fábián schrieb:
> I'm experimenting with a simple transmitter and I need an efficient PA (I'd
> like to SOTA a bit, but hate carrying large batteries ;) ), so good luck
> with the project! :)

You can use a 74AC gate for class-D. 30MHz is no problem. Paralleling
74AC is problematic. All of the family are buffered and like to
oscillate therefor. But there are newer logic families with unbuffered
typed. Those can be paralleled.

A single 74AC gate have Rout about 13 ohms.

- Henry

--
ehydra.dyndns.info
6983 2011-12-09 08:47:44 AD7ZU Re: Class D RF Amplifiers
there is a great NA5N class D / E tutorial which I had bookmarked here;
part 1:  http://www.aoc.nrao.edu/~pharden/hobby/_ClassDEF1.pdf
part 2:  http://www.aoc.nrao.edu/~pharden/hobby/_ClassDEF2.pdf
 
Randy
AD7ZU
 
 
 


________________________________
6984 2011-12-09 09:16:58 jasonb1963 Re: Class D RF Amplifiers
Hi Fábián,

Yes, I tried diplexers as well. They did not work for me.

I don't think the problem is related to the amplifier seeing a 50 ohm load. I believe the Class E amplifier is very sensitive to the load angle (it needs to be around 30 degrees in order to work right). Whether you use a regular low pass filter or a diplexer, you will generally alter this load angle to something else (unless you very carefully design the filter to produce a 0 degree phase delay at the frequency of interest).

There is some discussion of this in the literature
6986 2011-12-09 10:00:30 jasonb1963 Re: Class D RF Amplifiers
Hi Dave,

I have a spreadsheet made by WA0ITP which is based on an article you wrote for QRP Quarterly in the Summer 2009 issue. Unfortunately, I do not have access to that issue. However, I did examine the topology you use between the Class E portion of the amplifier and the output port and found it interesting. I would be curious to know the thought that goes behind it.

I had good success with this topology in terms of spurious responses, but I found that a large amount of tuning was necessary before I could get both decent tuning out of it along with low spurious emissions.

Now, on to your comments about Class D vs. Class E:

First of all, a Class E amplifier is limited as to its maximum frequency of operation according to the formula:

F(max) = Pout / (Cp * PI * VDD^2)

Where Cp is the parellel capacitance. If the transistor has a parallel capacitance greater than Cp, then you cannot use it at that frequency.

Now I should have been clear earlier that I am talking about CMCD (current mode Class D) which I believe was first proposed in 2001 by Kobayashi, Hinrichs, and Asbeck, so it is relatively new.

One of the advantages the CMCD has over the conventional Class D design is that the device capacitance is absorbed in the tank circuit. So any losses associated with the discharge of this capacitance is eliminated. As a result, the maximum frequency of operation of a CM Class D is much higher than it would be for a Class E using the same transistor. Furthermore, the maximum theoretical efficiency of this design is 100%, just as it is for a Class E amplifier.

Coupled with the insensitivity to load angle and lower peak drain/collector voltages, it seems like there is little reason to continue using the Class E topology.

Thanks for the suggestion regarding the clamping diodes on the drain. I will examine the waveforms there for any spikes like that which you menti
6987 2011-12-09 10:14:14 jasonb1963 Re: Class D RF Amplifiers
Hi Randy,

Thanks for the article reference. I read them some time ago when I was working with Class E amplifiers. Paul (NA5N) is talking about Voltage Mode Class D amplifiers, and I am talking about Current Mode Class D amplifiers -- they are different beasts. Again, my fault for not clearly stating this in my original posting.

Jason


6988 2011-12-09 10:27:27 jasonb1963 Re: Class D RF Amplifiers
Hi Dave,

One more thing I forgot to add in my previous reply. You state that the IRF510 and 2N7000 are good to at least 10M. This is partially at odds with what I have been able to achieve. I have found the IRF510 to be usable (at the 5W level) for class E operation only on 30M and below. The 2N7000 seems to work at 10M, though the efficiency is definitely down somewhat.

In Dan Tayloe's article on Class E amplifiers (http://www.norcalqrp.org/files/Class_E_Amplifiers.pdf), he states that the IRF510 makes a poor choice for a Class E design due to the large input capacitance, and relatively long turn on and turn off times. This has been my experience as well -- I have never been able to achieve much efficiency out of them
6989 2011-12-09 14:00:49 kb1gmx Re: Class D RF Amplifiers
6990 2011-12-09 15:50:02 jasonb1963 Re: Class D RF Amplifiers
Hi Allison,

6991 2011-12-09 16:55:12 ehydra Re: Class D RF Amplifiers
jasonb1963 schrieb:
> Now I should have been clear earlier that I am talking about CMCD (current mode Class D) which I believe was first proposed in 2001 by Kobayashi, Hinrichs, and Asbeck, so it is relatively new.

I googled for it and the very first papers look like classic Royer
converter. Or is there any difference I'm not aware of?

Mentioned 74AC gates won't work in this configuration.

- Henry


--
ehydra.dyndns.info
6992 2011-12-09 17:48:35 dixonglennb Re: Class D RF Amplifiers
Hi Jason, everyone,

I posted early this morning on this topic but it didn't show up for some reason. Apologies if the earlier one pops up.

My use of the class D nomenclature seems to be backwards from yours. I consider the topology that produces a squarewave drain voltage the 'voltage' mode, and the squarewave current waveform the 'current' mode. This because the former is driven with a constant voltage source, and the latter with a constant current source (usually a choke or large inductor). I think this is how Bostain and Raab classify them. The inductor-fed topology is my favorite and is more efficient than the voltage-fed because the drain is at or near zero volts when switching occurs.

At any rate, if you look through files and pics from ac7zn, you should find a design for a BS170 20 meter 10 watt inductor-fed circuit I posted over a year ago. I measured an efficiency of 96% (power out divided by 13.8V power in so does not include drivers).

That circuit has a double-tuned series bandpass filter
6993 2011-12-09 18:50:15 jasonb1963 Re: Class D RF Amplifiers
Hi Glenn,

I agree with your use of the class D nomenclature and I erred in my earlier posting by describing a voltage mode class D amplifier when I intended to describe a current mode class D amplifier.

I see a picture of your amplifier in the pictures section, but nothing in the files section of the group. I did manage to find a schematic of your amplifier at http://znradio.com/ARRLchallenge.htm but I fear it may not be up-to-date as you claim only 75 to 80% efficiency in the text.

I see many similarities in the two designs we're using. I'm driving the BS170s with two pairs of 74AC04 inverters, so that should give identical results to your 74AC86 XOR gates except I'm only presently driving one BS-170 with each pair (may try paralleling them later).

The only significant difference I see is in the drain circuit. I'm using a center tapped transformer to feed both drains (Vdd on the center tap) and the other side of the transformer as the output. I will investigate your arrangement to see if I get a better result with it.

I'm very pleased with the performance of this circuit so far and expect it will replace the class E designs I've used in the past. At this point, I can't think of a single way in which the class E design is superior (except perhaps that it only requires one transistor).

On the subject of class E amplifiers, one of the better designs I've tried comes from the May 1997 issue of QST which I adapted for QRP use. The authors use a clever arrangement of an L match combined with a series tuned circuit which shorts the second harmonic to ground. I found this arrangement produced some of the best efficiencies I've seen on class E amplifiers.

There is a paper by Bill Slade called, "Notes on Designing Class E Amplifiers." On page 4 he points out that class E operation requires the current to lag the voltage by 32.48 degrees. I've seen a number of theories about how to connect filters up to a class E amplifier without disturbing the performance -- from using a diplexer to various circuit topologies. However, I have experimentally verified that designing a filter that has a 0 degree phase delay at the fundamental frequency has the least effect on the efficiency, compared to other approaches. This is what leads me to believe that the load angle is the most important consideration when adding filters to a class E amplifier.

If you have a more recent schematic of your CMCD amplifier, I would love to see the schematic if there are any significant difference between it and the link I cite above. I presume there must be in order for you to improve the efficiency to 96%.

Jason


6994 2011-12-10 00:07:57 dixonglennb Re: Class D RF Amplifiers
Hi Jason,

Sorry, not under ac7zn but "20M xmt.pdf" under dixonglennb. This schematic matches the image you found.


The website schematic is indeed old and not as efficient.

Glenn


6995 2011-12-10 04:30:40 nm0s_qrp Re: Class D RF Amplifiers
A point to consider: In Class-D (including CMCD) and Class-E, the transistors are utilized as switches. Consequently, the drive requirements for a given device are essentially identical for Class-E as with Class-D. If anything, because Class-D requires creation af many harmonics at the drain to attain efficiency, the drive for it is more critical than for Class-E. Class-E is happy with a sine-wave drive and a low-powered resonant drive circuit.

The output loading phase angle requirements for class-D are actually more restrictive than for Class-E, as its efficiency is more degraded by a mistuned load.

The current mode Class-D has a lot of advantages over the more conventional voltage-mode. It also has many similarities with Class-E - higher efficiency, and narrower bandwidth. Depending on tuning, you can make it a push-pull Class-E, which I did for my ARRL Homebrew Challenge amp last year.

I built a 50 MHz, 5W Class-E amp using an IRF510 a couple years back, which had about 85% drain efficiency, but only 6dB gain. The CB guys use IRF520s in 11M amps all the time.

73 Dave NM0S
6996 2011-12-10 06:57:49 kb1gmx Re: Class D RF Amplifiers
6997 2011-12-10 07:00:29 kb1gmx Re: Class D RF Amplifiers
6998 2011-12-10 08:10:02 jasonb1963 Re: Class D RF Amplifiers
Hi Dave,

6999 2011-12-10 09:45:49 ehydra Re: Class D RF Amplifiers
kb1gmx schrieb:
> The effective impedance is 13 ohms but you will exceed the maximum current before you can drive that low an impedance. Also you may
> want more than 5V drive as some fets 5V is barely enough.

Yes. Consider:
1. The load current is not stationary.
2. FETs work as constant current source if 'overloaded'.

I model this all the day with SPICE.


But what is the TC1214N ? Is this a typo? Cannot find it. Something
maybe from Microchip.


- Henry


--
ehydra.dyndns.info
7002 2011-12-10 19:07:36 ehydra Re: Class D RF Amplifiers
> But what is the TC1214N ? Is this a typo? Cannot find it. Something
> maybe from Microchip.

Looks like TC1412N

http://www.microchip.com/TC1412N

- Henry


--
ehydra.dyndns.info
7003 2011-12-10 19:41:47 William Carver Re: Class D RF Amplifiers
I have some of the 9A versions of these for a 3 KW switching supply
project, but was looking of them for class E driver application, too.
Certainly good for 160m, and 80m, too I betcha. Henry have you tried
them higher than that?
W7AAZ



On Sat, 2011-12-10 at 21:13 +0100, ehydra wrote:
>
>
> > But what is the TC1214N ? Is this a typo? Cannot find it. Something
> > maybe from Microchip.
>
> Looks like TC1412N
>
> http://www.microchip.com/TC1412N
>
> - Henry
>
> --
> ehydra.dyndns.info
>
>
>
>
7006 2011-12-11 09:02:51 kb1gmx Re: Class D RF Amplifiers
7007 2011-12-11 17:53:58 ehydra Re: Class D RF Amplifiers
I never used the 1412. Most MOSFET driver will go a little over 1MHz and
then suffer from self-heating.

The AAT4900 goes till 10MHz. Almost ideal for class-D.

- Henry


William Carver schrieb:
> I have some of the 9A versions of these for a 3 KW switching supply
> project, but was looking of them for class E driver application, too.
> Certainly good for 160m, and 80m, too I betcha. Henry have you tried
> them higher than that?
> W7AAZ

--
ehydra.dyndns.info
7008 2011-12-11 17:54:16 ehydra Re: Class D RF Amplifiers
kb1gmx schrieb:
>>> The effective impedance is 13 ohms but you will exceed the maximum current before you can drive that low an impedance. Also you may
>>> want more than 5V drive as some fets 5V is barely enough.
>
> first 13 shoulld have been 30, price of having the cat help..

Not 30 ohms! Look here by example:
http://www.sysacom.ca/media/pdf/AN200805.pdf

30 ohms is a good value for 74HC.


>
> That refers to the surge impedance of 74ac gate outouts. the problem is they do not like reactive loads and are peak current limited. They
> make cheap but only fair drivers.
>
>> Yes. Consider:
>> 1. The load current is not stationary.
>> 2. FETs work as constant current source if 'overloaded'.
>>
>> I model this all the day with SPICE.
>>
>
> Same here, but the reference is to the 74ac gate as a driver
> not the IRF510. Those devices (74ACxx) use small geometry
> devices and are hard to characterize.

IBIS all what you need.

The IRF510 is a SMPS switcher MOSFET for the 100KHz region. Never made
for 1MHz+.


> Yes, typo. TC1412N. its a 2+A driver
>
> Used in the SWL (k1swl) Dave's Retro-75 QRP AM trx. That has a AM drain modulated class E final.

What is the speed available? The datasheet doesn't show important data.

IXYS and Intersil make the fastet MOSFET drivers available if one looks
for IC designs. Both are not cheap.


- Henry

--
ehydra.dyndns.info
7009 2011-12-12 15:37:17 kb1gmx Re: Class D RF Amplifiers