Forums

High speed (230K) RS232 drivers?

Started by linnix December 10, 2011
In article <14209592-a566-4454-832d-3da56fc7f877
@v6g2000yqv.googlegroups.com>, me@linnix.info-for.us says...
> > On Dec 10, 5:37&#2013266080;pm, Jamie > <jamie_ka1lpa_not_valid_after_ka1l...@charter.net> wrote: > > linnix wrote: > > > On Dec 10, 3:51 pm, miso <m...@sushi.com> wrote: > > > > >>>For a huge length of connection you should recommend an optical link. > > >>>There are (or at least were) cheap RS232 to optical converters. > > > > >>Note that 232 isn't set up to handle ground offsets that are likely to > > >>occur over long distances. Optical is a great way to get around that > > >>problem. > > > > >>Also there will be some asymmetry in rising and falling slew. > > > > >>Back on the subject of high speed 232, these devices with charge pumps > > >>on-board have jitter. Charge pumps create a lot of noise, and no amount > > >>of bypass can filter the substrate noise. > > > > >>I don't know if anyone specs the jitter, but it is easily measurable. > > >>There are internal limits on the jitter since ATE needs to know when to > > >>stop looking for a transition, but the limits are huge. Doing quick > > >>tests with a storage scope, you could hit the microsecond window > > >>rapidly. Ian communication, you are sending bits forever, so someday you > > >>will hit the perfect storm of jitter. Maybe 3uS for worst case margin. > > >>Now you can see how using a charge pump based 232 isn't all that robust > > >>for high speed 232. > > > > > So, would it help with bigger caps? &#2013266080;Are there alternative to charge > > > pumps? > > > > YOu need to put a scope on the other end. > > > > &#2013266080; Personally, I would make them use a twisted pair converter between > > long runs. That seems to work very nicely. > > > > We don't need a long run. It can be very short, just a few inches.
Even if "just a few inches" you have both ends to worry about. If does not leave the box why RS232 for "just a few inches" just put bus driver and receiver. Even simple picogate dedicated TTL inverter would do the job for short distance. If leave the box it may not be just the driver, but shitty cable or poor receiver, or both of those and transmitter. For each signal run both ways. -- Paul Carpenter | paul@pcserviceselectronics.co.uk <http://www.pcserviceselectronics.co.uk/> PC Services <http://www.pcserviceselectronics.co.uk/fonts/> Timing Diagram Font <http://www.gnuh8.org.uk/> GNU H8 - compiler & Renesas H8/H8S/H8 Tiny <http://www.badweb.org.uk/> For those web sites you hate
On Dec 10, 6:31=A0pm, Paul <p...@pcserviceselectronics.co.uk> wrote:
> In article <14209592-a566-4454-832d-3da56fc7f877 > @v6g2000yqv.googlegroups.com>, m...@linnix.info-for.us says... > > > > > > > > > > > > > On Dec 10, 5:37=A0pm, Jamie > > <jamie_ka1lpa_not_valid_after_ka1l...@charter.net> wrote: > > > linnix wrote: > > > > On Dec 10, 3:51 pm, miso <m...@sushi.com> wrote: > > > > >>>For a huge length of connection you should recommend an optical li=
nk.
> > > >>>There are (or at least were) cheap RS232 to optical converters. > > > > >>Note that 232 isn't set up to handle ground offsets that are likely=
to
> > > >>occur over long distances. Optical is a great way to get around tha=
t
> > > >>problem. > > > > >>Also there will be some asymmetry in rising and falling slew. > > > > >>Back on the subject of high speed 232, these devices with charge pu=
mps
> > > >>on-board have jitter. Charge pumps create a lot of noise, and no am=
ount
> > > >>of bypass can filter the substrate noise. > > > > >>I don't know if anyone specs the jitter, but it is easily measurabl=
e.
> > > >>There are internal limits on the jitter since ATE needs to know whe=
n to
> > > >>stop looking for a transition, but the limits are huge. Doing quick > > > >>tests with a storage scope, you could hit the microsecond window > > > >>rapidly. Ian communication, you are sending bits forever, so someda=
y you
> > > >>will hit the perfect storm of jitter. Maybe 3uS for worst case marg=
in.
> > > >>Now you can see how using a charge pump based 232 isn't all that ro=
bust
> > > >>for high speed 232. > > > > > So, would it help with bigger caps? =A0Are there alternative to cha=
rge
> > > > pumps? > > > > YOu need to put a scope on the other end. > > > > =A0 Personally, I would make them use a twisted pair converter betwee=
n
> > > long runs. That seems to work very nicely. > > > We don't need a long run. =A0It can be very short, just a few inches. > > Even if "just a few inches" you have both ends to worry about. > > If does not leave the box why RS232 for "just a few inches" just > put bus driver and receiver. Even simple picogate dedicated TTL > inverter would do the job for short distance. > > If leave the box it may not be just the driver, but shitty cable or > poor receiver, or both of those and transmitter. For each signal run > both ways. >
Our small box can sit right next to the big box. We might be able to specify and supply the cable if necessary, but i have to check with the captain.
On Dec 11, 6:16=A0am, linnix <m...@linnix.info-for.us> wrote:
> On Dec 10, 6:31=A0pm, Paul <p...@pcserviceselectronics.co.uk> wrote: > > > > > > > In article <14209592-a566-4454-832d-3da56fc7f877 > > @v6g2000yqv.googlegroups.com>, m...@linnix.info-for.us says... > > > > On Dec 10, 5:37=A0pm, Jamie > > > <jamie_ka1lpa_not_valid_after_ka1l...@charter.net> wrote: > > > > linnix wrote: > > > > > On Dec 10, 3:51 pm, miso <m...@sushi.com> wrote: > > > > > >>>For a huge length of connection you should recommend an optical =
link.
> > > > >>>There are (or at least were) cheap RS232 to optical converters. > > > > > >>Note that 232 isn't set up to handle ground offsets that are like=
ly to
> > > > >>occur over long distances. Optical is a great way to get around t=
hat
> > > > >>problem. > > > > > >>Also there will be some asymmetry in rising and falling slew. > > > > > >>Back on the subject of high speed 232, these devices with charge =
pumps
> > > > >>on-board have jitter. Charge pumps create a lot of noise, and no =
amount
> > > > >>of bypass can filter the substrate noise. > > > > > >>I don't know if anyone specs the jitter, but it is easily measura=
ble.
> > > > >>There are internal limits on the jitter since ATE needs to know w=
hen to
> > > > >>stop looking for a transition, but the limits are huge. Doing qui=
ck
> > > > >>tests with a storage scope, you could hit the microsecond window > > > > >>rapidly. Ian communication, you are sending bits forever, so some=
day you
> > > > >>will hit the perfect storm of jitter. Maybe 3uS for worst case ma=
rgin.
> > > > >>Now you can see how using a charge pump based 232 isn't all that =
robust
> > > > >>for high speed 232. > > > > > > So, would it help with bigger caps? =A0Are there alternative to c=
harge
> > > > > pumps? > > > > > YOu need to put a scope on the other end. > > > > > =A0 Personally, I would make them use a twisted pair converter betw=
een
> > > > long runs. That seems to work very nicely. > > > > We don't need a long run. =A0It can be very short, just a few inches. > > > Even if "just a few inches" you have both ends to worry about. > > > If does not leave the box why RS232 for "just a few inches" just > > put bus driver and receiver. Even simple picogate dedicated TTL > > inverter would do the job for short distance. > > > If leave the box it may not be just the driver, but shitty cable or > > poor receiver, or both of those and transmitter. For each signal run > > both ways. > > Our small box can sit right next to the big box. =A0We might be able to > specify and supply the cable if necessary, but i have to check with > the captain.- Hide quoted text - > > - Show quoted text -
We use a PNP (BC807), emitter to +5, collector to -5v via 4k7 and to output via 100 ohms. Base - emitter resistor of 3k3, drive from processor via 10k. Handles 3v3 processors OK and runs in excess of 115200 over 5 metres. Receiver is an NPN (BC817).
In article <A9GdndouELiujXnTnZ2dnUVZ_v6dnZ2d@giganews.com>, 
nospam@nowhere.com says...
> > linnix wrote: > > > We don't need a long run. It can be very short, just a few inches. > > Use general purpose CMOS inverters as drivers. All modern RS-232 ICs > have threshold at about +1.5V, so they work fine with CMOS levels. The > only problem could be some transceivers detecting low input as "cable > disconnected" condition and falling into sleep mode. >
If you cannot change other end, which transmits +/-9-12V levels make sure yout receiver can cope. Another thing that has caught some people out, is check the connectors. I have seen a few applications caught out by the use of Filtered D types for EMC, limit the bandwidth or add strange loads to signals. -- Paul Carpenter | paul@pcserviceselectronics.co.uk <http://www.pcserviceselectronics.co.uk/> PC Services <http://www.pcserviceselectronics.co.uk/fonts/> Timing Diagram Font <http://www.gnuh8.org.uk/> GNU H8 - compiler & Renesas H8/H8S/H8 Tiny <http://www.badweb.org.uk/> For those web sites you hate
On a sunny day (Sun, 11 Dec 2011 00:54:52 -0800 (PST)) it happened Rocky
<robertgush@gmail.com> wrote in
<75df1ce1-edb9-43e8-99a3-a47ab9989fa5@v24g2000prn.googlegroups.com>:

>We use a PNP (BC807), emitter to +5, collector to -5v via 4k7 and to >output via 100 ohms. Base - emitter resistor of 3k3, drive from >processor via 10k. Handles 3v3 processors OK and runs in excess of >115200 over 5 metres. Receiver is an NPN (BC817).
I have been using this: http://panteltje.com/panteltje/wap54g/max_pantel.txt Its gets it power from the DTR and RTS lines, the communication software needs to set those to + and -. also you need 2 extra wires.... Works OK at medium speed over a few meters, have not tried 115200 or higher.
On a sunny day (Sat, 10 Dec 2011 20:06:51 -0600) it happened Vladimir
Vassilevsky <nospam@nowhere.com> wrote in
<A9GdndouELiujXnTnZ2dnUVZ_v6dnZ2d@giganews.com>:

> > >linnix wrote: > >> We don't need a long run. It can be very short, just a few inches. > >Use general purpose CMOS inverters as drivers. All modern RS-232 ICs >have threshold at about +1.5V, so they work fine with CMOS levels. The >only problem could be some transceivers detecting low input as "cable >disconnected" condition and falling into sleep mode.
Beware of latchup if you plug an external source into a CMOS gate! Not a good solution.

Jan Panteltje wrote:

> On a sunny day (Sat, 10 Dec 2011 20:06:51 -0600) it happened Vladimir > Vassilevsky <nospam@nowhere.com> wrote
>>Use general purpose CMOS inverters as drivers. All modern RS-232 ICs >>have threshold at about +1.5V, so they work fine with CMOS levels. The >>only problem could be some transceivers detecting low input as "cable >>disconnected" condition and falling into sleep mode. > > Beware of latchup if you plug an external source into a CMOS gate! > Not a good solution.
What an incredible problem. Add two resistors and BAV99 to protect the input and one more resistor to protect the output. Done that 1000 times, works fine for tabletop distances except for the rare case of "cable disconnect" sensitive devices. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
On 11 Dec., 10:09, Paul <p...@pcserviceselectronics.co.uk> wrote:
> In article <A9GdndouELiujXnTnZ2dnUVZ_v6dn...@giganews.com>, > nos...@nowhere.com says... > > > > > linnix wrote: > > > > We don't need a long run. =A0It can be very short, just a few inches. > > > Use general purpose CMOS inverters as drivers. All modern RS-232 ICs > > have threshold at about +1.5V, so they work fine with CMOS levels. The > > only problem could be some transceivers detecting low input as "cable > > disconnected" condition and falling into sleep mode. > > If you cannot change other end, which transmits +/-9-12V levels make > sure yout receiver can cope. > > Another thing that has caught some people out, is check the connectors. > I have seen a few applications caught out by the use of Filtered D types > for EMC, limit the bandwidth or add strange loads to signals. >
good point about the filtered dsubs, it has got to be something like that I don't remember seeing much problem running 921kbit rs232, other than windows having trouble keeping up -Lasse
On a sunny day (Sun, 11 Dec 2011 11:26:22 -0600) it happened Vladimir
Vassilevsky <nospam@nowhere.com> wrote in
<3-qdnWwFkPouennTnZ2dnUVZ_sadnZ2d@giganews.com>:

> > >Jan Panteltje wrote: > >> On a sunny day (Sat, 10 Dec 2011 20:06:51 -0600) it happened Vladimir >> Vassilevsky <nospam@nowhere.com> wrote > >>>Use general purpose CMOS inverters as drivers. All modern RS-232 ICs >>>have threshold at about +1.5V, so they work fine with CMOS levels. The >>>only problem could be some transceivers detecting low input as "cable >>>disconnected" condition and falling into sleep mode. >> >> Beware of latchup if you plug an external source into a CMOS gate! >> Not a good solution. > >What an incredible problem. Add two resistors and BAV99 to protect the >input and one more resistor to protect the output. >Done that 1000 times, works fine for tabletop distances except for the >rare case of "cable disconnect" sensitive devices.
It is npt a problem if you do that, but you left it out. Many peopla read this, and could get the wrong idea.
>Vladimir Vassilevsky >DSP and Mixed Signal Design Consultant >http://www.abvolt.com >
On 12/10/2011 5:36 PM, linnix wrote:
> On Dec 10, 4:43 pm, linnix<m...@linnix.info-for.us> wrote: >> On Dec 10, 3:51 pm, miso<m...@sushi.com> wrote: >> >> >> >> >> >> >> >> >> >>>> For a huge length of connection you should recommend an optical link. >>>> There are (or at least were) cheap RS232 to optical converters. >> >>> Note that 232 isn't set up to handle ground offsets that are likely to >>> occur over long distances. Optical is a great way to get around that >>> problem. >> >>> Also there will be some asymmetry in rising and falling slew. >> >>> Back on the subject of high speed 232, these devices with charge pumps >>> on-board have jitter. Charge pumps create a lot of noise, and no amount >>> of bypass can filter the substrate noise. >> >>> I don't know if anyone specs the jitter, but it is easily measurable. >>> There are internal limits on the jitter since ATE needs to know when to >>> stop looking for a transition, but the limits are huge. Doing quick >>> tests with a storage scope, you could hit the microsecond window >>> rapidly. Ian communication, you are sending bits forever, so someday you >>> will hit the perfect storm of jitter. Maybe 3uS for worst case margin. >>> Now you can see how using a charge pump based 232 isn't all that robust >>> for high speed 232. >> >> So, would it help with bigger caps? Are there alternative to charge >> pumps? > > Well, you did say the noise is internal to the chip. If that's the > case, we might have to try chips w/o pumps and w external dc-dc > converters.
Certainly using an external DC/DC will help. However, I don't know if the old metal gate parts will work at high speed. You can measure the jitter easily with a storage scope or DSO. Test transmitter and receiver. Drive the transmitter input from a generator. Trigger off the generator. Store samples of the transmitter output. You will get a fat band at the edges due to jitter. You can either drive a receiver with the transmitter output or signal generator. Just make sure you are triggered on the receiver input. Look for fat traces on the receiver output. I assume with your own DCDC, you can use 1488 and 1489. But I don't know if they run at high speed either. You'd have to look that up.